In cases of statutory construction, this Court’s authority is limited. If the statutory language and legislative intent are plain, the judicial inquiry is at an end. Under our jurisprudence, it is presumed that ill-considered or unwise legislation will be corrected through the democratic process; a court is not permitted to distort a statute’s meaning in order to make it conform with the Justices’ own views of sound social policy. See TV A v. Hill, 437 U. S. 153 (1978).
Today’s decision flagrantly disregards these restrictions on judicial authority. The plurality ignores the plain meaning of the Occupational Safety and Health Act of 1970 in order to bring the authority of the Secretary of Labor in line with the plurality’s own views of proper regulatory policy. The unfortunate consequence is that the Federal Government’s efforts to protect American workers from cancer and other crippling diseases may be substantially impaired.
*689The first sentence of § 6 (b)(5) of the Act provides:
“The Secretary, in promulgating standards dealing with toxic materials or harmful physical agents under this subsection, shall set the standard which most adequately assures, to the extent feasible, on the basis of the best available evidence, that no employee will suffer material impairment of health or functional capacity even if such employee has regular exposure to the hazard dealt with by such standard for the period of his working life.” 29 U. S. C. §655 (b)(5).
In this case the Secretary of Labor found, on the basis of substantial evidence, that (1) exposure to benzene creates a risk of cancer, chromosomal damage, and a variety of nonmalignant but potentially fatal blood disorders, even at the level of 1 ppm; (2) no safe level of exposure has been shown; (3) benefits in the form of saved lives would be derived from the permanent standard; (4) the number of lives that would be saved could turn out to be either substantial or relatively small; (5) under the present state of scientific knowledge, it is impossible to calculate even in a rough way the number of lives that would be saved, at least without making assumptions that would appear absurd to much of the medical community; and (6) the standard would not materially harm the financial condition of the covered industries. The Court does not set aside any of these findings. Thus, it could not be plainer that the Secretary’s decision was fully in accord with his statutory mandate “most adequately [to] assur[e] . . . that no employee will suffer material impairment of health or functional capacity. . . .”
The plurality’s conclusion to the contrary is based on its interpretation of 29 TJ. S. C. § 652 (8), which defines an occupational safety and health standard as one “which requires conditions . . . reasonably necessary or appropriate to provide safe or healthful employment. . . .” According to the plurality, a standard is not “reasonably necessary or appropriate” *690unless the Secretary is able to show that it is “at least more likely than not,” ante, at 653, that the risk he seeks to regulate is a “significant” one. Ibid. Nothing in the statute’s language or legislative history, however, indicates that the “reasonably necessary or appropriate” language should be given this meaning. Indeed, both demonstrate that the plurality’s standard bears no connection with the acts or intentions of Congress and is based only on the plurality’s solicitude for the welfare of regulated industries. And the plurality uses this standard to’ evaluate not the agency’s decision in this case, but a strawman of its own creation.
Unlike the plurality, I do not purport to know whether the actions taken by Congress and its delegates to ensure occupational safety represent sound or unsound regulatory policy. The critical problem in cases like the ones at bar is scientific uncertainty. While science has determined that exposure to benzene at levels above 1 ppm creates a definite risk of health impairment, the magnitude of the risk cannot be quantified at the present time. The risk at issue has hardly been shown to be insignificant; indeed, future research may reveal that the risk is in fact considerable. But the existing evidence may frequently be inadequate to enable the Secretary to make the threshold finding of “significance” that the Court requires today. If so, the consequence of the plurality’s approach would be to subject American workers to a continuing risk of cancer and other fatal diseases, and to render the Federal Government powerless to take protective action on their behalf. Such an approach would place the burden of medical uncertainty squarely on the shoulders of the American worker, the intended beneficiary of the Occupational Safety and Health Act. It is fortunate indeed that at least a majority of the Justices reject the view that the Secretary is prevented from taking regulatory action when the magnitude of a health risk cannot be quantified on the basis of current techniques. See ante, at 666 (Powell, J., concurring in part *691and concurring in judgment); see also ante, at 656, and n. 63 (plurality opinion).
Because today’s holding has no basis in the Act, and because the Court has no authority to impose its own regulatory policies on the Nation, I dissent.
I
Congress enacted the Occupational Safety and Health Act as a response to what was characterized as “the grim history of our failure to heed the occupational health needs of our workers.” 1 The failure of voluntary action and legislation at the state level, see S. Rep. No. 91-1282, p. 4 (1970), Leg. Hist. 144, had resulted in a “bleak” and “worsening” 2 situation in which 14,500 persons had died annually as a result of conditions in the workplace. In the four years preceding the Act’s passage, more Americans were killed in the workplace than in the contemporaneous Vietnam War, S. Rep. No. 91-1283, at 2, Leg. Hist. 142. The Act was designed as “a safety bill of rights for close to 60 million workers.” 3 Its stated purpose is “to assure so far as possible every working man and woman in the Nation safe and healthful working conditions and to preserve our human resources.” 29 U. S. C. § 651 (b). See Atlas Roofing Co. v. Occupational Safety and Health Review Comm’n, 430 U. S. 442, 444-445 (1977).
The Act is enforced primarily through two provisions. First, a “general duty” is imposed upon employers to furnish employment and places of employment “free from recognized hazards that are causing or are likely to cause death or serious physical harm. . . .” 29 U. S. C. § 654 (a)(1). Second, the Secretary of Labor is authorized to set “occupational safety *692and health standards,” defined as standards requiring “conditions, or the adoption or use of one or more practices, means, methods, operations, or processes, reasonably necessary or appropriate to provide safe or healthful employment and places of employment.” 29 U. S. C. § 652 (8).
The legislative history of the Act reveals Congress’ particular concern for health hazards of “unprecedented complexity” that had resulted from chemicals whose toxic effects “are only now being discovered.” S. Rep. No. 91-1282, supra, at 2, Leg. Hist. 142. “Recent scientific knowledge points to hitherto unsuspected cause-and-effect relationships between occupational exposures and many of the so-called chronic diseases — cancer, respiratory ailments, allergies, heart disease, and others.” Ibid. Members of Congress made repeated references to the dangers posed by carcinogens and to the defects in our knowledge of their operation and effect.4 One of the primary purposes of the Act was to ensure regulation of these “insidious ‘silent’ killers.” 5
This special concern led to the enactment of the first sentence of 29 U. S. C. §655 (b)(5), which, as noted above, provides:
“The Secretary, in promulgating standards dealing with toxic materials or harmful physical agents under this subsection, shall set the standard which most adequately assures, to the extent feasible, on the basis of the best available evidence, that no employee will suffer material impairment of health or functional capacity even if such employee has regular exposure to the hazard dealt with by such standard for the period of his working life.”
This directive is designed to implement three legislative pur*693poses. First, Congress recognized that there may be substances that become dangerous only upon repeated or frequent exposure.6 The Secretary was therefore required to provide protection even from substances that would cause material impairment only upon exposure occurring throughout an employee’s working life. Second, the requirement that the Secretary act on the basis of “the best available evidence” was intended to ensure that the standard-setting process would not be destroyed by the uncertainty of scientific views. Recognizing that existing knowledge may be inadequate, Congress did not require the Secretary to wait until definitive information could be obtained. Thus “it is not intended that the Secretary be paralyzed by debate surrounding diverse medical opinions.” H. R. Rep. No. 91-1291, p. 18 (1970), Leg. Hist. 848. Third, Congress’ special concern for the “silent killers” was felt to justify an especially strong directive to the Secretary in the standard-setting process. 116 Cong. Rec. 37622 (1970), Leg. Hist. 502 (Sen. Dominick).
The authority conferred by §655 (b)(5), however, is not absolute. The subsection itself contains two primary limitations. The requirement of “material” impairment was designed to prohibit the Secretary from regulating substances that create a trivial hazard to affected employees.7 Moreover, all standards promulgated under the subsection must be “feasible.” During the floor debates Congress expressed concern that a prior version of the bill, not clearly embodying the feasibility requirement, would require the Secretary to close down whole industries in order to eliminate risks of impairment. This standard was criticized as unrealistic.8 *694The feasibility requirement was imposed as an affirmative limit on the standard-setting power.
The remainder of § 655 (b)(5), applicable to all safety and health standards, requires the Secretary to base his standards “upon research, demonstrations, experiments, and such other information as may be appropriate.” In setting standards, the Secretary is directed to consider “the attainment of the highest degree of health and safety protection for the employee” and also “the latest available scientific data in the field, the feasibility of the standards, and experience gained under this and other health and safety laws.”
The Act makes provision for judicial review of occupational safety and health standards promulgated pursuant to § 655 (b)(5). The reviewing court must uphold the Secretary’s *695determinations if they are supported by “substantial evidence in the record considered as a whole.” 29 TJ. S. C. § 655 (f). It is to that evidence that I now turn.
II
The plurality’s discussion of the record in this case is both extraordinarily arrogant and extraordinarily unfair. It is arrogant because the plurality presumes to make its own factual findings with respect to a variety of disputed issues relating to carcinogen regulation. See, e. g., ante, at 656-657, and n. 64. It should not be necessary to remind the Members of this Court that they were not appointed to undertake independent review of adequately supported scientific findings made by a technically expert agency.9 And the plurality’s discussion is unfair because its characterization of the Secretary’s report bears practically no resemblance to what the Secretary actually did in this case. Contrary to the plurality’s suggestion, the Secretary did not rely blindly on some Draconian carcinogen “policy.” See ante, at 624-625, 635-636. If he had, it would have been sufficient for him to have observed that *696benzene is a carcinogen, a proposition that respondents do not dispute. Instead, the Secretary gathered over 50 volumes of exhibits and testimony and offered a detailed and evenhanded discussion of the relationship between exposure to benzene at all recorded exposure levels and chromosomal damage, aplastic anemia, and leukemia. In that discussion he evaluated, and took seriously, respondents’ evidence of a safe exposure level. See also ante, at 666 (Powell, J., concurring in part and in judgment).
The hearings on the proposed standard were extensive, encompassing 17 days from July 19 through August 10, 1977. The 95 witnesses included epidemiologists, toxicologists, physicians, political economists, industry representatives, and members of the affected work force. Witnesses were subjected to exhaustive questioning by representatives from a variety of interested groups and organizations.
Three basic positions were presented at the hearings. The first position was that the proposed 1 ppm standard was necessary because exposure to benzene would cause material impairment of the health of workers no matter how low the exposure level. Some direct evidence indicated that exposure to benzene had caused chromosomal damage, blood disorders, and leukemia at or below the 10 ppm level itself. More important, it was suggested that the recorded effects of benzene at higher levels required an inference that leukemia and other disorders would result at levels of 1 ppm and lower, especially after the prolonged exposure typical in industrial settings. Therefore, the standard should be set at the lowest feasible level, which was 1 ppm.
The second position was that a 1 ppm exposure level would itself pose an unwarranted threat to employee health and safety and that the available evidence necessitated a significantly lower level. An exposure limit below 1 ppm, it was argued, would be feasible. There were suggestions that benzene was gradually being replaced in many of the affected *697industries and that most companies were already operating at or below the 1 ppm level.
The third position was that the 1971 standard should be retained. Proponents of this position suggested that evidence linking low levels of benzene exposure to leukemia was uncertain, that the current exposure limit was sufficiently safe, and that the benefits of the proposed standard would be insufficient to justify the standard’s costs. In addition, there was testimony that the expenses required by the proposed standard would be prohibitive.
The regulations announcing the permanent standard for benzene are accompanied by an extensive statement of reasons summarizing and evaluating the results of the hearings. The Secretary found that the evidence showed that exposure to benzene causes chromosomal damage, a variety of nonmalignant blood disorders, and leukemia. 43 Fed. Reg. 5921 (1978). He concluded that low concentrations imposed a hazard that was sufficiently grave to call for regulatory action under the Act.
Evidence of deleterious effects. The Secretary referred to studies which conclusively demonstrated that benzene could damage chromosomes in blood-forming cells. Id., at 5932. There was testimony suggesting a causal relationship between chromosomal damage and leukemia, although it could not be determined whether and to what extent such damage would impair health. Id., at 5933.10 Some studies had suggested chromosomal damage at exposure levels of 10-25 ppm and lower.11 No quantitative dose-response curve, showing the relationship between exposure levels and incidence of chromosomal damage, could yet be established. Id., at 5933-5934. The evidence of chromosomal damage was, in the Secretary’s view, a cause for “serious concern.” Id., at 5933.
The most common effect of benzene exposure was a de*698crease in the levels of blood platelets and red and white blood cells. If sufficiently severe, the result could be pancytopenia or aplastic anemia, noncancerous but potentially fatal diseases. There was testimony that some of the nonmalignant blood disorders caused by benzene exposure could progress to, or represented, a preleukemic stage which might eventually evolve into a frank leukemia. Id., at 5922.12
Considerable evidence showed an association between benzene and nonmalignant blood disorders at low exposure levels. Such an association had been established in one study in which the levels frequently ranged from zero to 25 ppm with some concentrations above 100 ppm, ibid.; in another they ranged from 5 to 30 ppm, id., at 5923. Because of the absence of adequate data, a dose-response curve showing the relationship between benzene exposure and blood disorders could not be constructed. There was considerable testimony, however, that such disorders had resulted from exposure to benzene at or near the current level of 10 ppm and lower.13 The Secretary concluded that the current standard did not provide adequate protection. He observed that a “safety factor” of 10 to 100 was generally used to discount the level at which a causal connection had been found in existing studies.14 Under this approach, he concluded that, quite apart from any leukemia risk, the permissible exposure limit should be set at a level considerably lower than 10 ppm.
Finally, there was substantial evidence that exposure to benzene caused leukemia. The Secretary concluded that the evidence established that benzene was a carcinogen. A causal relationship between benzene and leukemia was first reported in France in 1897, and since that time similar results had been found in a number of countries, including Italy, Turkey, Japan, Switzerland, the Soviet Union, and the United *699States. The latest study, undertaken by the National Institute for Occupational Safety and Health (NIOSH) in the 1970’s, reported a fivefold excess over the normal incidence of leukemia among workers exposed to benzene at industrial plants in Ohio. There was testimony that this study seriously understated the risk.15
The Secretary reviewed certain studies suggesting that low exposure levels of 10 ppm and more did not cause any excess incidence of leukemia. Those studies, he suggested, suffered from severe methodological defects, as their authors frankly acknowledged.16 Finally, the Secretary discussed a study suggesting a statistically significant excess in leukemia at levels of 2 to 9 ppm. Ibid.17 He found that, despite certain deficiencies in the study, it should be considered as consistent with other studies demonstrating an excess leukemia risk among employees exposed to benzene. Id., at 5928.
Areas oj uncertainty. The Secretary examined three areas *700of uncertainty that had particular relevance to his decision. First, he pointed to evidence that the latency period for benzene-induced leukemia could range from 2 to over 20 years. Id., at 5930. Since lower exposure levels lead to an increase in the latency period, it would be extremely difficult to obtain evidence showing the dose-response relationship between leu-kemiá and exposure to low levels of benzene. Because there has been no adequate monitoring in the past, it would be practically impossible to determine what the exposure levels were at a time sufficiently distant so that the latency period would have elapsed. The problem was compounded by the difficulty of conducting a suitable study. Because exposure levels approaching 10 ppm had been required only recently, direct evidence showing the relationship between leukemia and exposure levels between 1 and 10 ppm would be unavailable in the foreseeable future.
Second, the Secretary observed that individuals have differences in their susceptibility to leukemia. Ibid. Among those exposed to benzene was a group of unknown but possibly substantial size having various “predisposing factors” whose members were especially vulnerable to the disease. Id., at 5930, 5946. The permanent standard was designed to minimize the effects of exposure for these susceptible individuals as well as for the relatively insensitive, id., at 5946, and also to facilitate early diagnosis and treatment. Id., at 5930.
The Secretary discussed the contention that a safe level of exposure to benzene had been demonstrated. From the testimony of numerous scientists, he concluded that it had not. Id., at 5932.18 He also found that although no dose-response curve could be plotted, id., at 5946,19 the extent of the risk *701would decline with the exposure level. Ibid.20 Exposure at a level of 1 ppm would therefore be less dangerous than exposure at one of 10 ppm. The Secretary found that the existing evidence justified the conclusion that he should not “wait for answers" while employees continued to be exposed to benzene at hazardous levels.
Finally, the Secretary responded to the argument that the permissible exposure level should be zero or lower than 1 ppm. Id., at 5947.21 Even though many industries had already achieved the 1 ppm level, he found that a lower level would not be feasible. Ibid.
Costs and benefits. The Secretary offered a detailed discussion of the role that economic considerations should play in his determination. He observed that standards must be “feasible,” both economically and technologically. In his view the permanent standard for benzene was feasible under both tests. The economic impact would fall primarily on the more stable industries, such as petroleum refining and petrochemical production. Id., at 5934. These industries would be able readily to absorb the costs or to pass them on to consumers. None of the 20 affected industries, involving 157,000 facilities and 629,000 exposed employees, id., at 5935, would be unable to bear the required expenditures, id., at 5934. He concluded that the compliance costs were “well within the financial capability of the covered industries.” Id., at 5941. An extensive survey of the national economic impact of the standard, undertaken by a private contractor, found first-year operating costs of between $187 and $205 million, recurring annual costs of $34 million, and investment in engineering controls of about $266 million.22 Since respondents have not at*702tacked the Secretary’s basic conclusions as to cost, the Secretary’s extensive discussion need not be summarized here.
Finally, the Secretary discussed the benefits to be derived from the permanent standard. During the hearings, it had been argued that the Secretary should estimate the health benefits of the proposed regulation. To do this he would be required to construct a dose-response curve showing, at least in a rough way, the number of lives that would be saved at each possible exposure level. Without some estimate of benefits, it was argued, the Secretary’s decisionmaking would be defective. During the hearings an industry witness attempted to construct such a dose-response curve. Restricting himself to carcinogenic effects, he estimated that the proposed standard would save two lives every six years and suggested that this relatively minor benefit would not justify the regulation’s costs.
The Secretary rejected the hypothesis that the standard would save only two lives in six years. This estimate, he concluded, was impossible to reconcile with the evidence in the record. Ibid.23 He determined that, because of numer*703ous uncertainties in the existing data, it was impossible to construct a dose-response curve by extrapolating from those data to lower exposure levels.24 More generally, the Secre*704tary observed that it had not been established that there was a safe level of exposure for benzene. Since there was considerable testimony that the risk would decline with the exposure level, id., at 5940, the new standard would save lives. The number of lives saved “may be appreciable,” but there was no way to make a more precise determination.25 The question was “on the frontiers of scientific knowledge.” Ibid.
The Secretary concluded that, in light of the scientific uncertainty, he was not required to calculate benefits more precisely. Id., at 5941. In any event he gave “careful consideration” to the question of whether the admittedly substantial costs were justified in light of the hazards of benzene exposure. He concluded that those costs were “necessary” in order to promote the purposes of the Act.
Ill
A
This is not a case in which the Secretary found, or respondents established, that no benefits would be derived from a permanent standard, or that the likelihood of benefits was insignificant. Nor was it shown that a quantitative estimate of benefits could be made on the basis of “the best available evidence.” Instead, the Secretary concluded that benefits will result, that those benefits “may” be appreciable, but that the dose-response relationship of low levels of benzene *705exposure and leukemia, nonmalignant blood disorders, and chromosomal damage was impossible to determine. The question presented is whether, in these circumstances, the Act permits the Secretary to take regulatory action, or whether he must allow continued exposure until more definitive information becomes available.
As noted above, the Secretary’s determinations must be upheld if supported by “substantial evidence in the record considered as a whole.” 29 U. S. C. § 655 (f). This standard represents a legislative judgment that regulatory action should be subject to review more stringent than the traditional “arbitrary and capricious” standard for informal rulemaking. We have observed that the arbitrary and capricious standard itself contemplates a searching “inquiry into the facts” in order to determine “whether the decision was based on a consideration of the relevant factors and whether there has been a clear error of judgment.” Citizens to Preserve Overton Park v. Volpe, 401 U. S. 402, 416 (1971). Careful performance of this task is especially important when Congress has imposed the comparatively more rigorous “substantial evidence” requirement. As we have emphasized, however, judicial review under the substantial evidence test is ultimately deferential. See, e. g., Richardson v. Perales, 402 U. S. 389, 401 (1971) .; Consolo v. Federal Maritime Comm’n, 383 U. S. 607, 618-621 (1966). The agency’s decision is entitled to the traditional presumption of validity, and the court is not authorized to substitute its judgment for that of the Secretary. If the Secretary has considered the decisional factors and acted in conformance with the statute, his ultimate decision must be given a large measure of respect. Id., at 621.
The plurality is insensitive to three factors which, in my view, make judicial review of occupational safety and health standards under the substantial evidence test particularly difficult. First, the issues often reach a high level of technical complexity. In such circumstances the courts are required to immerse themselves in matters to which they are unaccus*706tomed by training or experience. Second, the factual issues with which the Secretary must deal are frequently not subject to any definitive resolution. Often “the factual finger points, it does not conclude.” Society of Plastics Industry, Inc. v. OSHA, 509 F. 2d 1301, 1308 (CA2) (Clark, J.), cert. denied, 421 U. S. 992 (1975). Causal connections and theoretical extrapolations may be uncertain.' Third, when the question involves determination of the acceptable level of risk, the ultimate decision must necessarily be based on considerations of policy ‘as well as empirically verifiable facts. Factual determinations can at most define the risk in some statistical way; the judgment whether that risk is tolerable cannot be based solely on a resolution of the facts.
The decision to take action in conditions of uncertainty bears little resemblance to the sort of empirically verifiable factual conclusions to which the substantial evidence test is normally applied. Such decisions were not intended to be unreviewable; they too must be scrutinized to ensure that the Secretary has acted reasonably and within the boundaries set by Congress. But a reviewing court must be mindful of the limited nature of its role. See Vermont Yankee Nuclear Power Corp. v. NBDC, 435 U. S. 519 (1978). It must recognize that the ultimate decision cannot be based solely on determinations of fact, and that those factual conclusions that have been reached are ones which the courts are ill-equipped to resolve on their own.
Under this standard of review, the decision to reduce the permissible exposure level to 1 ppm was well within the Secretary’s authority. The Court of Appeals upheld the Secretary’s conclusions that benzene causes leukemia, blood disorders, and chromosomal damage even at low levels, that an exposure level of 10 ppm is more dangerous than one of 1 ppm, and that benefits will result from the proposed standard. It did not set aside his finding that the number of lives that would be saved was not subject to quantification. *707Nor did it question his conclusion that the reduction was “feasible.”
In these circumstances, the Secretary’s decision was reasonable and in full conformance with the statutory language requiring that he “set the standard which most adequately assures, to the extent feasible, on the basis of the best available evidence, that no employee will suffer material impairment of health or functional capacity even if such employee has regular exposure to the hazard dealt with by such standard for the period of his working life.” 29 TJ. S. C. § 655 (b) (5). On this record, the Secretary could conclude that regular exposure above the 1 ppm level would pose a definite risk resulting in material impairment to some indeterminate but possibly substantial number of employees. Studies revealed hundreds of deaths attributable to benzene exposure. Expert after expert testified that no safe level of exposure had been shown and that the extent of the risk declined with the exposure level. There was some direct evidence of incidence of leukemia, nonmalignant blood disorders, and chromosomal damage at exposure levels of 10 ppm and below. Moreover, numerous experts testified that existing evidence required an inference that an exposure level above 1 ppm was hazardous. We have stated that “well-reasoned expert testimony — based on what is known and uncontradicted by empirical evidence— may in and of itself be 'substantial evidence’ when first-hand evidence on the question ... is unavailable.” FPC v. Florida Power & Light Co., 404 U. S. 453, 464-465 (1972). Nothing in the Act purports to prevent the Secretary from acting when definitive information as to the quantity of a standard’s benefits is unavailable.26 Where, as here, the deficiency in *708knowledge relates to the extent of the benefits rather than their existence, I see no reason to hold that the Secretary has exceeded his statutory authority.
B
The plurality avoids this conclusion through reasoning that may charitably be described as obscure. According to the plurality, the definition of occupational safety and health standards as those “reasonably necessary or appropriate to provide safe or healthful . . . working conditions” requires the Secretary to show that it is “more likely than not” that the risk he seeks to regulate is a “significant” one. Ante, at 653. The plurality does not show how this requirement can be plausibly derived from the “reasonably necessary or appropriate” clause. Indeed, the plurality’s reasoning is refuted by the Act’s language, structure, and legislative history, and it is foreclosed by every applicable guide to statutory construction. In short, the plurality’s standard is a fabrication bearing no connection with the acts or intentions of Congress.
At the outset, it is important to observe that “reasonably necessary or appropriate” clauses are routinely inserted in regulatory legislation, and in the past such clauses have uniformly been interpreted as general provisos that regulatory actions must bear a reasonable relation to those statutory purposes set forth in the statute’s substantive provisions. See, e. g., FCC v. National Citizens Committee for Broadcasting, 436 U. S. 775, 796-797 (1978); Mourning v. Family Publications Service, Inc., 411 U. S. 356, 369 (1973); Thorpe *709v. Housing Authority of City of Durham, 393 U. S. 268, 280-281 (1969). The Court has never — until today — interpreted a “reasonably necessary or appropriate” clause as having a substantive content that supersedes a specific congressional directive embodied in a provision that is focused more particularly on an agency’s authority. This principle, of course, reflects the common understanding that the determination of whether regulations are “reasonably necessary” may be made only by reference to the legislative judgment reflected in the statute; it must not be based on a court’s own, inevitably subjective view of what steps should be taken to promote perceived statutory goals.
The plurality suggests that under the “reasonably necessary” clause, a workplace is not “unsafe” unless the Secretary is able to convince a reviewing court that a “significant” risk is at issue. Ante, at 642. That approach is particularly embarrassing in this case, for it is contradicted by the plain language of the Act. The plurality’s interpretation renders utterly superfluous the first sentence of § 655 (b)(5), which, as noted above, requires the Secretary to set the standard “which most adequately assures . . . that no employee will suffer material impairment of health.” Indeed, the plurality’s interpretation reads that sentence out of the Act. By so doing, the plurality makes the test for standards regulating toxic substances and harmful physical agents substantially identical to the test for standards generally — plainly the opposite of what Congress intended. And it is an odd canon of construction that would insert in a vague and general definitional clause a threshold requirement that overcomes the specific language placed in a standard-setting provision. The most elementary principles of statutory construction demonstrate that precisely the opposite interpretation is appropriate. See, e. g., FPC v. Texaco Inc., 417 U. S. 380, 394-395 (1974); Clark v. Uebersee Finanz-Korp., 332 U. S. 480, 488-489 (1947). In short, Congress could have provided that the Secretary may not take regulatory action until the existing *710scientific evidence proves the risk at issue to be “significant,” 27 but it chose not to do so.
The plurality’s interpretation of the “reasonably necessary or appropriate” clause is also conclusively refuted by the legislative history. While the standard-setting provision that the plurality ignores received extensive legislative attention, the definitional clause received none at all. An earlier version of the Act, see n. 8, supra, did not embody a clear feasibility constraint and was not restricted to toxic substances or to “material” impairments. The “reasonably necessary or appropriate” clause was contained in this prior version of the bill, as it was at all relevant times. In debating this version, Members of Congress repeatedly expressed concern that it would require a risk-free universe. See, e. g., ante, at 646-649. The definitional clause was not mentioned at all, an omission that would be incomprehensible if Congress intended *711by that clause to require the Secretary to quantify the risk he sought to regulate in order to demonstrate that it was “significant.”
The only portions of the legislative history on which the plurality relies, see ibid.-, have nothing to do with the “reasonably necessary or appropriate” clause from which the “threshold finding” requirement is derived. Those portions consisted of criticisms directed toward the earlier version of the statute which already contained the definitional clause. These criticisms, in turn, were met by subsequent amendments that limited application of the strict “no employee will suffer” clause to toxic substances, inserted an explicit feasibility constraint, and modified the word “impairment” by the adjective “material.” It is disingenuous at best for the plurality to suggest that isolated statements in the legislative history, expressing concerns that were met by subsequent amendments not requiring any “threshold” finding, can justify reading such a requirement into a “reasonably necessary” clause that was in the Act all along.28
*712The plurality’s various structural arguments are also unconvincing. The fact that a finding of “grave danger” is required for temporary standards, see ante, at 640, n. 45, hardly implies that the Secretary must show for permanent standards that it is more probable than not that the substance to be regulated poses a “significant” risk. Nor is the reference to “toxic materials,” ante, at 643, in any way informative. And the priority-setting provision, ante, at 643-644, cannot plausibly be read to condition the Secretary’s standard-setting authority on an ability to meet the Court’s “threshold” requirement.
The plurality ignores applicable canons of construction, apparently because it finds their existence inconvenient. But as we stated quite recently, the inquiry into statutory purposes should be “informed by an awareness that the regulation is entitled to deference unless it can be said not to be a reasoned and supportable interpretation of the Act.” Whirlpool Corp. v. Marshall, 445 U. S. 1, 11 (1980). Can it honestly be said that the Secretary’s interpretation of the Act is “unreasoned” or “unsupportable”? And as we stated in the same case, “safety legislation is to be liberally construed to effectuate the congressional purpose.” Id., at 13. The plurality’s disregard of these principles gives credence to the frequently voiced criticism that they are honored only when the Court finds itself in substantive agreement with the agency action at issue.
In short, today’s decision represents a usurpation of deci-sionmaking authority that has been exercised by and properly belongs with Congress and its authorized representatives. *713The plurality’s construction has no support in the statute’s language, structure, or legislative history. The threshold finding that the plurality requires is the plurality’s own invention. It bears no relationship to the acts or intentions of Congress, and it can be understood only as reflecting the personal views of the plurality as to the proper allocation of resources for safety in the American workplace.
C
The plurality is obviously more interested in the consequences of its decision than in discerning the intention of Congress. But since the language and legislative history of the Act are plain, there is no need for conjecture about the effects of today’s decision. “It is not for us to speculate, much less act, on whether Congress would have altered its stance had the specific events of this case been anticipated.” TVA v. Hill, 437 U. S., at 185. I do not pretend to know whether the test the plurality erects today is, as a matter of policy, preferable to that created by Congress and its delegates: the area is too fraught with scientific uncertainty, and too dependent on considerations of policy, for a court to be able to determine whether it is desirable to require identification of a “significant” risk before allowing an administrative agency to take regulatory action. But in light of the tenor of the plurality opinion, it is necessary to point out that the question is not one-sided, and that Congress’ decision to authorize the Secretary to promulgate the regulation at issue here was a reasonable one.
In this case the Secretary found that exposure to benzene at levels above 1 ppm posed a definite albeit unquantifiable risk of chromosomal damage, nonmalignant blood disorders, and leukemia. The existing evidence was sufficient to justify the conclusion that such a risk was presented, but it did not permit even rough quantification of that risk. Discounting for the various scientific uncertainties, the Secretary gave *714“careful consideration to the question of whether th[e] substantial costs” of the standard “are justified in light of the hazards of exposure to benzene,” and concluded that “these costs are necessary in order to effectuate the statutory purpose . . . and to adequately protect employees from the hazards of exposure to benzene.” 43 Fed. Reg. 5941 (1978).
In these circumstances it seems clear that the Secretary found a risk that is “significant” in the sense that the word is normally used. There was some direct evidence of chromosomal damage, nonmalignant blood disorders, and leukemia at exposures at or near 10 ppm and below. In addition, expert after expert testified that the recorded effects of benzene exposure at higher levels justified an inference that an exposure level above 1 ppm was dangerous. The plurality’s extraordinarily searching scrutiny of this factual record reveals no basis for a conclusion that quantification is, on the basis of “the best available evidence,” possible at the present time. If the Secretary decided to wait until definitive information was available, American workers would be subjected for the indefinite future to a possibly substantial risk of benzene-induced leukemia and other illnesses. It is unsurprising, at least to me, that he concluded that the statute authorized him to take regulatory action now.
Under these circumstances, the plurality’s requirement of identification of a “significant” risk will have one of two consequences. If the plurality means to require the Secretary realistically to “quantify” the risk in order to satisfy a court that it is “significant,” the record shows that the plurality means to require him to do the impossible. But regulatory inaction has very significant costs of its own. The adoption of such a test would subject American workers to a continuing risk of cancer and other serious diseases; it would disable the Secretary from regulating a wide variety of carcinogens for which quantification simply cannot be undertaken at the present time.
*715There are encouraging signs that today’s decision does not extend that far.29 My Brother Powell concludes that the Secretary is not prevented from taking regulatory action “when reasonable quantification cannot be accomplished by any known methods.” See ante, at 666. The plurality also indicates that it would not prohibit the Secretary from promulgating safety standards when quantification of the benefits is impossible. See ante, at 656-657, and n. 63. The Court might thus allow the Secretary to attempt to make a very rough quantification of the risk imposed by a carcinogenic substance, and give considerable deference to his finding that the risk was significant. If so, the Court would permit the Secretary to promulgate precisely the same regulation involved in these cases if he had not relied on a carcinogen “policy,” but undertaken a review of the evidence and the *716expert testimony and concluded, on the basis of conservative assumptions, that the risk addressed is a significant one. Any other interpretation of the plurality’s approach would allow a court to displace the agency’s judgment with its own subjective conception of “significance,” a duty to be performed without statutory guidance.
The consequences of this second approach would hardly be disastrous; indeed, it differs from my own principally in its assessment of the basis for the Secretary’s decision in these cases. It is objectionable, however, for three reasons. First, the requirement of identification of a “significant” risk simply has no relationship to the statute that the Court today purports to construe. Second, if the “threshold finding” requirement means only that the Secretary must find “that there is a need for such a standard,” ante, at 643, n. 48, the requirement was plainly satisfied by the Secretary’s express statement that the standard’s costs “are necessary in order to effectuate the statutory purpose . . . and to adequately protect employees from the hazards of exposure to benzene.” 43 Fed. Reg. 5941 (1978). Third, the record amply demonstrates that in light of existing scientific knowledge, no purpose would be served by requiring the Secretary to take steps to quantify the risk of exposure to benzene at low levels. Any such quantification would be based not on scientific “knowledge” as that term is normally understood, but on considerations of policy. For carcinogens like benzene, the assumptions on which a dose-response curve must be based are necessarily arbitrary. To require a quantitative showing of a “significant” risk, therefore, would either paralyze the Secretary into inaction or force him to deceive the public by acting on the basis of assumptions that must be considered too speculative to support any realistic assessment of the relevant risk. See McGarity, Substantive and Procedural Discretion in Administrative Resolution of Science Policy Questions: Regulating Carcinogens in EPA and OSHA, 67 Geo. L. J. 729, 806 (1979). It is encouraging that the Court appears willing *717not to require quantification when it is not fairly possible. See ante, at 656-657, and n. 63.
Though it is difficult to see how a future Congress could be any more explicit on the matter than was the Congress that passed the Act in 1970, it is important to remember that today’s decision is subject to legislative reversal. Congress may continue to believe that the Secretary should not be prevented from protecting American workers from cancer and other fatal diseases until scientific evidence has progressed to a point where he can convince a federal court that the risk is “significant.” Today’s decision is objectionable not because it is final, but because it places the burden of legislative inertia on the beneficiaries of the safety and health legislation in question in these cases. By allocating the burden in this fashion, the Court requires the American worker to return to the political arena and to win a victory that he won once before in 1970. I am unable to discern any justification for that result.
D
Since the plurality’s construction of the “reasonably necessary or appropriate” clause is unsupportable, I turn to a brief discussion of the other arguments that respondents offer in support of the judgment below.
First, respondents characterize the Act as a pragmatic statute designed to balance the benefits of a safety and health regulation against its costs. Respondents observe that the statute speaks in terms of relative protection by providing that safety must be assured “so far as possible,” 29 U. S. C. § 651 (b), and by stating that the “no material impairment” requirement is to be imposed only “to the extent feasible.” 30 *718Respondents contend that the term “feasible” should be read to require consideration of the economic burden of a standard. not merely its technological achievability. I do not understand the Secretary to disagree. But respondents present no argument that the expenditure required by the benzene standard is not feasible in that respect. The Secretary concluded on the basis of substantial evidence that the costs of the standard would be readily absorbed by the 20 affected industries. One need not define the feasibility requirement with precision in order to conclude that the benzene standard is “feasible” in the sense that it will not materially harm the financial condition of the regulated industries.
Respondents suggest that the feasibility requirement should be understood not merely to refer to a standard’s expense, but also to mandate a finding that the benefits of an occupational safety and health standard bear a reasonable relation *719to its costs. I believe that the statute’s language, structure, and legislative history foreclose respondents’ position. In its ordinary meaning an activity is “feasible” if it is capable of achievement, not if its benefits outweigh its costs. See Webster’s Third New International Dictionary 831 (1976). Moreover, respondents’ interpretation would render § 655 (b) (5) internally inconsistent by reading into the term “feasible” a requirement irreconcilable with the express language authorizing the Secretary to set standards assuring that “no employee will suffer material impairment. . . .” Respondents’ position would render that language merely hortatory. As noted above, no cost-benefit analysis is referred to at any point in the statute or its legislative history, an omission which cannot be deemed inadvertent in light of the explicit cost-benefit requirements inserted into other regulatory legislation.31 Finally, the legislative history of the feasibility requirement, see n. 8, supra, demonstrates that Congress’ sole concern was that standards be economically and technologically achievable. The legislative intent was to prevent the Secretary from materially harming the financial condition of regulated industries in order to eliminate risks of impairment. Congress did not intend to preclude the Secretary from taking regulatory action where, as here, no such threat to industry is posed.32
*720In order to decide these cases, however, it is not necessary to resolve the question whether the term “feasible” may contemplate some balancing of the costs and benefits of regulatory action.33 Taking into account the uncertainties in existing knowledge, the Secretary made an express finding that the hazards of benzene exposure were sufficient to justify the regulation’s costs. 43 Fed. Reg. 5941 (1978). Any requirement to balance costs and benefits cannot be read to invalidate this wholly rational conclusion. A contrary result, forcing the Secretary to wait for quantitative data that may not be available in the foreseeable future, would run directly counter to the protective purposes of the Act.34
*721Finally, respondents suggest broadly that the Secretary did not fulfill his statutory responsibility to act on the basis of “research, demonstrations, experiments,” and to consider “the latest available scientific data in the field, the feasibility of the standards, and experience gained under this and other health and safety laws.” 29 U. S. C. §655 (b)(5). Here, they contend, the Secretary based his decision solely on “views and arguments.” Brief for Respondents American Petroleum Institute et al. 52. I disagree. The Secretary compiled an extensive record composed of over 50 volumes of exhibits. Most of those exhibits are the reported results of research and demonstrations representing “the latest available scientific data.” The Secretary offered a careful discussion of these data in the statement accompanying the permanent standard. His ultimate conclusions were grounded in extensive findings of fact. Where, as here, there are gaps in existing knowledge, the Secretary’s decision must necessarily be based on considerations of policy as well as on empirically verifiable facts.
In passing the Occupational Safety and Health Act of 1970, Congress was aware that it was authorizing the Secretary to regulate in areas of scientific uncertainty. But it intended to require stringent regulation even when definitive information was unavailable. In reducing the permissible level of exposure to benzene, the Secretary applied proper legal standards. His determinations are supported by substantial evi-*722denee. The Secretary’s decision was one, then, which the governing legislation authorized him to make.35
IV
In recent years there has been increasing recognition that the products of technological development may have harmful effects whose incidence and severity cannot be predicted with certainty. The responsibility to regulate such products has fallen to administrative agencies. Their task is not an enviable one. Frequently no clear causal link can be established between the regulated substance and the harm to be averted. Risks of harm are often uncertain, but inaction has considerable costs of its own. The agency must decide whether to take regulatory action against possibly substantial risks or to wait until more definitive information becomes available — a judg*723ment which by its very nature cannot be based solely on determinations of fact.36
Those delegations, in turn, have been made on the understanding that judicial review would be available to ensure that the agency’s determinations are supported by substantial evidence and that its actions do not exceed the limits set by Congress. In the Occupational Safety and Health Act, Congress expressed confidence that the courts would carry out this important responsibility. But in these cases the plurality has far exceeded its authority. The plurality’s “threshold finding” requirement is nowhere to be found in the Act and is antithetical to its basic purposes. “The fundamental policy questions appropriately resolved in Congress . . . are not subject to re-examination in the federal courts under the guise of judicial review of agency action.” ' Vermont Yankee Nuclear Power Corp. v. NRDC, 435 U. S., at 558 (emphasis in original). Surely this is no less true of the decision to ensure safety for the American worker than the decision to proceed with nuclear power. See ibid.
Because the approach taken by the plurality is so plainly irreconcilable with the Court’s proper institutional role, I am certain that it will not stand the test of time. In all likelihood, today’s decision will come to be regarded as an extreme reaction to a regulatory scheme that, as the Members of the plurality perceived it, imposed an unduly harsh burden on regulated industries. But as the Constitution “does not enact Mr. Herbert Spencer’s Social Statics,” Lochner v. New York, 198 U. S. 45, 75 (1905) (Holmes, J., dissenting), so the responsibility to scrutinize federal administrative action does not authorize this Court to strike its own balance between the *724costs and benefits of occupational safety standards. I am confident that the approach taken by the plurality today, like that in Lochner itself, will eventually be abandoned, and that the representative branches of government will once again be allowed to determine the level of safety and health protection to be accorded to the American worker.
Legislative History of the Occupational Safety and Health Act of 1970 (Committee Print compiled for the Senate Committee on Labor and Public Welfare), p. iii (1971) (Foreword by Sen. Williams) (hereinafter Leg. Hist.).
S. Rep. No. 91-1282, p. 2 (1970), Leg. Hist. 142.
Leg. Hist. iii.
S. Rep. No. 91-1282, p. 2 (1970), Leg. Hist. 142; 116 Cong. Rec. 37326 (1970), Leg. Hist. 415 (Sen. Williams); H. R. Rep. No. 91-1291, p. 19 (1970), Leg. Hist. 849; 116 Cong. Rec. 38392-38393 (1970), Leg. Hist. 1049 (Rep. Earth).
116 Cong. Rec. 38375 (1970), Leg. Hist. 1003 (Sen. Daniels).
116 Cong. Rec., at 37623, Leg. Hist. 503 (Sen. Dominick); H. R. No. 91-1291, p. 28 (1970), Leg. Hist. 858.
See n. 34, infra.
An earlier version of the bill had provided:
“The Secretary, in promulgating standards under this subsection, shall set the standard which most adequately and feasibly assures, on the basis of the best available evidence, that no employee will suffer any impairment *694of health or functional capacity, or diminished life expectancy even if such employee has regular exposure to the hazard dealt with by such standard for the period of his working life.” S. 2193, 91st Cong., 2d Sess., 39 (1970), Leg. Hist. 242.
This standard, it was feared, “could be read to require the Secretary to ban all occupations in which there remains some risk of injury, impaired health, or life expectancy. In the case of all occupations, it will be impossible to eliminate all risks to safety and health. Thus, the present criteria could, if literally applied, close every business in this nation. In addition, in many cases, the standard which might most 'adequately’ and 'feasibly’ assure the elimination of the danger would be the prohibition of the occupation itself.” 116 Cong. Rec. 36530 (1970), Leg. Hist. 367 (Statement on Amendment of Sen. Dominick). In explaining the present language, Senator Dominick stated:
“What we were trying to do in the bill — unfortunately, we did not have the proper wording or the proper drafting — was to say that when we are dealing with toxic agents or physical agents, we ought to take such steps as are feasible ánd practical to provide an atmosphere within which a person’s health or safety would not be affected. Unfortunately, we had language providing that anyone would be assured that no one would have a hazard ... so that no one would have any problem for the rest of his working life.
“It was an unrealistic standard. As modified, we would be approaching the problem by looking at the problem and setting a standard or criterion which would not result in harm.” 116 Cong. Rec., at 37622, Leg. Hist. 502.
I do not, of course, suggest that it is appropriate for a federal court reviewing agency action blindly to defer to the agency’s findings of fact and determinations of policy. Under Citizens to Preserve Overton Park, Inc. v. Volpe, 401 U. S. 402, 416 (1971), courts must undertake a “searching and careful” judicial inquiry into those factors. Such an inquiry is designed to require the agency to take a “hard look,” Kleppe v. Sierra Club, 427 U. S. 390, 410, n. 21 (1976) (citation omitted), by considering the proper factors and weighing them in a reasonable manner. There is also room for especially rigorous judicial scrutiny of agency decisions under a rationale akin to that offered in United States v. Carolene Products Co., 304 U. S. 144, 152, n. 4 (1938). See Environmental Defense Fund v. Ruckelshaus, 142 U. S. App. D. C. 74, 439 F. 2d 584 (1971).
I see no basis, however, for the approach taken by the plurality today, which amounts to nearly de novo review of questions of fact and of regulatory policy on behalf of institutions that are by no means unable to protect themselves in the political process. Such review is especially inappropriate when the factual questions at issue are ones about which the Court cannot reasonably be expected to have expertise.
Tr. 258-259, 1039.
Id., at 148,200-201, 258.
Id., at 145, 173-174, 352, 1227,1928, 3206; 15 Record, Ex. 43B, p. 166.
Id., at 149, 360-361, 997, 1023, 2543, 2689, 3203; 11 Record, Ex. 3.
Tr. 149, 1218, 2692, 2847.
Id., at 308, 314, 747, 768, 769-770, 874, 2445. As the Secretary observed, the issue of the exposure level in the NIOSH study was extensively debated during the hearings. A report from the Industrial Commission of Ohio suggested that concentrations generally ranged from zero to 10 or 15 ppm. But the Secretary concluded that evidence at the hearings showed that area exposures during the study period had sometimes substantially exceeded that level. Because of the conflicting evidence and the absence of monitoring data, he found that the excess leukemia risk observed in the NIOSH study could not be linked to any particular exposure level.
As to the study on which industry relied most heavily, for example, the Secretary, largely repeating the author’s own admissions, observed that (1) a number of employees included in the sample may not have been exposed to benzene at any time; (2) there was inadequate followup of numerous employees, so that persons who may have contracted leukemia were not included in the data; (3) the diagnoses were subject to serious question, and cases of leukemia may have gone unnoticed; (4) no determination of exposure levels had been made; and (5) the occupational histories of the workers were admittedly incomplete. 43 Fed. Reg. 5928 (1978).
Tr. 1023-1024, 1227 ; 22A Record, Ex. 154.
The testimony of Dr. Aksoy, one of the world’s leading experts, was typical: “[E]ven one ppm . . . causes leukemia.” Tr. 204. See also id., at 30, 150, 262, 328, 351-352, 363-364, 394, 745-746, 1057, 1210, 2420 ; 9 Record, Ex. 2.8-272, p. 1.
Tr. 130, 360, 414-415, 416-417, 760-761, 781-782, 925, 1055-1056; 17 Record, Ex. 75, p. 2; 1 Record, Ex. 2-4, p. 11.
Tr. 382, 401, 405, 1372, 2846, 2842-2843.
Id., at 148-149 ("the permissible exposure limit for benzene should be zero”) (testimony of Dr. Aksoy). See also id., at 1251 et seq., 3506 et seq.
The plurality’s estimate of the amount of expenditure per employee, see ante, at 629, is highly misleading. Most of the costs of the benzene *702standard would be incurred only onee and would thus protect an unascer-tainable number of employees in the future; that number will be much higher than the number of employees currently employed.
The projection, designed as an extrapolation from an amalgamation of existing studies, was dependent on a number of assumptions which the Secretary could reasonably view as questionable. Indeed, the witness himself stated that his estimate was based on “a lousy set of data,” was “slightly better than a guess,” Tr. 2772, and that there was “no real basis,” id., at 2719, for a dose-response curve on which the estimate was wholly dependent.
The witness’ assumptions were severely tested during the hearings, see id., at 2795 et seq., and the Secretary could reasonably reject them on the basis of the evidence in the record. For example: (1) The witness appeared to assume that in previous tests leukemia had been contracted after a lifetime of exposure; the evidence afforded no basis for that assumption, and the duration of exposure may have been quite short for particular employees. If the duration period was short, the witness’ estimate would have been much too low. (2) The witness assumed that exposure levels in the NIOSH study were around 100 ppm. The Secretary *703found, however, that no such assumption could be made, and there was evidence that exposure levels had generally been between zero and 10-15 ppm. (3) The witness assumed that the dose-response curve was linear at all levels, but there was no basis for that assumption. In the case of vinyl chloride (another carcinogen for which the Secretary has promulgated exposure standards), recent evidence suggested that the dose-response curve rises steeply at low doses and becomes less steep as the levels are increased. (4) Twenty-five percent of the workers in the NIOSH study had not been found, and the witness assumed that they were still alive and would not contract leukemia. Six hundred additional workers exposed in that study were still alive; the witness assumed they too would not contract leukemia. There was considerable testimony that, for these and other reasons, the NIOSH study significantly underestimated the risk. The witness assumes that it had not. (5) The NIOSH study found a fivefold excess risk from benzene exposure; the witness assumed that the excess was much lower, despite the NIOSH finding and the testimony that that finding was a significant understatement of the risk. In light of these uncertainties, the Secretary could conclude that the witness’ estimate was unsupportable.
Witnesses testifying to the inability to construct a dose-response curve referred primarily to the impossibility of correlating the incidence of leukemia, blood disorders, and chromosomal damage with the levels and duration of exposure in past studies. Thus Dr. Herman Kraybill of the National Cancer Institute testified:
“[W]e like to estimate risk factors. This has been done, as many of you recall, with vinyl chloride several years ago.
“. . . [T]o estimate the risk factors on [the basis of] experimental data, this presupposes if you have good toxicity data. When I say toxicity data, I mean good dose-response data on vinyl chloride, which indeed we did have that.
“But with benzene, it appeared that we didn't have this situation, so therefore, most of us gave up. . . .
“. . . With benzene, we sort of struck out.” Id., at 760-761.
Because of the enormous uncertainties in levels and duration of exposure in prior studies, any assumptions would necessarily be arbitrary. The possible range of assumptions was so great that the ultimate conclusion would be entirely uninformative. See id., at 360, 415, 1055-1056.
At one point the Secretary did indicate that appreciable benefits were “likely” to result. The Court of Appeals held that this conclusion was unsupported by substantial evidence. The Secretary’s suggestion, however, was made in the context of a lengthy discussion intended to show that appreciable benefits "may” be predicted but that their likelihood could not be quantified. The suggestion should not be taken as a definitive statement that appreciable benefits were more probable than not.
For reasons stated infra, there is nothing in the Act to prohibit the Secretary from acting when he is unable to conclude that appreciable benefits are more probable than not.
This is not to say that the Secretary is prohibited from examining relative costs and benefits in the process of setting priorities among hazardous substances, or that systematic consideration of costs and benefits is not to be attempted in the standard-setting process. Efforts to quantify costs and benefits, like statements of reasons generally, may help to promote informed consideration of decisional factors and facilitate *708judicial review. See Dunlop v. Bachowski, 421 U. S. 560, 571-574 (1975). The Secretary indicates that he has attempted to quantify costs and benefits in the past. See 43 Fed. Reg. 54354, 54427-54431 (1978) (lead); id at 27350, 27378-27379 (cotton dust).
It is not necessary in the present litigation to say whether the Secretary must show a reasonable relation between costs and benefits. Discounting for the scientific uncertainty, the Secretary expressly — and reasonably— found such a relation here.
It is useful to compare the Act with other regulatory statutes in which Congress has required a showing of a relationship between costs and benefits or of an "unreasonable risk.” In some statutes Congress has expressly required cost-benefit analysis or a demonstration of some reasonable relation between costs and benefits. See 33 U. S. C. § 701a (Flood Control Act of 1936); 42 U. S. C. §7545 (c)(2)(B) (1976 ed., Supp. II) (Clean Air Act); 33 U. S. C. § 1314 (b)(4)(B) (1976 ed., Supp. II) (Clean Water Act). In others Congress has imposed two independent requirements: that administrative action be “feasible” and justified by a balancing of costs and benefits, e. g., 43 U. S. C. § 1347 (b) (1976 ed., Supp. II) (Outer Continental Shelf Lands Act); 42 U. S. C. § 6295 (a) (2)(D) (1976 ed., Supp. II) (Energy Policy and Conservation Act). This approach demonstrates a legislative awareness of the difference between a feasibility constraint and a constraint based on weighing costs and benefits. See infra, at 719-720. In still others Congress has authorized regulation of “unreasonable risk,” a term which has been read by some courts to require a balancing of costs and benefits. See, e. g., Aqua Slide ‘N’ Dive Corp. v. Consumer Product Safety Comm’n, 569 F. 2d 831 (CA5 1978) (construing 15 U. S. C. §2058 (c)(2)(A) (Consumer Product Safety Act)); Forester v. Consumer Product Safety Comm’n, 182 U. S. App. D. C. 153, 559 F. 2d 774 (1977) (construing 15 U. S. C. § 1261 (s) (Child Protection and Toy Safety Act)).
The plurality also relies on its perception that if the “reasonably necessary” clause were not given the meaning it ascribes to it, there would be no guidance for “standards other than those dealing with toxic materials and harmful physical agents.” Ante, at 640, n. 45. For two reasons this argument is without force. First, even if the “reasonably necessary” clause does have independent content, and even if that content is as the plurality describes it, it cannot under any fairminded reading supersede the express language of § 655 (b) (5) for toxic substances and harmful physical agents.
Second, as noted above, an earlier version of the bill applied the “no employee will suffer” language to all substances. At that time, there was no “gap,” and accordingly it could not be argued that the “reasonably necessary or appropriate” clause had the content the plurality ascribes to it. In this light, the plurality’s reasoning must be that when Congress amended the bill to apply the strict § 655 (b) (5) requirements only to toxic substances, the definitional clause gained an independent meaning that in turn comprehended all standards. But surely this argument turns congressional purposes on their head. It reasons that when *712Congress singled out toxic substances for special regulation, it simultaneously created a more lenient (“reasonably necessary”) test for standards generally, and that once that more lenient test was applicable, it somehow superseded the strict requirements for toxic substances. That reasoning is both illogical and circular. Nor is there any basis for the plurality’s suggestion, see ante, at 649, n. 54, that the original bill’s application to all standards was “entirely inadvertent.”
The plurality suggests that it is for the agency “to determine, in the first instance, what it considers to be a 'significant’ risk,” and that the agency “is free to use conservative assumptions in interpreting the data. . . .” Ante, at 655, 656. Moreover, my Brother Powell would not require “quantification of risk in every case.” Ante, at 666 (opinion concurring in part and concurring in judgment). As I read his opinion, Mr. Justice Powell would have permitted the Secretary to promulgate the standard at issue here if the Secretary had provided a more carefully reasoned explanation of his conclusion that the risk at issue justified the admittedly significant costs of the benzene standard. Mr. Justice Powell also suggests that such a conclusion would be subject to relatively deferential review. Ante, at 670-671, n. 8.
In this respect, the differences between my approach and that of Mr. Justice Powell may be comparatively narrow. We are agreed on two propositions that I regard as critical to a fairminded interpretation of the Act: (1) the Secretary may regulate risks that are not subject to quantification on the basis of the “best available evidence”; and (2) the Secretary’s judgment that a particular health risk merits regulatory action is subject to limited judicial scrutiny. It is encouraging that at least five Members of the Court accept these basic propositions.
For reasons stated in the text, however, I disagree with my Brother Powell’s conclusion that it is appropriate to hold in these cases that the Act requires the Secretary to show a reasonable relationship between costs and benefits.
Finding obscurity in the word “feasible,” my Brother RehNQüist invokes the nondelegation doctrine, which was last used to invalidate an Act of Congress in 1935. A. L. A. Schecter Poultry Corp. v. United States, 295 U. S. 495 (1935). While my Brother Rehnquist eloquently argues that there remains a place for such a doctrine in our jurisprudence, I am frankly puzzled as to why the issue is thought to be of any relevance *718here. The nondelegation doctrine is designed to assure that the most fundamental decisions will be made by Congress, the elected representatives of the people, rather than- by administrators. Some minimal definiteness is therefore required in order for Congress to delegate its authority to administrative agencies.
Congress has been sufficiently definite here. The word “feasible” has a reasonably plain meaning, and its interpretation can be informed by other contexts in which Congress has used it. See n. 27, supra. Since the term is placed in the same sentence with the “no employee will suffer” language, it is clear that “feasible” means technologically and economically achievable. Under the Act, the Secretary is afforded considerably more guidance than are other administrators acting under different regulatory statutes. In short, Congress has made “the critical policy decisions” in these cases, see ante, at 687 (Rehnquist, J., concurring in ,j udgment).
The plurality’s apparent suggestion, see ante, at 646, that the nondele-gation doctrine might be violated if the Secretary were permitted to regulate definite but nonquantifiable risks is plainly wrong. Such a statute would be quite definite and would thus raise no constitutional question under Schechter Poultry. Moreover, Congress could rationally decide that it would be better to require industry to bear “feasible” costs than to subject American workers to an indeterminate risk of cancer and other fatal diseases.
See n. 27, supra.
Congress’ antipathy toward cost-benefit balancing is evident throughout the legislative history of the Act. For example:
“The costs that will be incurred by employers in meeting the standards of health and safety to be established under this bill are, in my view, reasonable and necessary costs of doing business. Whether we, as individuals, are motivated by simple humanity or by simple economics, we can no longer permit profits to be dependent upon an unsafe or unhealthy worksite.” 116 Cong. Rec 41766 (1970), Leg. Hist. 1150-1151 (Sen. Eagieton).
Similarly, Senator Yarborough stated:
“We are talking about people’s lives, not the indifference of some cost accountants. We are talking about assuring the men and women who *720work in our plants and factories thát they will go home after a day’s work with their bodies intact. We are talking about assuring our American workers who wo[r]k with deadly chemicals that when they have accumulated a few year’s seniority they will not have accumulated lung congestion and poison in their bodies, or something that will strike them down before they reach retirement age.” 116 Cong. Rec., at 37625, Leg. Hist. 510.
Nor need I discuss the possibility, raised by counsel for the federal parties in oral argument, that a decision to regulate a substance posing a negligible threat to health and safety could itself be challenged as arbitrary and capricious under the Administrative Procedure Act. See Tr. of Oral Arg. 23.
Respondents also rely on the statutory requirement that the Secretary may act only to prevent "material” impairment. They contend that the standard promulgated here does not fall within that category because the risk is so low. This interpretation derives no support from the statute or its legislative history. The statute itself states that standards should ensure that no employee will suffer “material impairment,” not material risk of impairment.
The language is consistent with the legislative history. In an early version of the Act, the word “impairment” was modified by “any” rather than “material.” See n. 8, supra. The feasibility and materiality requirements were added simultaneously as part of an effort to qualify the original language authorizing the Secretary to ensure that “no employee will suffer any impairment of health or functional capacity, or diminished life expectancy.” Senator Dominick was concerned that the *721phrase “any” impairment would require the Secretary to prevent insect bites. 116 Cong. Rec. 36522 (1970), Leg. Hist. 345.
The respondents’ construction would pose an enormous obstacle to efforts to regulate toxic substances under §655 (b)(5). The probability of contracting cancer will in most contexts be quite small with respect to any particular employee. If the statute were read to authorize the Secretary to act only to assure that “no employee will suffer material risk of impairment,” the Secretary would be disabled from regulating substances which poses a small risk with respect to any particular employee but which will nonetheless result in the death of numerous members of the employee pool.
Although the Court of Appeals accepted the Secretary’s finding that dermal contact with benzene could cause leukemia, it set aside the dermal contact standard because of the Secretary’s failure to perform an experiment recommended by an industry witness. The failure to conduct this test, according to the court, violated the statutory requirement that the Secretary act on the basis of “the best available evidence” and “the latest available scientific data in the field.”
In the hearings before the agency, respondents presented no substantial challenge to the position that benzene could be absorbed through the skin, and there was evidence in the record to support that position. Both animal and human studies had found such absorption. In these circumstances, the Secretary was not obligated to undertake additional studies simply because a witness testified that such studies would be informative. The imposition of such a requirement would paralyze the standard-setting process. The Secretary’s mandate is to act on the basis of “available” evidence, not evidence which may become available in the future.
In setting aside the dermal contact standard, the Court of Appeals also relied on its conclusion that the Secretary had not shown that quantifiable benefits would result from the standard. As the discussion above indicates, the court applied incorrect legal standards in so holding.
See W. Lowrance, Of Acceptable Risk: Science and the Determination of Safety (1976); Stewart, Paradoxes of Liberty, Integrity and Fraternity: The Collective Nature of Environmental Quality and Judicial Review of Administrative Action, 7 Environ. L. 463, 469-472 (1977).