*118Justice White

delivered the opinion of the Court.

These cases present the question whether the Environmental Protection Agency (EPA) may issue certain variances from toxic pollutant effluent limitations promulgated under the Clean Water Act, 86 Stat. 816, as amended, 33 U. S. C. § 1251 et seq. 1

I

As part of a consolidated lawsuit, respondent Natural Resources Defense Council (NRDC) sought a declaration that §301(1) of the Clean Water Act, 33 U. S. C. §1311(0, prohibited EPA from issuing “fundamentally different factor” (FDF) variances for pollutants listed as toxic under the Act.2 Petitioners EPA and Chemical Manufacturers Association (CMA) argued otherwise. To understand the nature of this controversy, some background with respect to the statute and the case law is necessary.

The Clean Water Act, the basic federal legislation dealing with water pollution, assumed its present form as the result of extensive amendments in 1972 and 1977. For direct dischargers — those who expel waste directly into navigable waters — the Act calls for a two-phase program of technology-based effluent limitations, commanding that dischargers comply with the best practicable control technology currently available (BPT) by July 1, 1977, and subsequently meet the generally more stringent effluent standard consistent with the best available technology economically achievable (BAT).3

*119Indirect dischargers — those whose waste water passes through publicly owned treatment plants — are similarly required to comply with pretreatment standards promulgated by EPA under §307 of the Act, 33 U. S. C. § 1317(b), for pollutants not susceptible to treatment by sewage systems or which would interfere with the operation of those systems. Relying upon legislative history suggesting that pretreatment standards are to be comparable to limitations for direct dischargers, see H. R. Rep. No. 95-830, p. 87 (1977), and pursuant to a consent decree,4 EPA has set effluent limitations for indirect dischargers under the same two-phase approach applied to those discharging waste directly into navigable waters.

Thus, for both direct and indirect dischargers, EPA considers specific statutory factors6 and promulgates regulations creating categories and classes of sources and setting uniform discharge limitations for those classes and categories. Since *120application of the statutory factors varies on the basis of the industrial process used and a variety of other factors, EPA has faced substantial burdens in collecting information adequate to create categories and classes suitable for uniform effluent limits, a burden complicated by the time deadlines it has been under to accomplish the task.6 Some plants may find themselves classified within a category of sources from which they are, or claim to be, fundamentally different in terms of the statutory factors. As a result, EPA has developed its FDF variance as a mechanism for ensuring that its necessarily rough-hewn categories do not unfairly burden atypical plants.7 Any interested party may seek an FDF *121variance to make effluent limitations either more or less stringent if the standards applied to a given source, because of factors fundamentally different from those considered by *122EPA in setting the limitation, are either too lenient or too strict.8

The 1977 amendments to the Clean Water Act reflected Congress’ increased concern with the dangers of toxic pollutants. The Act, as then amended, allows specific statutory modifications of effluent limitations for economic and water-*123quality reasons in §§ 301(c) and (g).9 Section 301(1), however, added by the 1977 amendments, provides:

“The Administrator may not modify any requirement of this section as it applies to any specific pollutant which is on the toxic pollutant list under section 307(a)(1) of this Act.” 91 Stat. 1590.

In the aftermath of the 1977 amendments, EPA continued its practice of occasionally granting FDF variances for BPT *124requirements. The Agency also promulgated regulations explicitly allowing FDF variances for pretreatment standards10 and BAT requirements.11 Under these regulations, EPA granted FDF variances, but infrequently.12

As part of its consolidated lawsuit, respondent NRDC here challenged pretreatment standards for indirect dischargers and sought a declaration that § 301(1) barred any FDF variance with respect to toxic pollutants.13 In an earlier case, the Fourth Circuit had rejected a similar argument, finding that § 301(1) was ambiguous on the issue of whether it applied to FDF variances and therefore deferring to the administrative agency’s interpretation that such variances were permitted. Appalachian Power Co. v. Train, 620 F. 2d 1040, 1047-1048 (1980). Contrariwise, the Third Circuit here ruled in favor of NRDC, and against petitioners EPA and CMA, holding that § 301(1) forbids the issuance of FDF variances for toxic pollutants. National Assn. of Metal Finish *125 ers v. EPA, 719 F. 2d 624 (1983). We granted certiorari to resolve this conflict between the Courts of Appeals and to decide this important question of environmental law. 466 U. S. 957 (1984). We reverse.

II

Section 301(1) states that EPA may not “modify any requirement of § 301 insofar as toxic materials are concerned. EPA insists that §301(1) prohibits only those modifications expressly permitted by other provisions of §301, namely, those that § 301(c) and § 301(g) would allow on economic or water-quality grounds. Section 301(1), it is urged, does not address the very different issue of FDF variances. This view of the agency charged with administering the statute is entitled to considerable deference; and to sustain it, we need not find that it is the only permissible construction that EPA might have adopted but only that EPA’s understanding of this very “complex statute” is a sufficiently rational one to preclude a court from substituting its judgment for that of EPA. Train v. NRDC, 421 U. S. 60, 75, 87 (1975); see also Chevron U. S. A. Inc. v. NRDC, 467 U. S. 837 (1984). Of course, if Congress has clearly expressed’ an intent contrary to that of the Agency, our duty is to enforce the will of Congress. Chevron, supra, at 843, n. 9; SEC v. Sloan, 436 U. S. 103, 117-118 (1978).

A

NRDC insists that the language of § 301(1) is itself enough to require affirmance of the Court of Appeals, since on its face it forbids any modifications of the effluent limitations that EPA must promulgate for toxic pollutants. If the word “modify” in §301(1) is read in its broadest sense, that is, to encompass any change or alteration in the standards, NRDC is correct. But it makes little sense to construe the section to forbid EPA to amend its. own standards, even to correct an error or to impose stricter requirements. Furthermore, *126reading § 301(1) in this manner would forbid what § 307(b)(2) expressly directs: EPA is there required to “revise” its pretreatment standards “from time to time, as control technology, processes, operating methods, or other alternatives change.” As NRDC does and must concede, Tr. of Oral Arg. 25-26, § 301(1) cannot be read to forbid every change in the toxic waste standards. The word “modify” thus has no plain meaning as used in §301(1), and is the proper subject of construction by EPA and the courts. NRDC would construe it to forbid the kind of alteration involved in an FDF variance, while the Agency would confine the section to prohibiting the partial modifications that § 301(c) would otherwise permit. Since EPA asserts that the FDF variance is more like a revision permitted by § 307 than it is like a § 301(c) or (g) modification, and since, as will become evident, we think there is a reasonable basis for such a position, we conclude that the statutory language does not foreclose the Agency’s view of the statute. We should defer to that view unless the legislative history or the purpose and structure of the statute clearly reveal a contrary intent on the part of Congress. NRDC submits that the legislative materials evince such a contrary intent. We disagree.

B

The legislative history of § 301(1) is best understood in light of its evolution. The 1972 amendments to the Act added § 301(c), which allowed EPA to waive BAT and pretreatment requirements on a case-by-case basis when economic circumstances justified such a waiver. Pub. L. 92-500, 86 Stat. 845. In 1977, the Senate proposed amending § 301(c) by prohibiting such waivers for toxic pollutants. See S. 1952, 92d Cong., 2d Sess., 30 (1977), Leg. Hist. 584,14 and S. Rep. *127No. 95-370, p. 44 (1977), Leg. Hist. 677. At the same time, the Senate bill added what became § 301(g), which allowed waivers from BAT and pretreatment standards where such waivers would not impair water quality, but which, like § 301(c), prohibited waivers for toxic pollutants. S. 1952, at 28-29, Leg. Hist. 582-583.15 The bill did not contain § 301(1). That section was proposed by the Conference Committee, which also deleted the toxic pollutant prohibition in § 301(c) and redrafted § 301(g) to prohibit water-quality waivers for conventional pollutants and thermal discharges as well as for toxic pollutants.16 While the Conference Committee Report did not explain the reason for proposing §301(1), Representative Roberts, the House floor manager, stated:

“Due to the nature of toxic pollutants, those identified for regulation will not be subject to waivers from or modification of the requirements prescribed under this section, specifically, neither section 301(c) waivers based on the economic capability of the discharger nor 301(g) waivers based on water quality considerations shall be available.” Leg. Hist. 328-329 (emphasis added).

Another indication that Congress did not intend to forbid FDF waivers as well as §§ 301(c) and (g) modifications is its silence on the issue. Under NRDC’s theory, the Conference Committee did not merely tinker with the wording of the Senate bill, but boldly moved to eliminate FDF variances. But if that was the Committee’s intention, it is odd that the *128Committee did not communicate it to either House, for only a few months before we had construed the Act to permit the very FDF variance NRDC insists the Conference Committee was silently proposing to abolish. In E. I. du Pont de Nemours & Co. v. Train, 430 U. S. 112 (1977), we upheld EPA’s class and category effluent limitations, relying on the availability of FDF waivers. Id., at 128. Congress was undoubtedly aware of Du Pont, 17 and absent an expression of legislative will, we are reluctant to infer an intent to amend the Act so as to ignore the thrust of an important decision. Edmonds v. Compagnie Generate Transatlantique, 443 U. S. 256, 266-267 (1979).18

*129NRDC argues that Congress’ discussion of the Act’s provisions supports its position. Several legislators’ comments seemed to equate “modifications” with “waivers” or “variances.”19 Many of these statements, however, came in the specific context of discussing the “waiver” provisions of §§ 301(c) and (g), not the prohibition in §301(1). See, e. g., 123 Cong. Rec. 39183-39184 (1977), Leg. Hist. 458, 461 (Sen. Muskie); 123 Cong. Rec. 38961 (1977), Leg. Hist. 331 (Rep. Roberts); S. Rep. No. 95-370, pp. 40-44, Leg. Hist. 673-677 (discussing water-quality based modifications). Simply because Members of Congress or Committees referred to modifications authorized by §§ 301(c) and (g) as “variance” provisions, does not mean that FDF variances are also modifications barred by §301(1).

After examining the wording and legislative history of the statute, we agree with EPA and CMA that the legislative history itself does not evince an unambiguous congressional intention to forbid all FDF waivers with respect to toxic materials. Chevron, 467 U. S., at 842-843, and n. 9.

C

Neither are we convinced that FDF variances threaten to frustrate the goals and operation of the statutory scheme set *130up by Congress. The nature of FDF variances has been spelled out both by this Court and by the Agency itself. The regulation explains that its purpose is to remedy categories which were not accurately drawn because information was either not available to or not considered by the Administrator in setting the original categories and limitations. 40 CFR §403.13(b) (1984). An FDF variance does not excuse compliance with a correct requirement, but instead represents an acknowledgment that not all relevant factors were taken sufficiently into account in framing that requirement originally, and that those relevant factors, properly considered, would have justified — indeed, required — the creation of a subcategory for the discharger in question. As we have recognized, the FDF variance is a laudable corrective mechanism, “an acknowledgment that the uniform . . . limitation was set without reference to the full range of current practices, to which the Administrator was to refer.” EPA v. National Crushed Stone Assn., 449 U. S. 64, 77-78 (1980). It is, essentially, not an exception to the standard-setting process, but rather a more fine-tuned application of it.20

We are not persuaded by NRDC’s argument that granting FDF variances is inconsistent with the goal of uniform effluent limitations under the Act. Congress did intend uniformity among sources in the same category, demanding that “similar point sources with similar characteristics . . . meet similar effluent limitations,” S. Rep. No. 92-1236, p. 126 (1972). EPA; however, was admonished to take into account the diversity within each industry by establishing appropriate subcategories. Leg. Hist. 455.

*131NRDC concedes that EPA could promulgate rules under §307 of the Act21 creating a subcategory for each source which is fundamentally different from the rest of the class under the factors the EPA must consider in drawing categories. The same result is produced by the issuance of an FDF variance for the same failure properly to subdivide a broad category.22 Since the dispute is therefore reduced to an argument over the means used by EPA to define subcategories of indirect dischargers in order to achieve the goals of the Act, these are particularly persuasive cases for deference to the Agency’s interpretation. Cf. Vermont Yankee Nuclear Power Corp. v. NRDC, 435 U. S. 519, 543 (1978); NLRB v. Bell Aerospace Co., 416 U. S. 267, 293 (1974).

NRDC argues, echoing the concern of the Court of Appeals below, that allowing FDF variances will render meaningless the § 301(1) prohibition against modifications on the basis of economic and water-quality factors. That argument ignores the clear difference between the purpose of FDF waivers and that of §§ 301(c) and (g) modifications, a difference we explained in National Crushed Stone. A discharger that satisfies the requirements of § 301(c) qualifies for a variance “simply because [it] could not afford a compliance cost that is not fundamentally different from those the Administrator has already considered” in creating a category and setting an effluent limitation. 449 U. S., at 78. A § 301(c) modification forces “a displacement of calculations already performed, not because those calculations were incomplete or had unexpected effects, but only because the costs happened to fall on *132one particular operator, rather than on another who might be economically better off.” Ibid. FDF variances are specifically unavailable for the grounds that would justify the statutory modifications. 40 CFR §§403.13(e)(3) and (4) (1984). Both a source’s inability to pay the foreseen costs, grounds for a § 301(c) modification, and the lack of a significant impact on water quality, grounds for a § 301(g) modification, are irrelevant under FDF variance procedures. Ibid.; see also Crown Simpson Pulp Co. v. Costle, 642 F. 2d 323 (CA9), cert. denied, 454 U. S. 1053 (1981).

EPA and CMA point out that the availability of FDF variances makes bearable the enormous burden faced by EPA in promulgating categories of sources and setting effluent limitations. Acting under stringent timetables,23 EPA must collect and analyze large amounts of technical information concerning complex industrial categories.24 Understand*133ably, EPA may not be apprised of and will fail to consider unique factors applicable to atypical plants during the categorical rulemaking process, and it is thus important that EPA’s nationally binding categorical pretreatment standards for indirect dischargers be tempered' with the flexibility that the FDF variance mechanism offers, a mechanism repugnant to neither the goals nor the operation of the Act.25

*134HH HH HH

Viewed in its entirety, neither the language nor the legislative history of the Act demonstrates a clear congressional intent to .forbid EPA’s sensible variance mechanism for tailoring the categories it promulgates. In the absence of a congressional directive to the contrary, we accept EPA’s conclusion that § 301(1) does not prohibit FDF variances. That interpretation gives the term “modify” a consistent meaning in §§ 301(c), (g), and (1), and draws support from the legislative evolution of §301(1) and from congressional silence on whether it intended to forbid FDF variances altogether and thus to obviate our decision in Du Pont.

Here we are not dealing with an agency’s change of position with the advent of a different administration, but rather with EPA’s consistent interpretation since the 1970’s.26 NRDC argues that its construction of the statute is better supported by policy considerations. But we do not sit to judge the relative wisdom of competing statutory interpretations. Here EPA’s construction, fairly understood, is not inconsistent with the language, goals, or operation of the Act. Nor does the administration of EPA’s regulation undermine the will of Congress.27

The judgment of the Court of Appeals is reversed.

It is so ordered.

Justice Marshall,

with whom Justice Blackmun and Justice Stevens join, and with whom Justice O’Connor joins as to Parts I, II, and III,

dissenting.

In these cases, the Environmental Protection Agency (EPA) maintains that it may issue, on a case-by-case basis, individualized variances from the national standards that limit the discharge of toxic water pollutants. EPA asserts *135this power in the face of a provision of the Clean Water Act that expressly withdraws from the agency the authority to “modify” the national standards for such pollutants. The Court today defers to EPA’s interpretation of the Clean Water Act even though that interpretation is inconsistent with the clear intent of Congress, as evidenced by the statutory language, history, structure, and purpose. I had not read our cases to permit judicial deference to an agency’s construction of a statute when that construction is inconsistent with the clear intent of Congress.

I

The Clean Water Act requires the EPA Administrator to regulate two types of industrial facilities: (1) “direct” dischargers, i. e., facilities that discharge waste water directly into navigable waters; and (2) “indirect” dischargers, i. e., facilities that discharge waste water into publicly owned treatment works prior to discharge into navigable waters. For both types of requirements, EPA conducts rulemaking proceedings and promulgates nationwide, categorical limitations, that is, limitations applicable to categories of discharg-ers (e. g., iron and steel plants).

The Act provides for the phased implementation of progressively more stringent requirements for direct discharg-ers. By July 1, 1977, existing direct dischargers were required to meet effluent limitations based on the “best practicable control technology currently available” (BPT). § 301(b)(1)(A), 86 Stat. 844, 33 U. S. C. § 1311(b)(1)(A). By July 1, 1984, such dischargers were obligated to meet limitations based on the “best available technology economically achievable” (BAT). § 301(b)(2)(A).1

Indirect dischargers are subject to “pretreatment” standards applicable to pollutants, including toxic pollutants, that *136are not susceptible to treatment by or would interfere with the operation of public treatment facilities. § 307(b). Pursuant to a consent decree, EPA has set limitations on existing indirect dischargers using the same two-phase scheme used for direct dischargers. See ante, at 119. Thus, pretreatment standards for existing indirect dischargers are set by reference to BPT and BAT levels.

In 1978, EPA issued pretreatment regulations that contained a variance provision for “fundamentally different factors” (FDF). See 43 Fed. Reg. 27757 (1978). An FDF variance is a case-by-case adjustment of the relevant nationwide standard. See 40 CFR §403.13(b)(1) (1984). A discharger may obtain such a variance if the factors relating to its discharges are fundamentally different from those taken into account by EPA in setting the nationwide standard. § 403.13(c)(ii).

In a petition for review filed in the Court of Appeals for the Third Circuit, respondent NRDC challenged the FDF variance provision on two grounds. First, it argued that EPA lacked the inherent authority to issue such variances. Second, it argued that even if, in general, EPA had the authority to grant such variances, it could not do so in the case of toxic pollutants, because § 301(1), which was enacted as part of the 1977 amendments to the Act, bans all “modifications” from the toxic standards. The Third Circuit agreed with the latter argument, holding that § 301(1) prohibits FDF variances in the case of toxic pollutants. National Assn. of Metal Finishers v. EPA, 719 F. 2d 624, 644-646 (1983).2 The *137court remanded the variance provision back to EPA without considering the question of EPA’s inherent authority to grant such variances.3

EPA advances — and the Court defers to — two independent statutory constructions in support of its position that § 301(1) does not ban FDF variances from the toxic standards. First, EPA argues that § 301(1) prohibits only modifications otherwise expressly allowed by two other statutory provisions — §§ 301(c) and (g) — and thus does not apply to FDF variances, which are nonstatutory. The plain meaning of §301(1), the changes made prior to enactment to the bill containing this provision, and the clearly expressed congressional objectives in enacting §301(1) — to deal vigorously and comprehensively with the extremely serious environmental problem caused by toxic pollutants — establish that this provision’s scope was meant to be considerably broader than that attributed to it by EPA. As part of its effort to strengthen the control of toxic pollutants, Congress clearly intended to prohibit all exceptions to the nationwide, categorical standards.

Second, in a strained attempt to characterize the challenged variances in a way that would bring them outside the scope of the § 301(1) prohibition, EPA contends that the case-by-case FDF variance procedure provides a permissible alternative to the statutory mechanism for “revising” standards. The Court defers to this argument, and in so doing, it ignores the relevance of the central feature of the 1972 amendments to the Act — that Congress pointedly determined that water pollution control standards should take the form of general rules, to apply uniformly to categories of dischargers. As a result, the Court validates outcomes substantially less protective of the environment than those *138mandated by Congress. The only view of FDF variances consistent with the scheme of the Clean Water Act is that they are individual exceptions that soften the hardship of general rules. As such, they are undoubtedly disallowed by §301(1).

These cases are not about whether exceptions are useful adjuncts to regulatory schemes of general applicability. That is a policy choice on which courts should defer to Congress in the first instance, and to the administrative agency in the absence of a clear congressional mandate. Here, Congress has made the policy choice. It has weighed competing goals and determined that, whatever the general merits of exceptions schemes, they are simply inappropriate in the context of the control of toxic water pollution. As a result, an exceptions scheme such as the one challenged here simply cannot stand.

II

I first consider EPA’s argument that §301(1) proscribes only those modifications otherwise authorized by §§ 301(c) and (g). Under these provisions, EPA can “modify” the categorical standard if a discharger makes an adequate showing that such a standard is not within the discharger’s economic capability and that a less stringent standard would nonetheless result in reasonable environmental progress, § 301(c),4 or that a less stringent standard adequately pro*139tects water quality, § 301(g).5 This limited view of §301(l)’s scope is clearly inconsistent with congressional intent; the plain meaning of the statute and its legislative history show a clear congressional intent to ban all “modifications.”

A

Section 301(1) provides:

“The Administrator may not modify any requirement of this section as it applies to any specific pollutant which is on the toxic pollutant list under section 307(a)(1) of this Act.” 91 Stat. 1590, 33 U. S. C. § 1311(Z).

The statute does not define either “modify” or “modification.” The phrase “may not modify any requirement,” however, expressly proscribes all “modifications” of the standards for toxics. Nothing on the face of the statute suggests that Congress intended that qualifying language be read into this prohibition. On the contrary, the prohibition is unqualified.

EPA’s argument that § 301(1) bans only those modifications otherwise authorized by §§ 301(c) and (g) is therefore incon*140sistent with the plain meaning of the statute. By its terms, the statutory prohibition has universal scope, not the limited scope attributed to it by EPA.

B

Moreover, the legislative history demonstrates that Congress meant what it said, and it evidences a clear congressional intent to ban all “modifications.” First, the legislative history firmly establishes that § 301(1) was enacted as part of a program to deal effectively and comprehensively with the problem of toxic pollutants, and that its prohibition was an integral part of this program. Under any canon of statutory construction, the congressional purposes in enacting a provision would be deemed relevant to the question of the scope of that provision, but the Court simply fails to discuss this issue.

In 1977, when it enacted the amendments to the Clean Water Act containing §301(1), Congress regarded the problem of toxic pollution as a very serious one. For example, Senator Muskie, the major drafter and Senate manager of the bill containing § 301(1), remarked:

“The seriousness of the toxics problem is just beginning to be understood. New cases are reported each day of unacceptable concentrations of materials in the aquatic environment, in fish and shellfish, and even in mother’s milk. Empirical evidence has shown a statistical correlation between materials in New Orleans’ drinking water and cancer mortality rates; Repone has destroyed the James River, one of America’s most productive, ánd most historic rivers; PCB’s are pervasive and have ruined the fishing in the Hudson River and the Great Lakes; carbon tetrachloride is only the most recent material to contaminate the Ohio River; the *141pesticide endrin has been found in Mississippi; perhaps worst of all, are the ones we do not know yet.

“The more we find out, the more cause there is for concern. It is imperative that these materials be controlled.” 123 Cong. Rec. 39181 (1977), Legislative History of the Clean Water Act of 1977, p. 454 (1978) (1977 Leg. Hist.).6

Similarly, Representative Roberts, the House manager of the bill, stated:

“[Toxics] have not only polluted drinking water and destroyed both commercial and sport fishing, but in many major water bodies they also constitute a hazard to aquatic environment and public health that has yet to be fully recognized.” 123 Cong. Rec. 38960 (1977), 1977 Leg. Hist. 327.

See also 1977 Leg. Hist. 334 (House Subcommittee memorandum).

The primary purpose of the 1977 amendments was to strengthen the regulation “of the increasingly evident toxic hazard.” 123 Cong. Rec. 38960 (1977), 1977 Leg. Hist. 326 (Rep. Roberts). See also 123 Cong. Rec. 39219 (1977), 1977 Leg. Hist. 549 (Sen. Moynihan) (“There is no room for compromise here: toxics must be controlled”). The §301(1) ban on “modifications” was an integral part of this effort to make the environment safe from toxics, and through it, Congress sought to prevent any weakening of the categorical standards for the control of toxic pollutants. It is clear that Congress knew full well what effects the rule might have on industry, and that it went forward nonetheless. For example, the legislators were aware that the prohibition against *142“modifications” of the standards for toxic pollutants could lead to “new regulations more restrictive than any previously contemplated.” 123 Cong. Rec. 38993 (1977), 1977 Leg. Hist. 411 (Rep. Buchanan). Congress also realized that such regulations would cost industry “millions of dollars and result only in a little more clean-up of our waters.” 123 Cong. Rec. 38952 (1977), 1977 Leg. Hist. 305 (Rep. Roberts). But Congress found that for toxics, unlike for other pollutants, ibid., such high costs of pollution control were justified in view of the serious environmental dangers at stake. Cf. §502(13) (defining “toxic” pollutants as pollutants that “cause death, disease, behavioral abnormalities, cancer, genetic mutations, physiological malfunctions (including malfunctions in reproduction) or physical deformations”).

It is readily apparent that a complete ban on modifications would most directly and completely accomplish the congressional goal. EPA offers no evidence in the legislative history to explain why this goal would be promoted by banning the statutory modifications of §§ 301(c) and (g), but would not more effectively be advanced by banning other modifications as well. It points to no evidence that Congress singled out the §§ 301(c) and (g) modifications as more pernicious from the standpoint of an effective toxic control program than modifications based on other factors. In fact, the statutory scheme suggests that the converse is true, as Congress specifically provided for statutory exemptions in these areas but not in other areas.

In the case of § 301(c), Congress was aware that certain firms would be driven to bankruptcy if they were required to comply strictly with the categorical standards. Congress determined that avoiding bankruptcies was an important social goal, and one that was not automatically outweighed by the goal of protecting the environment. Section 301(c) reflects the tension between these two goals: As long as a firm can make reasonable pollution control progress, it will not be driven to bankruptcy by its inability to meet higher pollution control standards.

*143Similarly, in the case of § 301(g) water-quality modifications, Congress decided not to force dischargers to meet standards higher than those that could be justified by legitimate environmental considerations. Thus, as long as a discharge did not interfere with the attainment of adequate water quality, a discharger would not be forced to expend additional resources in pollution control merely because a higher standard was “economically achievable.” Cf. 123 Cong. Rec. 38960 (1977), 1977 Leg. Hist. 326 (Rep. Roberts).

If these two modifications are the only ones now prohibited, the result is wholly counterintuitive. EPA is in effect contending that economic and water-quality factors present the most compelling case for modification of the standard in the nontoxic context — as they are explicitly authorized by statute — but the least compelling case for modification in the toxic context — as they are the only modifications prohibited by §301(1). As might be expected, EPA does not present any theory, much less a logical argument, or evidence in the legislative history, to support this extremely inconsistent result.

Moreover, if Congress had not intended to prohibit all modifications, it would almost certainly either have defined explicitly the scope of permissible modifications, or given the agency some guidance on how to go about doing so. Only in this way would Congress have had any assurance that modifications would be allowed only when they promoted interests of sufficient importance to outweigh Congress’ foremost goal of protecting the environment against toxic pollution.

C

The changes made in conference to the 1977 amendments, which ultimately included §301(1), provide further support for the proposition that Congress did not intend to limit §301(1) in the way suggested by EPA. Of the three provisions that undergird EPA’s theory — subsections (c), (g), and (1) of § 301 — only subsection (c) was adopted before the 1977 *144amendments, as part of the 1972 amendments. See 33 U. S. C. § 1311(c). The 1977 Senate bill contained two provisions of interest here. First, the bill proposed amending subsection (c) to prohibit, in the case of toxic pollutants, variances based on economic factors. S. 1952, 95th Cong., 1st Sess., §26(c) (1977), 1977 Leg. Hist. 584. Second, the Senate proposed what ultimately became subsection (g), which authorized modifications that did not interfere with water-quality goals. Like the proposed amendment to subsection (c), subsection (g) prohibited modifications in the case of toxic pollutants. The Senate bill did not contain subsection (1).

The Conference Committee changed the Senate bill in three relevant ways. First, it took out of subsection (c) the ban against modifications for toxics. Second, it reworded subsection (g) to prohibit water-quality modifications for conventional pollutants and for all thermal discharges, but it left unaffected the Senate bill’s prohibition against modifications for toxic pollutants. Third, it added subsection (1), which creates a ban of general applicability on modifications for toxic pollutants.

In explaining these changes, petitioner CMA contends that during the Conference Committee deliberations, “it was decided that, rather than repeating the identical limiting clause [for toxic pollutants] at the end of § 301(c) and what had become § 301(g) of the Act, the limitation would be put into a separate §301(1).” Brief for Petitioners in No. 83-1013, pp. 29-30. The debates of the Conference bill do not suggest that such a thing was “decided”; in fact, the reasons for the changes are not discussed at all. Moreover, if cleaning up the statutory language was in fact the objective of the changes, the Conference Committee was remarkably unsuccessful at doing so. Indeed, while the Committee took the prohibition against toxic modifications out of subsection (c), it left this prohibition undisturbed in subsection (g). Thus, the language of the Act simply belies CMA’s explanation.

*145More importantly, the wording of §301(1) strongly suggests that the purpose of the change was not to improve the style of the statute, but to expand the scope of the prohibition against “modifications.” Indeed, there is an important difference in the wording of subsections (c), (g), and (1). Subsections (c) and (g), which authorize exceptions, apply by their terms only to modifications of “the requirements of subsection (b)(2)(A).”7 If the Conference Committee was attempting merely to consolidate the bans on modifications of toxic standards, then it would similarly have limited the applicability of subsection (1) to subsection (b)(2)(A) requirements. Instead, subsection (1) applies to “any requirement of this section,” which includes numerous standards in addition to those of subsection (b)(2)(A).8

In fact, it appears that EPA once agreed that the changes made in conference expanded the scope of the ban on “modifications.” In the past, EPA construed §301(1) to prohibit, in the case of toxics, not only subsection (c) and (g) modifications, but also modifications from secondary treatment standards otherwise authorized by subsection (h), Brief for EPA on Petition to Enforce Mandate and Petitions for Review 24 in Appalachian Power Co. v. Train, 620 F. 2d 1040 (CA4 1980). Cf. FMC v. Seatrain Lines, Inc., 411 U. S. 726, 745 *146(1973) (administrative interpretation entitled to additional deference if “longstanding”).

In summary, the Conference changes provide further support for a broad reading of §301(1). See FTC v. Raladam Co., 283 U. S. 643, 648 (1931). The Court, however, appears to draw the opposite conclusion. But in doing so, it completely ignores the difference in the scope of §§ 301(c) and (g) on the one hand, and §301(1) on the other, and instead rests on an explanation of congressional activity that in fact explains almost nothing. See ante, at 126-127.

D

The Court and EPA both attach great importance to the congressional silence regarding FDF variances. EPA argues that E. I. du Pont de Nemours & Co. v. Train, 430 U. S. 112 (1977), held that FDF variances are “appropriate.” According to EPA, if Congress had intended to reverse this result it would have made its intention clear. See Brief for EPA 28-29. This contention, which the Court finds persuasive, see ante, at 127-128, is based on a misunderstanding of what was at stake in Du Pont. That case did not authorize the issuance of variances in any context that is relevant here.

Du Pont involved a challenge to EPA’s authority to issue, to direct dischargers, categorical effluent limitations for BPT and BAT. The Court had little difficulty in upholding such categorical limitations in the BAT context, as the statute provided that the limitations be set for “categories and classes” of dischargers, § 301(b)(1)(B). See Du Pont, supra, at 127. In contrast, the statute provided that BPT limitations be set for “point sources.” § 301(b)(1)(A). Several chemical manufacturers argued that, given this language, individualized BPT limitations were necessary, and that regulation by categories and classes of dischargers was inappropriate. This Court rejected the industry’s challenge, holding that BPT *147limitations could be set by industrywide regulation, so long as some allowance — such as FDF variances — was made for variations in individual plants. 430 U. S., at 128.

In support of its position that the Court broadly endorsed the issuance of FDF variances and that the congressional silence is noteworthy, EPA cites as dispositive one sentence in the opinion, which reads:

“We conclude that the statute authorizes the 1977 limitations [BPT] as well as the 1983 limitations [BAT] to be set by regulation, so long as some allowance is made for variations in individual plants, as EPA has done by including a variance clause in its 1977 limitations.” Ibid.

Only by taking this sentence out of context can one find support for the proposition that Du Pont requires FDF variances from BAT limitations, just as it does in the case of BPT limitations.9 When read in context, the sentence cited by EPA clearly means that BPT standards, like BAT standards, can be set by regulation, but if EPA does so in the BPT context, it must allow for variances. Indeed, the Court had earlier concluded that “§301 unambiguously provides for the use of regulations to establish the [BAT] effluent limitations.” Du Pont, supra, at 127. The Court did not qualify this conclusion in any way, but instead went on to discuss the BPT problem. The sentence that EPA refers to comes at the end of the discussion of BPT limitations, and is thus logically related to that discussion.

*148Furthermore, the Court upheld the regulations challenged in Du Pont even though they did not contain an FDF variance clause for BAT limitations. See 430 U. S., at 123,127.10 If the sentence in question has the meaning that EPA now ascribes to it, the Court would presumably have had to reverse on that point.

In summary, the portion of Du Pont on which EPA relies, has absolutely no bearing on the question of whether FDF variances are “appropriate” — to use the language employed by EPA, see n. 9, supra — when the statute calls for limitations for categories or classes of dischargers. See EPA v. National Crushed Stone Assn., 449 U. S. 64, 72 (1980) (“[Du Pont] indicated that a variance provision was a necessary aspect of BPT limitations applicable by regulation to classes and categories of point sources”); id., at 73, n. 12 {“[DuPont] held that a uniform BPT limitation must contain a variance provision, if it is to be valid”). Both the facts and the rationale of this portion of Du Pont are of relevance only to cases in which EPA issues categorical standards in the face of a statutory scheme that calls for regulation of “point sources.”

This distinction is of crucial significance because the standards for toxic pollutants, like all BAT and pretreatment standards, are to be set not for “point sources,” but instead “for the applicable category or class of point sources.” § 307(a)(2) (emphasis added) (toxics); see also § 301(b)(2)(A) (BAT); § 307(b)(3) (pretreatment). Du Pont did not consider whether such standards are necessary, or even appropriate, in this context.11 We should scarcely attribute any signifi-*149canee to the legislative failure to discuss Du Pont because Du Pont considered a fundamentally different scheme of regulation. It may be that one day the Clean Water Act will be read to permit, for nontoxic pollutants, FDF variances from BAT and pretreatment standards; however, there is no reason why Congress should have said anything in 1977, when it enacted § 301(1), about a legal development that has not yet taken place, eight years later.

There is, moreover, another reason for the legislative silence on FDF variances. The legislative history of the 1977 amendments shows that Congress believed — correctly, as it turns out — that the courts had not yet determined whether FDF variances were permissible in the BAT context. See S. Rep. No. 95-370 (1977), 1977 Leg. Hist. 674.12 Only by misreading Du Pont and ignoring the relevant legislative history can the Court say that Du Pont “construed the Act to permit the very FDF variance NRDC insists the Conference Committee was silently proposing to abolish.” Ante, at 128.13

*150E

EPA also relies heavily on a statement by Representative Roberts:

“Due to the nature of toxic pollutants, those identified for regulation will not be subject to waivers from or modification of the requirements prescribed under this section, specifically, neither section § 301(c) waivers based on the economic capability of the discharger nor 301(g) waivers based on water quality considerations shall be available.” 123 Cong. Rec. 38960 (1977), 1977 Leg. Hist. 328-329 (emphasis added).

However, other statements in the debates fail similarly to restrict the scope of the provision. For example, Senator Muskie stated:

“Like toxic pollutants for which there are no waivers or modifications, there are no potential waivers or modifications for conventional pollutants.” 123 Cong. Rec. 39183 (1977), 1977 Leg. Hist. 458 (emphasis added).

See also 123 Cong; Rec. 38952 (1977), 1977 Leg. Hist. 305 (“Strict requirements are still in effect for damaging pollutants, such as toxics. However, for certain other pollutants, industry may get a waiver”) (Rep. Roberts); 123 Cong. Rec. 38993 (1977), 1977 Leg. Hist. 411 (referring to “denial of any waiver” with respect to toxics) (Rep. Buchanan) (emphasis added).

Taken as a whole, the legislative history firmly supports the plain meaning of the statute, namely, that § 301(1) bans all *151“modifications,” and not just those otherwise permitted by §§ 301(c) and (g). EPA’s strongest argument in support of its position on this score is that, during the course of debates, one of the bill’s managers used the word “specifically” instead of “for example. ” Under any accepted canon of construction, this choice of words is insufficient to overcome the other, more probative indications of congressional intent that emerge from an analysis of the legislative history. And, with the language and the legislative history pointing so definitely in the same direction, there can be no doubt that congressional intent was clear.

F

The determination that Congress clearly intended that § 301(1) do more than just ban modifications otherwise permitted by §§ 301(c) and (g) compels the conclusion that EPA’s construction to the contrary cannot stand. As this Court has repeatedly stated:

“The interpretation put on the statute by the agency charged with administering it is entitled to deference, but the courts are the final authorities on issues of statutory construction. They must reject administrative constructions of the statute, whether reached by adjudication or by rulemaking, that are inconsistent with the statutory mandate or that frustrate the policy that Congress sought to implement.” FEC v. Democratic Senatorial Campaign Committee, 454 U. S. 27, 31-32 (1981) (citations omitted).

See also SEC v. Sloan, 436 U. S. 103, 117-118 (1978); FMC v. Seatrain Lines, Inc., 411 U. S., at 745-746; Volkswagenwerk v. FMC, 390 U. S. 261, 272 (1968); NLRB v. Brown, 380 U. S. 278, 291 (1965); Social Security Board v. Nierotko, 327 U. S. 358, 369 (1946); Burnet v. Chicago Portrait Co., 285 U. S. 1, 16 (1932); Webster v. Luther, 163 U. S. 331, 342 (1896).

*152This case is thus unlike Chevron U. S. A. Inc. v. NRDC, 467 U. S. 837 (1984), on which the Court and EPA rely. In Chevron, the Court reviewed an EPA regulation that treated all pollution-emitting devices within the same industrial grouping as though they were encased within a single “bubble.” This regulation was challenged on the ground that it was not based on a proper construction of the statutory term “stationary source.” Analyzing the statutory language, the Court concluded that “parsing of general terms in the text of the statute” would not reveal the actual intent of Congress. Id., at 861. Similarly, it found the legislative history “unilluminating.” Id., at 862. Given these two conclusions, the Court determined that deference to the Agency’s reasonable interpretation was appropriate.

Chevron’s deference requirement, however, was explicitly limited to cases in which congressional intent cannot be discerned through the use of the traditional techniques of statutory interpretation. Indeed, Chevron reaffirmed the principle that “[t]he judiciary is the final authority on issues of statutory construction and must reject administrative constructions which are contrary to clear congressional intent.” Id., at 843, n. 9.14

My disagreement with the Court does not center on its reading of Chevron, but instead on its analysis of the congressional purposes behind §301(1). If I agreed with the Court’s analysis of the statute and the legislative history, I too would conclude that Chevron commands deference to the administrative construction.

EPA’s second construction of the statutory scheme is, on the surface, a more plausible one. EPA argues that FDF *153variances do not excuse compliance with the correct standards, but instead provide a means for setting more appropriate standards. It is clear that, pursuant to § 307(b)(2), EPA can “revise” the pretreatment standards, as long as it does so “following the procedure established ... for the promulgation of such standards.” The statute contemplates that the standards will be set and revised through notice-and-comment rulemaking and will be applicable to categories of sources. See §§ 307(b)(2), (3); see also Brief for EPA 9. EPA argues that such a “revision,” which is clearly not proscribed by § 301(1), would be substantively indistinguishable from an FDF variance. Thus, according to the Agency, NRDC’s concern stems not from the result achieved when an FDF variance is granted, but rather from the procedure employed in reaching that result. EPA relies on Vermont Yankee Nuclear Power Corp. v. NRDC, 435 U. S. 519 (1978), for the proposition that an agency is free to choose between two procedures for reaching the same substantive ends. See Brief for EPA 11, 36.

To support its argument, EPA points out that the factors that may justify an FDF variance are the same factors that may be taken into account in setting and revising the national pretreatment standards. Compare § 304(b)(2) (statutory standard) with 40 CFR §403.13(d) (1984) (FDF variance provision). EPA also points out that, in considering whether an FDF variance will be granted, it cannot take into account factors that could not have justified a change in the national standards. See Brief for EPA 31; 40 CFR §403.13(e)(1984). EPA acknowledges that the statute requires that the national pretreatment standards be established — and therefore revised — for “categories” of dis-chargers, see § 307(b)(3) (pretreatment standards); Brief for EPA 11; see also § 307(a)(2) (toxic standards), and not on a case-by-case basis. It argues, however, that nothing in the Clean Water Act precludes EPA from defining a subcategory that has only one discharger. See Brief for EPA 31.

*154The logic of EPA’s position is superficially powerful. If EPA can, through rulemaking, define a subcategory that includes only one discharger, why should it not be able to do so through a variance procedure? In fact, if rulemaking and the variance procedure were alternative means to the same end, I might have no quarrel with EPA’s position, which the Court has accepted. Ante, at 132-133. Indeed, “[ajbsent constitutional constraints or extremely compelling circumstances the administrative agencies should be free to fashion their own rules of procedure and to pursue methods of inquiry capable of permitting them to discharge their multitudinous duties.” Vermont Yankee, supra, at 543 (citations omitted); see also SEC v. Chenery Corp., 332 U. S. 194, 202-203 (1947).

However, the Agency’s position does not withstand more than superficial analysis. An examination of the legislative history of the 1972 amendments to the Clean Water Act — the relevance of which both the Court and EPA ignore — reveals that Congress attached great substantive significance to the method used for establishing pollution control requirements.

The Conference Committee Report directed EPA to “make the determination of the economic impact of an effluent limitation on the basis of classes and categories of point sources, as distinguished from a plant by plant determination.” 1972 Leg. Hist. 304 (emphasis added).15 Representative Dingell, one of the House conferees, described this principle as “very important” and stated that “a plant-by-plant determination of the economic impact of an effluent limitation is neither expected, nor desired, and, in fact, it should be avoided.” 118 Cong. Rec. 33758 (1972), 1972 Leg. Hist. 254-255.

*155Similarly, Senator Muskie stated:

“The Conferees intend that the factors described in section 304(b) be considered only within categories and classes of point sources and that such factors not be considered at the time of the application of an effluent limitation to an individual point source within such a category or class.” 118 Cong. Rec. 33697 (1972), 1972 Leg. Hist. 172 (emphasis added).

See also Du Pont, 430 U. S., at 130; American Iron & Steel Institute v. EPA, 526 F. 2d 1027, 1051 (CA3 1975) (“Congress clearly intended that the Administrator consider costs on a class or category basis, rather than on a plant-by-plant basis”) (emphasis added). Moreover, in a letter urging the President to approve the 1972 amendments, William Ruckelshaus, EPA’s Administrator, observed that the Act’s standards should be set “for industrial categories, taking into account processes involved, age of equipment, and cost, considered on a national, industry-unde basis.” 118 Cong. Rec. 36775 (1972), 1972 Leg. Hist. 145 (emphasis added). It is difficult to imagine a legislative history that would make more clear that standards should not be set — and therefore should not be revised — on an individual basis.

The legislative history also makes clear why Congress found it so important that the standards be set for “categories” of dischargers, and not for individual dischargers. Congress intended to use the standards as a means to “force” the introduction of more effective pollution control technology. Thus, Congress directed EPA to establish BPT levels by reference to “the average of the best existing performance by plants of various sizes, ages, and unit processes within each industrial category.” 118 Cong. Rec. 33696 (1972), 1972 Leg. Hist. 169 (Sen. Muskie). In establishing BAT levels, it directed EPA to look at “the best performer in an industrial category.” 118 Cong. Rec. 33696 (1972), 1972 Leg. Hist. 170. By requiring that the standards be set by reference to *156either the “average of the best” or very “best” technology, the Act seeks to foster technological innovation. 118 Cong. Rec. 33696 (1972), 1972 Leg. Hist. 170. See generally La Pierre, Technology-Forcing and Federal Environmental Protection Statutes, 62 Iowa L. Rev. 771, 805-829 (1977); Note, Forcing-Technology: The Clean Air Act Experience, 88 Yale L. J. 1713 (1979).

Unlike the statutory revision mechanism of § 307(b), FDF variances are set not by reference to a category of dis-chargers, but instead by reference to a single discharger. In evaluating an application for a variance, EPA does not look at the group of dischargers in the same position as the applicant, but instead focuses solely on the characteristics of the applicant itself. Under the FDF program, there is no mechanism for EPA to ascertain whether there are any other dis-chargers in that position. Moreover, there is no mechanism for EPA to group together similarly situated dischargers. Quite to the contrary, a scheme in which the initial screening may be done by the individual States, at times determined by when the variance application is filed, is unlikely to lead to the identification of new subcategories. See 40 CFR §403.13(k) (1984).

The FDF variance procedure leads to substantive results that are different in two fundamental ways from those attained through the rulemaking for categories of dischargers contemplated in § 307(b). First, it is less protective of the environment. If, for example, a discharger shows that its production processes — and, as a result, its costs of compliance — are significantly different from those taken into account in setting the categorical standards, that discharger would be eligible for an FDF variance, and EPA could set a new requirement based on the applicant’s peculiar situation. See 40 CFR §§403.13(d)(5), (6) (1984); Tr. of Oral Arg. 14. It may turn out, however, that there are many other discharg-ers in the same situation, and that all of these dischargers use production processes that make pollution control possible *157át a much lower cost. If EPA took into account the production processes of these more efficient dischargers — as it presumably would have to do if it proceeded through rule-making on a categorical scale — it would set a requirement far more stringent than that adopted as part of the FDF variance mechanism.

In the aggregate, if EPA defines a new pretreatment subcategory through rulemaking, the BAT-level pollution control requirement of each discharger would be determined by reference to the capability of the “best” performer. In contrast, if EPA provides individual variances to each plant in this group, only one discharger would have a requirement based on the capability of the best performer — the best performer itself. The others would necessarily be subject to less stringent standards.16

The second important difference is that FDF variances do not spur technological innovation to the same extent as § 307(b) revisions. In the preceding example, the discharger with environmentally unsound production processes would probably be compelled to purchase new technology if it were subjected to a pollution control requirement set by reference to the characteristics of the “best” discharger. Under the less stringent requirement adopted through the FDF variance procedure, it might not need to do so. The additional demand for new technology that results from the § 307(b) procedures creates incentives for technological innovation. In the long run, such innovation would lead to even better technology and to the possibility of further tightening of the pollution control requirements, as such technology became cheaper. In fact, Congress envisioned that this iterative procedure would ultimately lead to an elimination of harmful discharges. See 118 Cong. Rec. 33696 (1972), 1972 Leg. Hist. 170 (Sen. Muskie).

*158It is true, of course, that even the statutory revision procedure might identify a subcategory with only one discharger. That procedure, however, will have established that this discharger is indeed uniquely situated. In contrast, an FDF variance sets an individual requirement even where there may be similarly situated dischargers.

In summary, whatever else FDF variances might do, they do not further the same congressional goals as the notice- and-comment rulemaking required for § 307(b) revisions.17 Vermont Yankee is simply inapposite; Congress intended, for substantive reasons, that the pretreatment standards be set and revised through rulemaking for categories of dischargers.18 The Court’s conclusion to the contrary stems exclusively from its failure to consider why Congress chose to require categorical standards.

> HH

The analysis of Parts II and III compels the conclusion that neither of the alternative arguments advanced to support EPA’s construction of the statute can stand. That analysis *159also leads directly to the conclusion that § 301(1) in fact disallows FDF variances from the standards for toxic pollutants. Congress clearly intended that § 301(1) ban variances such as those at issue here, and the language and legislative history permit no other interpretation.

A

Part II shows that the language of § 301(1), the purposes that led to the adoption of the provision, and the changes made by the Conference Committee, indicate a clear congressional intent to ban all “modifications” to the standards for toxics, not merely those otherwise authorized by §§ 301(c) and (g). The legislative history also establishes that Congress banned “modifications” because it wanted to ensure that the serious problem of toxic pollution not be exacerbated by the granting of exceptions to the general rulemaking standards. See Part II-B, supra.

It is true, of course, that in many cases exceptions serve the important purpose of softening the impact of rules of general applicability. They mediate between demands for comprehensive solutions on the one hand, and individualized application of law on the other. See generally Diver, Policy-making Paradigms in Administrative Law, 95 Harv. L. Rev. 393 (1981).

Exceptions, however, are not without costs. For example, they are inappropriate where small errors could lead to irreversible or catastrophic results.19 In such cases, indi*160vidual equity should give way to comprehensive rationality. See id., at 431-432; Note, Regulatory Values and the Exceptions Process, 93 Yale L. J. 938, 955, and n. 85 (1984).

The decision of when exceptions are required, when they are permissible, and when they are prohibited is, in the first instance, one for Congress to make. It is an administrative decision only where Congress has left a gap for the agency to fill. See Chevron, 467 U. S., at 843-844. In this case, Congress determined that the flexibility resulting from exceptions would interfere with the furtherance of the more important goal of controlling toxic pollution. There is no question that courts should defer to this congressional judgment.

In fact, when Congress has attached great importance to certain environmental goals, we have disallowed exceptions even in the absence of an explicit statutory ban. For example, in TV A v. Hill, 437 U. S. 153 (1978), we reviewed a provision of the Endangered Species Act that required federal agencies “to insure that actions authorized, funded, or carried out by them do not jeopardize the continued existence” of an endangered species or “result in the destruction or modification of habitat of such species . . . .” 16 U. S. C. § 1536 (1976 ed.). Even though Congress had not expressly banned exceptions from the statutory requirements, the Court focused on the quoted language and found that it “admits of no exception.” Id., at 173. It further found that both the language and the legislative history “clearly” showed that Congress viewed the preservation of endangered species as a goal of great importance. Id., at 187-188. In light of this statutory construction, the Court concluded that any exemption from the statute’s requirements — other than exemptions specifically approved by Congress — would be inappropriate.20

*161Similarly, in Du Pont itself, the Court disallowed FDF variances from the Clean Water Act’s standards of performance for new sources, reasoning that such variances would be inconsistent with the environmental goals expressed in the statute and the legislative history. There, the Court stated that FDF variances “would be inappropriate in a standard that' was intended to insure national uniformity and ‘maximum feasible control of new sources.’” 430 U. S., at 138 (citation omitted). In this case, of course, Congress has not only indicated that the environmental goal at stake is extremely important, but it has also explicitly disallowed exceptions. Under such circumstances, it would be especially inappropriate to defer to the Agency’s decision to create exceptions.

B

Part III establishes that FDF variances are not an alternative way of complying with the statutory command to set rules of general applicability. They do not implement the Clean Water Act’s technology-based requirements; instead, like §§ 301(c) and (g) modifications, they are case-by-case departures from such requirements. In fact, in the past, EPA itself has referred to FDF variances as “exception[sj to [a] general rule of applicability.” Brief for EPA 47 in NRDC v. EPA, 537 F. 2d 642 (CA2 1976).

FDF variances not only take the same form as §§ 301(c) and (g) modifications, but they also serve closely analogous functions. As I have discussed, the purpose of exceptions is to soften the harshness of general rules. See supra, at 159. A § 301(c) modification, for example, relieves a firm of its obligation to meet an applicable rule when compliance with *162that rule would place the firm in a serious hardship. See EPA v. National Crushed Stone Assn., 449 U. S., at 78; S. Rep. No. 95-370, p. 41 (1977), 1977 Leg. Hist. 674 (Sen. Muskie); Brief for EPA 32-33. FDF variances also temper — albeit in a slightly different way — the effects of the nationwide, categorical standards. They relieve a firm of its obligation to comply with a rule that would impose on that firm a disproportionate share of the regulatory burden. See Tr. of Oral Arg. 14.21 In fact, EPA itself has characterized FDF variances as “ ‘safety valves’ in regulatory schemes of general applicability.” Brief for EPA 44 in NRDC v. EPA, 537 F. 2d 642 (CA2 1976); see also Hearings on Possible Amendments to the Federal Water Pollution Control Act *163before the Subcommittee on Water Resources of the House Committee on Public Works and Transportation, 98th Cong., 1st Sess., 2706, 2741 (1988) (EPA Administrator Ruckelshaus describing FDF variances as “safety valves”); NRDC v. EPA, 637 F. 2d, at 646 (“[T]he ‘variance’ clause was assertedly adopted as an administrative safety valve”). Thus, FDF variances are exceptions that provide the type of flexibility that § 301(1) sought to ban.22

The Court accepts EPA’s present characterization that FDF variances are a hybrid: “more like” a revision permitted *164by §307 than like a §§ 301(c) and (g) modification. Ante, at 126. But a requirement that, by definition, applies to only one discharger cannot be considered “more like” a rule of general applicability than like an exception to such a rule. Clearly, it is an exception.23

The Court’s error is to overlook the distinction between general rules and exceptions. Instead, it focuses on the differences between the grounds for exceptions provided by §§ 301(c) and (g) on the one hand, and by the FDF provisions on the other. Thus, the Court makes its cuts along an entirely different — and irrelevant — axis. For EPA to prevail, the Court must show that Congress found that exceptions based on economic capability or water-quality factors were especially undesirable. If this were true, then exceptions based on other factors would be less undesirable, and it would make sense to decide the cases on the basis of the extent to which the factors taken into account in granting FDF variances differ from §§ 301(c) and (g) factors. The Court’s position, however, is inconsistent with the clear purpose of §301(1). As I have shown, there is absolutely no reason to believe that this provision was designed to ban §§ 301(c) and (g) modifications because there was something particularly pernicious about such exceptions. See supra, at 143. Rather, the congressional concern was that exceptions would weaken the standards for the control of toxic pollutants. This concern defines the relevant criterion: whether something is a general rule or an exception to such a rule. Sections 301(c) and (g) modifications are at one end of the axis not because they are based on economic or water-*165quality factors, but because they are exceptions to general rules. Section 307(b) revisions are at the other end of the axis not because they are based on factors taken into account in setting the standards, but because they are rules of general applicability. Of course, FDF variances, which are nothing but exceptions to general rules, are at the same end of the axis as §§ 301(c) and (g) modifications.

For the foregoing reasons, it is apparent that §301(1) prohibits FDF variances from the pretreatment standards for toxic pollutants. I therefore dissent.

Justice O’Connor,

dissenting.

I join Parts I, II, and III of Justice Marshall’s dissent. They accurately demonstrate that the Court’s interpretation of §301(1) of the Clean Water Act, 33 U. S. C. §1311(0, is inconsistent with the language of the statute and its legislative history. In my opinion, this alone is sufficient grounds for affirming the judgment of the Court of Appeals. I express no view as to Part IV of the dissent because I think it is not necessary to the disposition of these cases.