FN* The syllabus constitutes no part of the opinion of the Court but has been prepared by the Reporter of Decisions for the convenience of the reader. See United States v. Detroit Lumber Co., 200 U.S. 321, 337, 26 S.Ct. 282, 287, 50 L.Ed. 499.
*690 Justice STEVENS delivered the opinion of the Court.
The Endangered Species Act of 1973 (ESA or Act), 87 Stat. 884, 16 U.S.C. § 1531 (1988 ed. and Supp. V), contains a variety of protections designed to save from extinction species that the Secretary of the Interior designates as endangered or threatened. Section 9 of the Act makes it unlawful for any person to “take” any endangered or threatened species. The Secretary has promulgated a regulation that defines the statute's prohibition on takings to include “significant habitat modification or degradation where it actually kills or injures wildlife.” This case presents the question whether the Secretary exceeded his authority under the Act by promulgating that regulation.
I
Section 9(a)(1) of the Act provides the following protection for endangered species: FN1
FN1. The Act defines the term “endangered species” to mean “any species which is in danger of extinction throughout all or a significant portion of its range other than a species of the Class Insecta determined by the Secretary to constitute a pest whose protection under the provisions of this chapter would present an overwhelming and overriding risk to man.” 16 U.S.C. § 1532(6).
“Except as provided in sections 1535(g)(2) and 1539 of this title, with respect to any endangered species of fish or wildlife listed pursuant to section 1533 of this title it is unlawful for any person subject to the jurisdiction of the United States to-
The Act does not further define the terms it uses to define “take.” The Interior Department regulations that implement the statute, however, define the statutory term “harm”:
FN2. The Secretary, through the Director of the Fish and Wildlife Service, originally promulgated the regulation in 1975 and amended it in 1981 to emphasize that actual death or injury of a protected animal is necessary for a violation. See 40 Fed.Reg. 44412, 44416 (1975); 46 Fed.Reg. 54748, 54750 (1981).
In addition to the prohibition on takings, the Act provides several other protections for endangered species. Section 4, 16 U.S.C. § 1533, commands the Secretary to identify species of fish or wildlife that are in danger of extinction and to publish from time to time lists of all species he determines to *692 be endangered or threatened. Section 5, 16 U.S.C. § 1534, authorizes the Secretary, in cooperation with the States, see § 1535, to acquire land to aid in preserving such species. Section 7 requires federal agencies to ensure that none of their activities, including the granting of licenses and permits, will jeopardize the continued existence of endangered species “or result in the destruction or adverse modification of habitat of such species which is determined by the Secretary ... to be critical.” 16 U.S.C. § 1536(a)(2).
Respondents in this action are small landowners, logging companies, and families dependent on the forest products industries in the Pacific Northwest and in the Southeast, and organizations that represent their interests. They brought this declaratory judgment action against petitioners, the Secretary of the Interior and the Director of the Fish and Wildlife Service, in the United States District Court for the District of Columbia to challenge the statutory validity of the Secretary's regulation defining “harm,” particularly the inclusion of habitat modification and degradation in the definition. FN3 Respondents challenged the regulation on its face. Their complaint alleged that application of the “harm” regulation to the red-cockaded woodpecker, an endangered species, FN4 and the northern spotted owl, a threatened species, FN5 had injured them economically. App. 17-23.
FN3. Respondents also argued in the District Court that the Secretary's definition of “harm” is unconstitutionally void for vagueness, but they do not press that argument here.
FN4. The woodpecker was listed as an endangered species in 1970 pursuant to the statutory predecessor of the ESA. See 50 CFR § 17.11(h) (1994), issued pursuant to the Endangered Species Conservation Act of 1969, 83 Stat. 275.
FN5. See 55 Fed.Reg. 26114 (1990). Another regulation promulgated by the Secretary extends to threatened species, defined in the ESA as “any species which is likely to become an endangered species within the foreseeable future throughout all or a significant portion of its range,” 16 U.S.C. § 1532(20), some but not all of the protections endangered species enjoy. See 50 CFR § 17.31(a) (1994). In the District Court respondents unsuccessfully challenged that regulation's extension of § 9 to threatened species, but they do not press the challenge here.
FN6. Senate 1983, reprinted in Hearings on S. 1592 and S. 1983 before the Subcommittee on Environment of the Senate Committee on Commerce, 93d Cong., 1st Sess., 27 (1973).
The District Court considered and rejected each of respondents' arguments, finding “that Congress intended an expansive interpretation of the word ‘take,’ an interpretation that encompasses habitat modification.” 806 F.Supp. 279, 285 (1992). The court noted that in 1982, when Congress was aware of a judicial decision that had applied the Secretary's regulation, see Palila v. Hawaii Dept. of Land and Natural Resources, 639 F.2d 495 (CA9 1981) ( Palila I ), it amended the Act without using the opportunity to change the definition of “take.” 806 F.Supp., at 284. The court stated that, even had it found the ESA “ ‘silent or ambiguous' ” as to the authority for the Secretary's definition of “harm,” it would nevertheless have upheld the regulation as a reasonable interpretation of the statute. Id., at 285 (quoting *694 Chevron U.S.A. Inc. v. Natural Resources Defense Council, Inc., 467 U.S. 837, 843, 104 S.Ct. 2778, 2782, 81 L.Ed.2d 694 (1984)). The District Court therefore entered summary judgment for petitioners and dismissed respondents' complaint.
A divided panel of the Court of Appeals initially affirmed the judgment of the District Court. 1 F.3d 1 (CADC 1993). After granting a petition for rehearing, however, the panel reversed. 17 F.3d 1463 (CADC 1994). Although acknowledging that “[t]he potential breadth of the word ‘harm’ is indisputable,” id., at 1464, the majority concluded that the immediate statutory context in which “harm” appeared counseled against a broad reading; like the other words in the definition of “take,” the word “harm” should be read as applying only to “the perpetrator's direct application of force against the animal taken.... The forbidden acts fit, in ordinary language, the basic model ‘A hit B.’ ” Id., at 1465. The majority based its reasoning on a canon of statutory construction called noscitur a sociis, which holds that a word is known by the company it keeps. See Neal v. Clark, 95 U.S. 704, 708-709, 24 L.Ed. 586 (1878).
The majority claimed support for its construction from a decision of the Ninth Circuit that narrowly construed the word “harass” in the Marine Mammal Protection Act of 1972, 16 U.S.C. § 1372(a)(2)(A), see United States v. Hayashi, 5 F.3d 1278, 1282 (1993); from the legislative history of the ESA; FN7 from its view that Congress must not have intended the purportedly broad curtailment of private property rights that the Secretary's interpretation permitted; and from the ESA's land acquisition provision in § 5 and restriction on federal agencies' activities regarding habitat in § 7, both of which the court saw as evidence that Congress had not intended the § 9 “take” prohibition to reach habitat modification. *695 Most prominently, the court performed a lengthy analysis of the 1982 amendment to § 10 that provided for “incidental take permits” and concluded that the amendment did not change the meaning of the term “take” as defined in the 1973 statute. FN8
FN7. Judge Sentelle filed a partial concurrence in which he declined to join the portions of the court's opinion that relied on legislative history. See 17 F.3d 1463, 1472 (CADC 1994).
FN8. The 1982 amendment had formed the basis on which the author of the majority's opinion on rehearing originally voted to affirm the judgment of the District Court. Compare 1 F.3d 1, 11 (CADC 1993) (Williams, J., concurring in part), with 17 F.3d, at 1467-1472.
Chief Judge Mikva, who had announced the panel's original decision, dissented. See 17 F.3d, at 1473. In his view, a proper application of Chevron indicated that the Secretary had reasonably defined “harm,” because respondents had failed to show that **2412 Congress unambiguously manifested its intent to exclude habitat modification from the ambit of “take.” Chief Judge Mikva found the majority's reliance on noscitur a sociis inappropriate in light of the statutory language and unnecessary in light of the strong support in the legislative history for the Secretary's interpretation. He did not find the 1982 “incidental take permit” amendment alone sufficient to vindicate the Secretary's definition of “harm,” but he believed the amendment provided additional support for that definition because it reflected Congress' view in 1982 that the definition was reasonable.
The Court of Appeals' decision created a square conflict with a 1988 decision of the Ninth Circuit that had upheld the Secretary's definition of “harm.” See Palila v. Hawaii Dept. of Land and Natural Resources, 852 F.2d 1106 (1988) ( Palila II ). The Court of Appeals neither cited nor distinguished Palila II, despite the stark contrast between the Ninth Circuit's holding and its own. We granted certiorari to resolve the conflict. 513 U.S. 1072, 115 S.Ct. 714, 130 L.Ed.2d 621 (1995). Our consideration of the text and structure of the Act, its legislative history, and the significance of the 1982 amendment persuades us that the Court of Appeals' judgment should be reversed.
Because this case was decided on motions for summary judgment, we may appropriately make certain factual assumptions in order to frame the legal issue. First, we assume respondents have no desire to harm either the red-cockaded woodpecker or the spotted owl; they merely wish to continue logging activities that would be entirely proper if not prohibited by the ESA. On the other hand, we must assume, arguendo, that those activities will have the effect, even though unintended, of detrimentally changing the natural habitat of both listed species and that, as a consequence, members of those species will be killed or injured. Under respondents' view of the law, the Secretary's only means of forestalling that grave result-even when the actor knows it is certain to occur FN9 -is to use his § 5 authority to purchase *697 the lands on which the survival of the species depends. The Secretary, on the other hand, submits that the § 9 prohibition on takings, which Congress defined to include “harm,” places on respondents a duty to avoid harm that habitat alteration will cause the birds unless respondents first obtain a permit pursuant to § 10.
FN9. As discussed above, the Secretary's definition of “harm” is limited to “act[s] which actually kil[l] or injur[e] wildlife.” 50 CFR § 17.3 (1994). In addition, in order to be subject to the Act's criminal penalties or the more severe of its civil penalties, one must “knowingly violat[e]” the Act or its implementing regulations. 16 U.S.C. §§ 1540(a)(1), (b)(1). Congress added “knowingly” in place of “willfully” in 1978 to make “criminal violations of the act a general rather than a specific intent crime.” H.R.Conf.Rep. No. 95-1804, p. 26 (1978). The Act does authorize up to a $500 civil fine for “[a]ny person who otherwise violates” the Act or its implementing regulations. 16 U.S.C. § 1540(a)(1). That provision is potentially sweeping, but it would be so with or without the Secretary's “harm” regulation, making it unhelpful in assessing the reasonableness of the regulation. We have imputed scienter requirements to criminal statutes that impose sanctions without expressly requiring scienter, see, e.g., Staples v. United States, 511 U.S. 600, 114 S.Ct. 1793, 128 L.Ed.2d 608 (1994), but the proper case in which we might consider whether to do so in the § 9 provision for a $500 civil penalty would be a challenge to enforcement of that provision itself, not a challenge to a regulation that merely defines a statutory term. We do not agree with the dissent that the regulation covers results that are not “even foreseeable ... no matter how long the chain of causality between modification and injury.” Post, at 2421. Respondents have suggested no reason why either the “knowingly violates” or the “otherwise violates” provision of the statute-or the “harm” regulation itself-should not be read to incorporate ordinary requirements of proximate causation and foreseeability. In any event, neither respondents nor their amici have suggested that the Secretary employs the “otherwise violates” provision with any frequency.
The text of the Act provides three reasons for concluding that the Secretary's interpretation is reasonable. First, an ordinary understanding of the word “harm” supports it. The dictionary definition of the verb form of “harm” is “to cause hurt or damage to: injure.” Webster's Third New International Dictionary 1034 (1966). In the context of the ESA, that definition naturally encompasses habitat modification that results in actual **2413 injury or death to members of an endangered or threatened species.
Respondents argue that the Secretary should have limited the purview of “harm” to direct applications of force against protected species, but the dictionary definition does not include the word “directly” or suggest in any way that only direct or willful action that leads to injury constitutes “harm.” FN10 Moreover, unless the statutory term “harm” encompasses *698 indirect as well as direct injuries, the word has no meaning that does not duplicate the meaning of other words that § 3 uses to define “take.” A reluctance to treat statutory terms as surplusage supports the reasonableness of the Secretary's interpretation. See, e.g., Mackey v. Lanier Collection Agency & Service, Inc., 486 U.S. 825, 837, and n. 11, 108 S.Ct. 2182, 2189, and n. 11, 100 L.Ed.2d 836 (1988). FN11
FN10. Respondents and the dissent emphasize what they portray as the “established meaning” of “take” in the sense of a “wildlife take,” a meaning respondents argue extends only to “the effort to exercise dominion over some creature, and the concrete effect of [ sic ] that creature.” Brief for Respondents 19; see post, at 2422-2423. This limitation ill serves the statutory text, which forbids not taking “some creature” but “tak[ing] any [endangered] species ”-a formidable task for even the most rapacious feudal lord. More importantly, Congress explicitly defined the operative term “take” in the ESA, no matter how much the dissent wishes otherwise, see post, at 2422-2424, 2424-2425, thereby obviating the need for us to probe its meaning as we must probe the meaning of the undefined subsidiary term “harm.” Finally, Congress' definition of “take” includes several words-most obviously “harass,” “pursue,” and “wound,” in addition to “harm” itself-that fit respondents' and the dissent's definition of “take” no better than does “significant habitat modification or degradation.”
FN11. In contrast, if the statutory term “harm” encompasses such indirect means of killing and injuring wildlife as habitat modification, the other terms listed in § 3-“harass,” “pursue,” “hunt,” “shoot,” “wound,” “kill,” “trap,” “capture,” and “collect”-generally retain independent meanings. Most of those terms refer to deliberate actions more frequently than does “harm,” and they therefore do not duplicate the sense of indirect causation that “harm” adds to the statute. In addition, most of the other words in the definition describe either actions from which habitat modification does not usually result ( e.g., “pursue,” “harass”) or effects to which activities that modify habitat do not usually lead ( e.g., “trap,” “collect”). To the extent the Secretary's definition of “harm” may have applications that overlap with other words in the definition, that overlap reflects the broad purpose of the Act. See infra, at 2413-2414.
*699 In Hill, we construed § 7 as precluding the completion of the Tellico Dam because of its predicted impact on the survival of the snail darter. See 437 U.S., at 193, 98 S.Ct., at 2301. Both our holding and the language in our opinion stressed the importance of the statutory policy. “The plain intent of Congress in enacting this statute,” we recognized, “was to halt and reverse the trend toward species extinction, whatever the cost. This is reflected not only in the stated policies of the Act, but in literally every section of the statute.” Id., at 184, 98 S.Ct., at 2297. Although the § 9 “take” prohibition was not at issue in Hill, we took note of that prohibition, placing particular emphasis on the Secretary's inclusion of habitat modification in his definition of “harm.” FN12 In light of that **2414 provision for habitat protection, we could “not understand how TVA intends to operate Tellico Dam without ‘harming’ the snail darter.” Id., at 184, n. 30, 98 S.Ct., at 2297. Congress' intent to provide comprehensive protection for endangered and threatened species supports the permissibility of the Secretary's “harm” regulation.
FN12. We stated: “The Secretary of the Interior has defined the term ‘harm’ to mean ‘an act or omission which actually injures or kills wildlife, including acts which annoy it to such an extent as to significantly disrupt essential behavioral patterns, which include, but are not limited to, breeding, feeding or sheltering; significant environmental modification or degradation which has such effects is included within the meaning of “harm.” ’ ” TVA v. Hill, 437 U.S., at 184-185, n. 30, 98 S.Ct., at 2297, n. 30 (citations omitted; emphasis in original).
FN13. The dissent incorrectly asserts that the Secretary's regulation (1) “dispenses with the foreseeability of harm” and (2) “fail[s] to require injury to particular animals,” post, at 2429. As to the first assertion, the regulation merely implements the statute, and it is therefore subject to the statute's “knowingly violates” language, see 16 U.S.C. §§ 1540(a)(1), (b)(1), and ordinary requirements of proximate causation and foreseeability. See n. 9, supra. Nothing in the regulation purports to weaken those requirements. To the contrary, the word “actually” in the regulation should be construed to limit the liability about which the dissent appears most concerned, liability under the statute's “otherwise violates” provision. See n. 9, supra; post, at 2424-2425, 2429. The Secretary did not need to include “actually” to connote “but for” causation, which the other words in the definition obviously require. As to the dissent's second assertion, every term in the regulation's definition of “harm” is subservient to the phrase “an act which actually kills or injures wildlife.”
FN14. The dissent acknowledges the legislative history's clear indication that the drafters of the 1982 amendment had habitat modification in mind, see post, at 2428, but argues that the text of the amendment requires a contrary conclusion. This argument overlooks the statute's requirement of a “conservation plan,” which must describe an alternative to a known, but undesired, habitat modification.
FN15. The dissent makes no effort to defend the Court of Appeals' reading of the statutory definition as requiring a direct application of force. Instead, it tries to impose on § 9 a limitation of liability to “affirmative conduct intentionally directed against a particular animal or animals.” Post, at 2424. Under the dissent's interpretation of the Act, a developer could drain a pond, knowing that the act would extinguish an endangered species of turtles, without even proposing a conservation plan or applying for a permit under § 10(a)(1)(B); unless the developer was motivated by a desire “to get at a turtle,” post, at 2424, no statutory taking could occur. Because such conduct would not constitute a taking at common law, the dissent would shield it from § 9 liability, even though the words “kill” and “harm” in the statutory definition could apply to such deliberate conduct. We cannot accept that limitation. In any event, our reasons for rejecting the Court of Appeals' interpretation apply as well to the dissent's novel construction.
FN16. Respondents' reliance on United States v. Hayashi, 22 F.3d 859 (CA9 1993), is also misplaced. Hayashi construed the term “harass,” part of the definition of “take” in the Marine Mammal Protection Act of 1972, 16 U.S.C. § 1361 et seq., as requiring a “direct intrusion” on wildlife to support a criminal prosecution. 22 F.3d, at 864. Hayashi dealt with a challenge to a single application of a statute whose “take” definition includes neither “harm” nor several of the other words that appear in the ESA definition. Moreover, Hayashi was decided by a panel of the Ninth Circuit, the same court that had previously upheld the regulation at issue here in Palila II , 852 F.2d 1106 (1988). Neither the Hayashi majority nor the dissent saw any need to distinguish or even to cite Palila II.
FN17. Congress recognized that §§ 7 and 9 are not coextensive as to federal agencies when, in the wake of our decision in Hill in 1978, it added § 7( o ), 16 U.S.C. § 1536( o ), to the Act. That section provides that any federal project subject to exemption from § 7, 16 U.S.C. § 1536(h), will also be exempt from § 9.
FN18. Respondents also argue that the rule of lenity should foreclose any deference to the Secretary's interpretation of the ESA because the statute includes criminal penalties. The rule of lenity is premised on two ideas: First, “ ‘a fair warning should be given to the world in language that the common world will understand, of what the law intends to do if a certain line is passed’ ”; second, “legislatures and not courts should define criminal activity.” United States v. Bass, 404 U.S. 336, 347-350, 92 S.Ct. 515, 522-523, 30 L.Ed.2d 488 (1971) (quoting McBoyle v. United States, 283 U.S. 25, 27, 51 S.Ct. 340, 341, 75 L.Ed. 816 (1931)). We have applied the rule of lenity in a case raising a narrow question concerning the application of a statute that contains criminal sanctions to a specific factual dispute-whether pistols with short barrels and attachable shoulder stocks are short-barreled rifles-where no regulation was present. See United States v. Thompson/Center Arms Co., 504 U.S. 505, 517-518, and n. 9, 112 S.Ct. 2102, 2109-2110, and n. 9, 119 L.Ed.2d 308 (1992). We have never suggested that the rule of lenity should provide the standard for reviewing facial challenges to administrative regulations whenever the governing statute authorizes criminal enforcement. Even if there exist regulations whose interpretations of statutory criminal penalties provide such inadequate notice of potential liability as to offend the rule of lenity, the “harm” regulation, which has existed for two decades and gives a fair warning of its consequences, cannot be one of them.
Our conclusion that the Secretary's definition of “harm” rests on a permissible construction of the ESA gains further support from the legislative history of the statute. The Committee Reports accompanying the bills that became the ESA do not specifically discuss the meaning of “harm,” but they make clear that Congress intended “take” to apply broadly to cover indirect as well as purposeful actions. The Senate Report stressed that “ ‘[t]ake’ is defined ... in the broadest possible manner to include every conceivable way in which a person can ‘take’ or attempt to ‘take’ any fish or wildlife.” S.Rep. No. 93-307, p. 7 (1973). U.S.Code Cong. & Admin.News 1973, pp. 2989, 2995. The House Report stated that “the broadest possible terms” were used to define restrictions on takings. H.R.Rep. No. 93-412, p. 15 (1973). The House Report underscored the breadth of the *705 “take” definition by noting that it included “harassment, whether intentional or not. ” Id., at 11 (emphasis added). The Report explained that the definition “would allow, for example, the Secretary to regulate or prohibit the activities of birdwatchers where the effect of those activities might disturb the birds and make it difficult for them to hatch or raise their young.” Ibid. These comments, ignored in the dissent's welcome but selective foray into legislative history, see post, at 2426-2429, support the Secretary's interpretation that the term “take” in § 9 reached far more than the deliberate actions of hunters and trappers.
Two endangered species bills, S. 1592 and S. 1983, were introduced in the Senate and referred to the Commerce Committee. Neither bill included the word “harm” in its definition of “take,” although the definitions otherwise closely resembled the one that appeared in the bill as ultimately enacted. See Hearings on S. 1592 and S. 1983 before the Subcommittee on Environment of the Senate Committee on Commerce, 93d Cong., 1st Sess., pp. 7, 27 (1973) (hereinafter Hearings). Senator Tunney, the floor manager of the bill in the Senate, subsequently introduced a floor amendment that added “harm” to the **2417 definition, noting that this and accompanying amendments would “help to achieve the purposes of the bill.” 119 Cong.Rec. 25683 (1973). Respondents argue that the lack of debate about the amendment that added “harm” counsels in favor of a narrow interpretation. We disagree. An obviously broad word that the Senate went out of its way to add to an important statutory definition is precisely the sort of provision that deserves a respectful reading.
The definition of “take” that originally appeared in S. 1983 differed from the definition as ultimately enacted in one other significant respect: It included “the destruction, modification, or curtailment of [the] habitat or range” of fish and wildlife. Hearings, at 27. Respondents make much of the fact that the Commerce Committee removed this phrase *706 from the “take” definition before S. 1983 went to the floor. See 119 Cong.Rec. 25663 (1973). We do not find that fact especially significant. The legislative materials contain no indication why the habitat protection provision was deleted. That provision differed greatly from the regulation at issue today. Most notably, the habitat protection provision in S. 1983 would have applied far more broadly than the regulation does because it made adverse habitat modification a categorical violation of the “take” prohibition, unbounded by the regulation's limitation to habitat modifications that actually kill or injure wildlife. The S. 1983 language also failed to qualify “modification” with the regulation's limiting adjective “significant.” We do not believe the Senate's unelaborated disavowal of the provision in S. 1983 undermines the reasonableness of the more moderate habitat protection in the Secretary's “harm” regulation. FN19
FN19. Respondents place heavy reliance for their argument that Congress intended the § 5 land acquisition provision and not § 9 to be the ESA's remedy for habitat modification on a floor statement by Senator Tunney:
“Many species have been inadvertently exterminated by a negligent destruction of their habitat. Their habitats have been cut in size, polluted, or otherwise altered so that they are unsuitable environments for natural populations of fish and wildlife. Under this bill, we can take steps to make amends for our negligent encroachment. The Secretary would be empowered to use the land acquisition authority granted to him in certain existing legislation to acquire land for the use of the endangered species programs.... Through these land acquisition provisions, we will be able to conserve habitats necessary to protect fish and wildlife from further destruction.“Although most endangered species are threatened primarily by the destruction of their natural habitats, a significant portion of these animals are subject to predation by man for commercial, sport, consumption, or other purposes. The provisions in S. 1983 would prohibit the commerce in or the importation, exportation, or taking of endangered species....” 119 Cong.Rec. 25669 (1973).Similarly, respondents emphasize a floor statement by Representative Sullivan, the House floor manager for the ESA:“For the most part, the principal threat to animals stems from destruction of their habitat.... H.R. 37 will meet this problem by providing funds for acquisition of critical habitat.... It will also enable the Department of Agriculture to cooperate with willing landowners who desire to assist in the protection of endangered species, but who are understandably unwilling to do so at excessive cost to themselves.“Another hazard to endangered species arises from those who would capture or kill them for pleasure or profit. There is no way that Congress can make it less pleasurable for a person to take an animal, but we can certainly make it less profitable for them to do so.” Id., at 30162.Each of these statements merely explained features of the bills that Congress eventually enacted in § 5 of the ESA and went on to discuss elements enacted in § 9. Neither statement even suggested that § 5 would be the Act's exclusive remedy for habitat modification by private landowners or that habitat modification by private landowners stood outside the ambit of § 9. Respondents' suggestion that these statements identified § 5 as the ESA's only response to habitat modification contradicts their emphasis elsewhere on the habitat protections in § 7. See supra, at 2415.
When it enacted the ESA, Congress delegated broad administrative and interpretive power to the Secretary. See 16 U.S.C. §§ 1533, 1540(f). The task of defining and listing endangered and threatened species requires an expertise and attention to detail that exceeds the normal province of Congress. Fashioning appropriate standards for issuing permits under § 10 for takings that would otherwise violate § 9 necessarily requires the exercise of broad discretion. The proper interpretation of a term such as “harm” involves a complex policy choice. When Congress has entrusted the Secretary with broad discretion, we are especially reluctant to substitute our views of wise policy for his. See Chevron, 467 U.S., at 865-866, 104 S.Ct., at 2793. In this case, that reluctance accords with our conclusion, based on the text, structure, and legislative history of the ESA, that the Secretary reasonably construed the intent of Congress when he defined “harm” to include “significant habitat modification or degradation that actually kills or injures wildlife.”
In the elaboration and enforcement of the ESA, the Secretary and all persons who must comply with the law will confront difficult questions of proximity and degree; for, as all recognize, the Act encompasses a vast range of economic and social enterprises and endeavors. These questions must be addressed in the usual course of the law, through case-by-case resolution and adjudication.
The judgment of the Court of Appeals is reversed.
It is so ordered.
Justice O'CONNOR, concurring.
In my view, the regulation is limited by its terms to actions that actually kill or injure individual animals. Justice SCALIA disagrees, arguing that the harm regulation “encompasses injury inflicted, not only upon individual animals, but upon populations of the protected species.” Post, at 2422. At one level, I could not reasonably quarrel with this observation; death to an individual animal always reduces the size of the population in which it lives, and in that sense, “injures” that population. But by its insight, the dissent means something else. Building upon the regulation's use of the word “breeding,” Justice SCALIA suggests that the regulation facially bars significant habitat modification that actually kills or injures hypothetical animals (or, perhaps more aptly, causes potential additions to the population not to come **2419 into being). Because “[i]mpairment of breeding does not ‘injure’ living creatures,” Justice SCALIA reasons, the regulation must contemplate application to “ a population of animals which would otherwise have maintained or increased its numbers.” Post, at 2422, 2430.
I disagree. As an initial matter, I do not find it as easy as Justice SCALIA does to dismiss the notion that significant impairment of breeding injures living creatures. To raze the last remaining ground on which the piping plover currently *710 breeds, thereby making it impossible for any piping plovers to reproduce, would obviously injure the population (causing the species' extinction in a generation). But by completely preventing breeding, it would also injure the individual living bird, in the same way that sterilizing the creature injures the individual living bird. To “injure” is, among other things, “to impair.” Webster's Ninth New Collegiate Dictionary 623 (1983). One need not subscribe to theories of “psychic harm,” cf. post, at 2430, n. 5, to recognize that to make it impossible for an animal to reproduce is to impair its most essential physical functions and to render that animal, and its genetic material, biologically obsolete. This, in my view, is actual injury.
In any event, even if impairing an animal's ability to breed were not, in and of itself, an injury to that animal, interference with breeding can cause an animal to suffer other, perhaps more obvious, kinds of injury. The regulation has clear application, for example, to significant habitat modification that kills or physically injures animals which, because they are in a vulnerable breeding state, do not or cannot flee or defend themselves, or to environmental pollutants that cause an animal to suffer physical complications during gestation. Breeding, feeding, and sheltering are what animals do. If significant habitat modification, by interfering with these essential behaviors, actually kills or injures an animal protected by the Act, it causes “harm” within the meaning of the regulation. In contrast to Justice SCALIA, I do not read the regulation's “breeding” reference to vitiate or somehow to qualify the clear actual death or injury requirement, or to suggest that the regulation contemplates extension to nonexistent animals.
There is no inconsistency, I should add, between this interpretation and the commentary that accompanied the amendment of the regulation to include the actual death or injury requirement. See 46 Fed.Reg. 54748 (1981). Quite the contrary. It is true, as Justice SCALIA observes, post, at 2422, *711 that the Fish and Wildlife Service states at one point that “harm” is not limited to “direct physical injury to an individual member of the wildlife species,” see 46 Fed.Reg. 54748 (1981). But one could just as easily emphasize the word “direct” in this sentence as the word “individual.” FN* Elsewhere in the commentary, the Service makes clear that “section 9's threshold does focus on individual members of a protected species.” Id., at 54749. Moreover, the Service says that the regulation has no application to speculative harm, explaining that its insertion of the word “actually” was intended “to bulwark the need for proven injury to a species due to a party's actions.” Ibid.; see also ibid. (approving language that “[h]arm covers actions ... which actually (as opposed to potentially), cause injury”). That a protected animal could have eaten the leaves of a fallen tree or could, perhaps, have fruitfully multiplied in its branches is not sufficient under the regulation. Instead, as the commentary reflects, the regulation requires demonstrable effect ( i.e., actual injury or death) on actual, individual members of the protected species.
FN* Justice SCALIA suggests that, if the word “direct” merits emphasis in this sentence, then the sentence should be read as an effort to negate principles of proximate causation. See post, at 2430-2431, n. 5. As this case itself demonstrates, however, the word “direct” is susceptible of many meanings. The Court of Appeals, for example, used “direct” to suggest an element of purposefulness. See 17 F.3d 1463, 1465 (CADC 1994). So, occasionally, does the dissent. See post, at 2423 (describing “affirmative acts ... which are directed immediately and intentionally against a particular animal”) (emphasis added). It is not hard to imagine conduct that, while “indirect” ( i.e., nonpurposeful), proximately causes actual death or injury to individual protected animals, cf. post, at 2429; indeed, principles of proximate cause routinely apply in the negligence and strict liability contexts.
*713 Proximate causation is not a concept susceptible of precise definition. See Keeton, supra, at 280-281. It is easy enough, of course, to identify the extremes. The farmer whose fertilizer is lifted by a tornado from tilled fields and deposited miles away in a wildlife refuge cannot, by any stretch of the term, be considered the proximate cause of death or injury to protected species occasioned thereby. At the same time, the landowner who drains a pond on his property, killing endangered fish in the process, would likely satisfy any formulation of the principle. We have recently said that proximate causation “normally eliminates the bizarre,” Jerome B. Grubart, Inc. v. Great Lakes Dredge & Dock Co., 513 U.S. 527, 536, 115 S.Ct. 1043, 1049, 130 L.Ed.2d 1024 (1995), and have noted its “functionally equivalent” alternative characterizations in terms of foreseeability, see Milwaukee & St. Paul R. Co. v. Kellogg, 94 U.S. 469, 475, 24 L.Ed. 256 (1877) (“natural and probable consequence”), and duty, see Palsgraf v. Long Island R. Co., 248 N.Y. 339, 162 N.E. 99 (1928). Consolidated Rail Corporation v. Gottshall, 512 U.S. 532, 546, 114 S.Ct. 2396, 2405-2406, 129 L.Ed.2d 427 (1994). Proximate causation depends to a great extent on considerations of the fairness of imposing liability for remote consequences. The task of determining whether proximate causation exists in the limitless fact patterns sure to arise is best left to lower courts. But I note, at the least, that proximate cause principles inject a foreseeability element into the statute, and hence, the regulation, that would appear to alleviate some of the problems noted by the dissent. See, e.g., post, at 2423 (describing “a farmer who tills his field and causes erosion that makes silt run into a nearby river which depletes oxygen and thereby [injures] protected fish”).
In my view, then, the “harm” regulation applies where significant habitat modification, by impairing essential behaviors, proximately (foreseeably) causes actual death or injury to identifiable animals that are protected under the Endangered Species Act. Pursuant to my interpretation, Palila II -under which the Court of Appeals held that a state *714 agency committed a “taking” by permitting **2421 mouflon sheep to eat mamane-naio seedlings that, when full grown, might have fed and sheltered endangered palila-was wrongly decided according to the regulation's own terms. Destruction of the seedlings did not proximately cause actual death or injury to identifiable birds; it merely prevented the regeneration of forest land not currently sustaining actual birds.
This case, of course, comes to us as a facial challenge. We are charged with deciding whether the regulation on its face exceeds the agency's statutory mandate. I have identified at least one application of the regulation ( Palila II ) that is, in my view, inconsistent with the regulation's own limitations. That misapplication does not, however, call into question the validity of the regulation itself. One can doubtless imagine questionable applications of the regulation that test the limits of the agency's authority. However, it seems to me clear that the regulation does not on its terms exceed the agency's mandate, and that the regulation has innumerable valid habitat-related applications. Congress may, of course, see fit to revisit this issue. And nothing the Court says today prevents the agency itself from narrowing the scope of its regulation at a later date.
With this understanding, I join the Court's opinion.
Justice SCALIA, with whom THE CHIEF JUSTICE and Justice THOMAS join, dissenting.
The Endangered Species Act of 1973 (Act), 16 U.S.C. § 1531 et seq. (1988 ed. and Supp. V), provides that “it is unlawful for any person subject to the jurisdiction of the United States to take-... any [protected] species within the United States.” § 1538(a)(1)(B). The term “take” is defined as “to harass, harm, pursue, hunt, shoot, wound, kill, trap, capture, or collect, or to attempt to engage in any such conduct.” § 1532(19) (emphasis added). The challenged regulation defines “harm” thus:
“ Harm in the definition of ‘take’ in the Act means an act which actually kills or injures wildlife. Such act may include significant habitat modification or degradation where it actually kills or injures wildlife by significantly impairing essential behavioral patterns, including breeding, feeding or sheltering.” 50 CFR § 17.3 (1994).
In my view petitioners must lose-the regulation must fall-even under the test of Chevron U.S.A. Inc. v. Natural Resources Defense Council, Inc., 467 U.S. 837, 843, 104 S.Ct. 2778, 2782, 81 L.Ed.2d 694 (1984), so I shall assume that the Court is correct to apply Chevron. See ante, at 2415-2416, and n. 18.
The regulation has three features which, for reasons I shall discuss at length below, do not comport with the statute. First, it interprets the statute to prohibit habitat modification that is no more than the cause-in-fact of death or injury to wildlife. Any “significant habitat modification” that in fact produces that result by “impairing essential behavioral patterns” is made unlawful, regardless of whether that result is intended or even foreseeable, and no matter how long the chain of causality between modification and injury. See, e.g., Palila v. Hawaii Dept. of Land and Natural Resources, 852 F.2d 1106, 1108-1109 (CA9 1988) ( Palila II ) (sheep grazing constituted “taking” of palila birds, since although sheep do not destroy full-grown mamane trees, they do destroy mamane seedlings, which will not grow *716 to full-grown trees, on which the palila feeds and nests). See also Davison, Alteration of Wildlife Habitat as a Prohibited Taking under the Endangered Species Act, 10 J.Land Use & Envtl.L. 155, 190 (1995) (regulation requires only causation-in-fact).
**2422 Second, the regulation does not require an “act”: The Secretary's officially stated position is that an omission will do. The previous version of the regulation made this explicit. See 40 Fed.Reg. 44412, 44416 (1975) ( “ ‘Harm’ in the definition of ‘take’ in the Act means an act or omission which actually kills or injures wildlife ...”). When the regulation was modified in 1981 the phrase “or omission” was taken out, but only because (as the final publication of the rule advised) “the [Fish and Wildlife] Service feels that ‘act’ is inclusive of either commissions or omissions which would be prohibited by section [1538(a)(1)(B) ].” 46 Fed.Reg. 54748, 54750 (1981). In their brief here petitioners agree that the regulation covers omissions, see Brief for Petitioners 47 (although they argue that “[a]n ‘omission’ constitutes an ‘act’ ... only if there is a legal duty to act”), ibid.
The third and most important unlawful feature of the regulation is that it encompasses injury inflicted, not only upon individual animals, but upon populations of the protected species. “Injury” in the regulation includes “significantly impairing essential behavioral patterns, including breeding, ” 50 CFR § 17.3 (1994) (emphasis added). Impairment of breeding does not “injure” living creatures; it prevents them from propagating, thus “injuring” a population of animals which would otherwise have maintained or increased its numbers. What the face of the regulation shows, the Secretary's official pronouncements confirm. The Final Redefinition of “Harm” accompanying publication of the regulation said that “harm” is not limited to “direct physical injury to an individual member of the wildlife species,” 46 Fed.Reg. 54748 (1981), and refers to “injury to a population, ” id., at 54749 (emphasis added). See also Palila II, supra, at 1108; *717 Davison, supra, at 190, and n. 177, 195; M. Bean, The Evolution of National Wildlife Law 344 (1983). FN1
FN1. The Court and Justice O'CONNOR deny that the regulation has the first or the third of these features. I respond to their arguments in Part III, infra.
The Act's definition of “take” does expand the word slightly (and not unusually), so as to make clear that it includes not just a completed taking, but the process of taking, and all of the acts that are customarily identified with or accompany that process (“to harass, harm, pursue, hunt, shoot, wound, kill, trap, capture, or collect”); and so as to include attempts. § 1532(19). The tempting fallacy-which the Court commits with abandon, see ante, at 2413, n. 10-is to assume that once defined, “take” loses any significance, and it is only the definition that matters. The Court treats the statute as though Congress had directly enacted the § 1532(19) definition as a self-executing prohibition, and had not enacted § 1538(a)(1)(B) at all. But § 1538(a)(1)(B) is there, and if the terms contained in the definitional section are susceptible of two readings, one of which comports with the standard meaning of “take” as used in application to wildlife, and one of which does not, an agency regulation that adopts the latter reading is necessarily unreasonable, for it reads the defined term “take”-the only operative term-out of the statute altogether. FN2
FN2. The Court suggests halfheartedly that “take” cannot refer to the taking of particular animals, because § 1538(a)(1)(B) prohibits “tak [ing] any [endangered] species. ” Ante, at 2413, n. 10. The suggestion is halfhearted because that reading obviously contradicts the statutory intent. It would mean no violation in the intentional shooting of a single bald eagle-or, for that matter, the intentional shooting of 1,000 bald eagles out of the extant 1,001. The phrasing of § 1538(a)(1)(B), as the Court recognizes elsewhere, see[, e.g., ante, at 2412, is shorthand for “take any member of an endangered] species.”
Here the evidence shows the opposite. “Harm” is merely one of 10 prohibitory words in § 1532(19), and the other 9 fit the ordinary meaning of “take” perfectly. To “harass, pursue, hunt, shoot, wound, kill, trap, capture, or collect” are *720 all affirmative acts (the provision itself describes them as “conduct,” see § 1532(19)) which are directed immediately and intentionally against a particular animal-not acts or omissions that indirectly and accidentally cause injury to a population of animals. The Court points out that several of the words (“harass,” “pursue,” “wound,” and “kill”) “refer to actions or effects that do not require direct applications of force. ” Ante, at 2414 (emphasis added). That is true enough, but force is not the point. Even “taking” activities in the narrowest sense, activities traditionally engaged in by hunters and trappers, do not all consist of direct applications of force; pursuit and harassment are part of the business of “taking” the prey even before it has been **2424 touched. What the nine other words in § 1532(19) have in common-and share with the narrower meaning of “harm” described above, but not with the Secretary's ruthless dilation of the word-is the sense of affirmative conduct intentionally directed against a particular animal or animals.
I am not the first to notice this fact, or to draw the conclusion that it compels. In 1981 the Solicitor of the Fish and Wildlife Service delivered a legal opinion on § 1532(19) that is in complete agreement with my reading:
I would call it noscitur a sociis, but the principle is much the same: The fact that “several items in a list share an attribute *721 counsels in favor of interpreting the other items as possessing that attribute as well,” Beecham v. United States, 511 U.S. 368, 371, 114 S.Ct. 1669, 1671, 128 L.Ed.2d 383 (1994). The Court contends that the canon cannot be applied to deprive a word of all its “independent meaning,” ante, at 2415. That proposition is questionable to begin with, especially as applied to long lawyers' listings such as this. If it were true, we ought to give the word “trap” in the definition its rare meaning of “to clothe” (whence “trappings”)-since otherwise it adds nothing to the word “capture.” See Moskal v. United States, 498 U.S. 103, 120, 111 S.Ct. 461, 471, 112 L.Ed.2d 449 (1990) (SCALIA, J., dissenting). In any event, the Court's contention that “harm” in the narrow sense adds nothing to the other words underestimates the ingenuity of our own species in a way that Congress did not. To feed an animal poison, to spray it with mace, to chop down the very tree in which it is nesting, or even to destroy its entire habitat in order to take it (as by draining a pond to get at a turtle), might neither wound nor kill, but would directly and intentionally harm.
The penalty provisions of the Act counsel this interpretation as well. Any person who “knowingly” violates § 1538(a)(1)(B) is subject to criminal penalties under § 1540(b)(1) and civil penalties under § 1540(a)(1); moreover, under the latter section, any person “who otherwise violates” the taking prohibition ( i.e., violates it un knowingly) may be assessed a civil penalty of $500 for each violation, with the stricture that “[e]ach such violation shall be a separate offense.” This last provision should be clear warning that the regulation is in error, for when combined with the regulation it produces a result that no legislature could reasonably be thought to have intended: A large number of routine private activities-for example, farming, ranching, roadbuilding, construction and logging-are subjected to strict-liability penalties when they fortuitously injure protected wildlife, no matter how remote the chain of causation and no matter how difficult to foresee (or to disprove) the “injury” may be ( e.g., *722 an “impairment” of breeding). The Court says that “[the strict-liability provision] is potentially sweeping, but it would be so with or without the Secretary's ‘harm’ regulation.” Ante, at 2412, n. 9. That is not correct. Without the regulation, the routine “habitat modifying” activities that people conduct to make a daily living would not carry exposure to strict penalties; only acts directed at animals, like those described by the other words in § 1532(19), would risk liability.
The Court says that “[to] read a requirement of intent or purpose into the words used to define ‘take’ ... ignore[s] [§ 1540's] express provision that a ‘knowin[g]’ action is enough to violate the Act.” Ante, at 2415. This presumably means that because the reading of § 1532(19) advanced here ascribes an element of purposeful injury to the prohibited acts, it makes superfluous (or inexplicable) the more severe penalties provided for a “knowing” violation. That conclusion does not follow, for it is quite possible to take protected wildlife purposefully without doing so knowingly. A requirement that a violation **2425 be “knowing” means that the defendant must “know the facts that make his conduct illegal,” Staples v. United States, 511 U.S. 600, 606, 114 S.Ct. 1793, 1797, 128 L.Ed.2d 608 (1994). The hunter who shoots an elk in the mistaken belief that it is a mule deer has not knowingly violated § 1538(a)(1)(B)-not because he does not know that elk are legally protected (that would be knowledge of the law, which is not a requirement, see ante, at 2412, n. 9), but because he does not know what sort of animal he is shooting. The hunter has nonetheless committed a purposeful taking of protected wildlife, and would therefore be subject to the (lower) strict-liability penalties for the violation.
So far I have discussed only the immediate statutory text bearing on the regulation. But the definition of “take” in § 1532(19) applies “[f]or the purposes of this chapter,” that is, it governs the meaning of the word as used everywhere in the Act. Thus, the Secretary's interpretation of “harm” is wrong if it does not fit with the use of “take” throughout *723 the Act. And it does not. In § 1540(e)(4)(B), for example, Congress provided for the forfeiture of “[a]ll guns, traps, nets, and other equipment ... used to aid the taking, possessing, selling, [etc.]” of protected animals. This listing plainly relates to “taking” in the ordinary sense. If environmental modification were part (and necessarily a major part) of taking, as the Secretary maintains, one would have expected the list to include “plows, bulldozers, and backhoes.” As another example, § 1539(e)(1) exempts “the taking of any endangered species” by Alaskan Indians and Eskimos “if such taking is primarily for subsistence purposes”; and provides that “[n]on-edible byproducts of species taken pursuant to this section may be sold ... when made into authentic native articles of handicrafts and clothing.” Surely these provisions apply to taking only in the ordinary sense, and are meaningless as applied to species injured by environmental modification. The Act is full of like examples. See, e.g., § 1538(a)(1)(D) (prohibiting possession, sale, and transport of “species taken in violation” of the Act). “[I]f the Act is to be interpreted as a symmetrical and coherent regulatory scheme, one in which the operative words have a consistent meaning throughout,” Gustafson v. Alloyd Co., 513 U.S. 561, 569, 115 S.Ct. 1061, 1066-1067, 131 L.Ed.2d 1 (1995), the regulation must fall.
The broader structure of the Act confirms the unreasonableness of the regulation. Section 1536 provides:
The Act defines “critical habitat” as habitat that is “essential to the conservation of the species,” §§ 1532(5)(A)(i), (A)(ii), with “conservation” in turn defined as the use of methods *724 necessary to bring listed species “to the point at which the measures provided pursuant to this chapter are no longer necessary,” § 1532(3).
These provisions have a double significance. Even if §§ 1536(a)(2) and 1538(a)(1)(B) were totally independent prohibitions-the former applying only to federal agencies and their licensees, the latter only to private parties-Congress's explicit prohibition of habitat modification in the one section would bar the inference of an implicit prohibition of habitat modification in the other section. “[W]here Congress includes particular language in one section of a statute but omits it in another ..., it is generally presumed that Congress acts intentionally and purposely in the disparate inclusion or exclusion.” Keene Corp. v. United States, 508 U.S. 200, 208, 113 S.Ct. 2035, 2040, 124 L.Ed.2d 118 (1993) (internal quotation marks omitted). And that presumption against implicit prohibition would be even stronger where the one section which uses the language carefully defines and limits its application. That is to say, it would be passing strange for Congress carefully to define “critical habitat” as used in § 1536(a)(2), but leave it to the Secretary to evaluate, willy-nilly, impermissible “habitat modification” **2426 (under the guise of “harm”) in § 1538(a)(1)(B).
In fact, however, §§ 1536(a)(2) and 1538(a)(1)(B) do not operate in separate realms; federal agencies are subject to both, because the “person[s]” forbidden to take protected species under § 1538 include agencies and departments of the Federal Government. See § 1532(13). This means that the “harm” regulation also contradicts another principle of interpretation: that statutes should be read so far as possible to give independent effect to all their provisions. See Ratzlaf v. United States, 510 U.S. 135, 140-141, 114 S.Ct. 655, 658-659, 126 L.Ed.2d 615 (1994). By defining “harm” in the definition of “take” in § 1538(a)(1)(B) to include significant habitat modification that injures populations of wildlife, the regulation makes the habitat-modification restriction in § 1536(a)(2) almost wholly superfluous. As “critical habitat” is habitat “essential to the conservation of the *725 species,” adverse modification of “critical” habitat by a federal agency would also constitute habitat modification that injures a population of wildlife.
Petitioners try to salvage some independent scope for § 1536(a)(2) by the following contortion: Because the definition of critical habitat includes not only “the specific areas within the geographical area occupied by the species [that are] essential to the conservation of the species,” § 1532(5)(A)(i), but also “specific areas outside the geographical area occupied by the species at the time it is listed [as a protected species] ... [that are] essential to the conservation of the species,” § 1532(5)(A)(ii), there may be some agency modifications of critical habitat which do not injure a population of wildlife. See Brief for Petitioners 41, and n. 27. This is dubious to begin with. A principal way to injure wildlife under the Secretary's own regulation is to “significantly impai[r] ... breeding,” 50 CFR § 17.3 (1994). To prevent the natural increase of a species by adverse modification of habitat suitable for expansion assuredly impairs breeding. But even if true, the argument only narrows the scope of the superfluity, leaving as so many wasted words the § 1532(5)(A)(i) definition of critical habitat to include currently occupied habitat essential to the species' conservation. If the Secretary's definition of “harm” under § 1538(a)(1)(B) is to be upheld, we must believe that Congress enacted § 1536(a)(2) solely because in its absence federal agencies would be able to modify habitat in currently unoccupied areas. It is more rational to believe that the Secretary's expansion of § 1538(a)(1)(B) carves out the heart of one of the central provisions of the Act.
II
The Court makes four other arguments. First, “the broad purpose of the [Act] supports the Secretary's decision to extend protection against activities that cause the precise harms Congress enacted the statute to avoid.” Ante, at 2413. *726 I thought we had renounced the vice of “simplistically ... assum [ing] that whatever furthers the statute's primary objective must be the law.” Rodriguez v. United States, 480 U.S. 522, 526, 107 S.Ct. 1391, 1393, 94 L.Ed.2d 533 (1987) ( per curiam) (emphasis in original). Deduction from the “broad purpose” of a statute begs the question if it is used to decide by what means (and hence to what length ) Congress pursued that purpose; to get the right answer to that question there is no substitute for the hard job (or, in this case, the quite simple one) of reading the whole text. “The Act must do everything necessary to achieve its broad purpose” is the slogan of the enthusiast, not the analytical tool of the arbiter. FN3
FN3. This portion of the Court's opinion, see ante, at 2413, n. 12, discusses and quotes a footnote in TVA v. Hill, 437 U.S. 153, 184-185, n. 30, 98 S.Ct. 2279, 2297, n. 30, 57 L.Ed.2d 117 (1978), in which we described the then-current version of the Secretary's regulation, and said that the habitat modification undertaken by the federal agency in the case would have violated the regulation. Even if we had said that the Secretary's regulation was authorized by § 1538, that would have been utter dictum, for the only provision at issue was § 1536. See id., at 193, 98 S.Ct., at 2301. But in fact we simply opined on the effect of the regulation while assuming its validity, just as courts always do with provisions of law whose validity is not at issue.
Much of the Court's discussion of legislative history is devoted to two items: first, the Senate floor manager's introduction of an amendment that added the word “harm” to the definition of “take,” with the observation that (along with other amendments) it would “ ‘help to achieve the purposes of the bill’ ”; second, the relevant Committee's removal from the definition of a provision stating that “take” includes “ ‘the destruction, modification or curtailment of [the] habitat or range’ ” of fish and wildlife. See ante, at 2416-2417. The Court inflates the first and belittles the second, even though the second is on its face far more pertinent. But this elaborate inference from various pre-enactment actions and inactions is quite unnecessary, since we have direct evidence of what those who brought the legislation to the floor thought it meant-evidence as solid as any ever to be found in legislative history, but which the Court banishes to a footnote. See ante, at 2417, n. 19.
Both the Senate and House floor managers of the bill explained it in terms which leave no doubt that the problem of habitat destruction on private lands was to be solved principally by the land acquisition program of § 1534, while § 1538 solved a different problem altogether-the problem of takings. Senator Tunney stated:
“Although most endangered species are threatened primarily by the destruction of their natural habitats, a significant portion of these animals are subject to predation by man for commercial, sport, consumption, or other purposes. The provisions of [the bill] would prohibit the commerce in or the importation, exportation, or taking of endangered species....” 119 Cong.Rec. 25669 (1973) (emphasis added).
*728 The House floor manager, Representative Sullivan, put the same thought in this way:
“[T]he principal threat to animals stems from destruction of their habitat.... [ The bill] will meet this problem by providing funds for acquisition of critical habitat .... It will also enable the Department of Agriculture to cooperate with willing landowners who desire to assist in the protection of endangered species, but who are understandably unwilling to do so at excessive cost to themselves.
“Another hazard to endangered species arises from those who would capture or kill them for pleasure or profit. There is no way that the Congress can make it less pleasurable for a person to take an animal, but we can certainly make it less profitable for them to do so.” Id., at 30162 (emphasis added).
Habitat modification and takings, in other words, were viewed as different problems, addressed by different provisions of the Act. The Court really has no explanation for these statements. All it can say is that “[n]either statement even suggested that [the habitat acquisition funding provision in § 1534] would be the Act's exclusive remedy for habitat modification by private landowners or that habitat modification by private landowners stood outside the ambit of [§ 1538].” Ante, at 2417, n. 19. That is to say, the statements are not as bad as they might have been. Little in life is. They are, however, quite bad enough to destroy the Court's legislative-history case, since they display the **2428 clear understanding (1) that habitat modification is separate from “taking,” and (2) that habitat destruction on private lands is to be remedied by public acquisition, and not by making particular unlucky landowners incur “excessive cost to themselves.” The Court points out triumphantly that they do not display the understanding (3) that the land acquisition program is “the [Act's] only response to habitat modification.” *729 Ibid. Of course not, since that is not so (all public lands are subject to habitat-modification restrictions); but (1) and (2) are quite enough to exclude the Court's interpretation. They identify the land acquisition program as the Act's only response to habitat modification by private landowners, and thus do not in the least “contradic[t],” ibid., the fact that § 1536 prohibits habitat modification by federal agencies.
“[T]he Secretary may permit, under such terms and conditions as he shall prescribe-
....
“any taking otherwise prohibited by section 1538(a)(1)(B) ... if such taking is incidental to, and not the purpose of, the carrying out of an otherwise lawful activity.” 16 U.S.C. § 1539(a)(1)(B).
This provision does not, of course, implicate our doctrine that reenactment of a statutory provision ratifies an extant judicial or administrative interpretation, for neither the taking prohibition in § 1538(a)(1)(B) nor the definition in § 1532(19) was reenacted. See Central Bank of Denver, N.A. v. First Interstate Bank of Denver, N.A., 511 U.S. 164, 185, 114 S.Ct. 1439, 1452, 128 L.Ed.2d 119 (1994). The Court claims, however, that the provision “strongly suggests that Congress understood [§ 1538(a)(1)(B) ] to prohibit indirect as well as deliberate takings.” Ante, at 2414. That would be a valid inference if habitat modification were the only substantial “otherwise lawful activity” that might incidentally and nonpurposefully cause a prohibited “taking.” Of course it is not. This provision applies to the many otherwise lawful takings that incidentally take a protected species-as when fishing for unprotected salmon also takes an endangered species of salmon, see Pacific Northwest Generating Cooperative v. Brown, 38 F.3d 1058, 1067 (CA9 1994). *730 Congress has referred to such “incidental takings” in other statutes as well-for example, a statute referring to “the incidental taking of ... sea turtles in the course of ... harvesting [shrimp]” and to the “rate of incidental taking of sea turtles by United States vessels in the course of such harvesting,” 103 Stat. 1038, § 609(b)(2), note following 16 U.S.C. § 1537 (1988 ed., Supp. V); and a statute referring to “the incidental taking of marine mammals in the course of commercial fishing operations,” 108 Stat. 546, § 118(a). The Court shows that it misunderstands the question when it says that “[n]o one could seriously request an ‘incidental’ take permit to avert ... liability for direct, deliberate action against a member of an endangered or threatened species. ” Ante, at 2414 (emphasis added). That is not an incidental take at all. FN4
FN4. The statutory requirement of a “conservation plan” is as consistent with this construction as with the Court's. See ante, at 2414, and n. 14. The commercial fisherman who is in danger of incidentally sweeping up protected fish in his nets can quite reasonably be required to “minimize and mitigate” the “impact” of his activity. 16 U.S.C. § 1539(a)(2)(A).
Fourth and lastly, the Court seeks to avoid the evident shortcomings of the regulation on the ground that the respondents are challenging it on its face rather than as applied. See ante, at 2414; see also ante, at 2418 (O'CONNOR, J., concurring). The Court seems to say that even if the regulation dispenses with the foreseeability of harm that it acknowledges the statute to require, that does not matter because this is a facial challenge: So long as habitat modification that would foreseeably cause harm is prohibited by the statute, the regulation must be sustained. Presumably it would apply the same reasoning to all the other defects of the regulation: The regulation's failure to require injury to particular animals survives the present challenge, because at least some environmental modifications kill particular animals. This evisceration of the facial challenge is unprecedented. It is one thing to say that a facial challenge to a regulation that omits statutory element x must be rejected if there is any set of facts on which the statute does not require x. It is something quite different-and unlike any doctrine of “facial challenge” I have ever encountered-to say that the challenge must be rejected if the regulation could be applied to a state of facts in which element x happens to be present. On this analysis, the only regulation susceptible to facial attack is one that not only is invalid in all its applications, but also does not sweep up any person who could have been held liable under a proper application of the statute. That is not the law. Suppose a statute that prohibits “premeditated killing of a human being,” and an implementing regulation that prohibits “killing a human *732 being.” A facial challenge to the regulation would not be rejected on the ground that, after all, it could be applied to a killing that happened to be premeditated. It could not be applied to such a killing, because it does not require the factfinder to find premeditation, as the statute requires. In other words, to simplify its task the Court today confuses lawful application of the challenged regulation with lawful application of a different regulation, i.e., one requiring the various elements of liability that this regulation omits.
III
In response to the points made in this dissent, the Court's opinion stresses two points, neither of which is supported by the regulation, and so cannot validly be used to uphold it. First, the Court and the concurrence suggest that the regulation should be read to contain a requirement of proximate causation or foreseeability, principally because the statute does -and “[n]othing in the regulation purports to weaken those requirements [of the statute].” See ante, at 2412, n. 9; 2414, n. 13; see also ante, at 2419-2420 (O'CONNOR, J., concurring). I quite agree that the statute contains such a limitation, because the verbs of purpose in § 1538(a)(1)(B) denote action directed at animals. But the Court has rejected that reading. The critical premise on which it has upheld the regulation is that, despite the weight of the other words in § 1538(a)(1)(B), “the statutory term ‘harm’ encompasses indirect as well as direct injuries,” ante, at 2413. See also ante, at 2413, n. 11 (describing “the sense of indirect causation that ‘harm’ adds to the statute”); ante, at 2415 (stating that the Secretary permissibly interprets “ ‘harm’ ” to include “indirectly injuring endangered animals”). Consequently, unless there is some strange category of causation that is indirect and yet also proximate, the Court has already rejected its own basis for finding a proximate-cause limitation in the regulation. In fact “proximate” causation simply means “direct” causation. See, e.g., Black's Law Dictionary 1103 *733 5th ed. 1979) (defining “[p]roximate” as “Immediate; nearest; direct ”) (emphasis added); Webster's New International Dictionary 1995 (2d ed. **2430 1949) (“[P]roximate cause. A cause which directly, or with no mediate agency, produces an effect”) (emphasis added).
The only other reason given for finding a proximate-cause limitation in the regulation is that “by use of the word ‘actually,’ the regulation clearly rejects speculative or conjectural effects, and thus itself invokes principles of proximate causation.” Ante, at 2420 (O'CONNOR, J., concurring); see also ante, at 2414, n. 13 (majority opinion). Non sequitur, of course. That the injury must be “actual” as opposed to “potential” simply says nothing at all about the length or foreseeability of the causal chain between the habitat modification and the “actual” injury. It is thus true and irrelevant that “[t]he Secretary did not need to include ‘actually’ to connote ‘but for’ causation,” ibid.; “actually” defines the requisite injury, not the requisite causality.
The regulation says (it is worth repeating) that “harm” means (1) an act that (2) actually kills or injures wildlife. If that does not dispense with a proximate-cause requirement, I do not know what language would. And changing the regulation by judicial invention, even to achieve compliance with the statute, is not permissible. Perhaps the agency itself would prefer to achieve compliance in some other fashion. We defer to reasonable agency interpretations of ambiguous statutes precisely in order that agencies, rather than courts, may exercise policymaking discretion in the interstices of statutes. See Chevron, 467 U.S., at 843-845, 104 S.Ct., at 2782. Just as courts may not exercise an agency's power to adjudicate, and so may not affirm an agency order on discretionary grounds the agency has not advanced, see SEC v. Chenery Corp., 318 U.S. 80, 63 S.Ct. 454, 87 L.Ed. 626 (1943), so also this Court may not exercise the Secretary's power to regulate, and so may not uphold a regulation by adding to it even the most reasonable of elements it does not contain.
*734 The second point the Court stresses in its response seems to me a belated mending of its holding. It apparently concedes that the statute requires injury to particular animals rather than merely to populations of animals. See ante, at 2414, n. 13; ante, at 2412 (referring to killing or injuring “ members of [listed] species” (emphasis added)). The Court then rejects my contention that the regulation ignores this requirement, since, it says, “every term in the regulation's definition of ‘harm’ is subservient to the phrase ‘an act which actually kills or injures wildlife.’ ” Ante, at 2414, n. 13. As I have pointed out, see supra, at 2422, this reading is incompatible with the regulation's specification of impairment of “breeding” as one of the modes of “kill[ing] or injur[ing] wildlife.” FN5
FN5. Justice O'CONNOR supposes that an “impairment of breeding” intrinsically injures an animal because “to make it impossible for an animal to reproduce is to impair its most essential physical functions and to render that animal, and its genetic material, biologically obsolete.” Ante, at 2419 (concurring opinion). This imaginative construction does achieve the result of extending “impairment of breeding” to individual animals; but only at the expense of also expanding “injury” to include elements beyond physical harm to individual animals. For surely the only harm to the individual animal from impairment of that “essential function” is not the failure of issue (which harms only the issue), but the psychic harm of perceiving that it will leave this world with no issue (assuming, of course, that the animal in question, perhaps an endangered species of slug, is capable of such painful sentiments). If it includes that psychic harm, then why not the psychic harm of not being able to frolic about-so that the draining of a pond used for an endangered animal's recreation, but in no way essential to its survival, would be prohibited by the Act? That the concurrence is driven to such a dubious redoubt is an argument for, not against, the proposition that “injury” in the regulation includes injury to populations of animals. Even more so with the concurrence's alternative explanation: that “impairment of breeding” refers to nothing more than concrete injuries inflicted by the habitat modification on the animal who does the breeding, such as “physical complications [suffered] during gestation,” ibid. Quite obviously, if “impairment of breeding” meant such physical harm to an individual animal, it would not have had to be mentioned.
The concurrence entangles itself in a dilemma while attempting to explain the Secretary's commentary to the harm regulation, which stated that “harm” is not limited to “direct physical injury to an individual member of the wildlife species,” 46 Fed.Reg. 54748 (1981). The concurrence denies that this means that the regulation does not require injury to particular animals, because “one could just as easily emphasize the word ‘direct’ in this sentence as the word ‘individual.’ ” Ante, at 2419. One could; but if the concurrence does, it thereby refutes its separate attempt to exclude indirect causation from the regulation's coverage, see ante, at 2419-2420. The regulation, after emerging from the concurrence's analysis, has acquired both a proximate-cause limitation and a particular-animals limitation-precisely the one meaning that the Secretary's quoted declaration will not allow, whichever part of it is emphasized.
* * *
The Endangered Species Act is a carefully considered piece of legislation that forbids all persons to hunt or harm endangered animals, but places upon the public at large, *736 rather than upon fortuitously accountable individual landowners, the cost of preserving the habitat of endangered species. There is neither textual support for, nor even evidence of congressional consideration of, the radically different disposition contained in the regulation that the Court sustains. For these reasons, I respectfully dissent.