concurring in part and dissenting in part.
I join parts I, II, and IIIA of the majority’s opinion, but must respectfully dissent from part IIIB. The majority concludes that the EPA’s interpretation of the secondary treatment provisions is “permissible” and therefore valid under Chevron, U.S.A, Inc. v. Natural Resources Defense Council, Inc., 467 U.S. 837, 104 S.Ct. 2778, 81 L.Ed.2d 694 (1984). I cannot agree for two reasons. First, one “permissible” interpretation identified by the majority is not in fact advanced by the EPA. We cannot defer under Chevron to an agency construction when the agency has not construed the language at issue. Second, on this record, the EPA’s construction of the term “secondary treatment” is not permissible. While the Clean Water Act (“CWA”) gives the Administrator discretion to define secondary treatment pursuant to the statute, that discretion cannot be exercised in a manner inconsistent both with the structure and legislative history of the statute and with the Administrator’s own prior interpretation of the term. In allowing the substitution of quality-based controls for generally-applicable, technology-based effluent limitations, the majority allows the EPA to return clean water regulation to the pre-1972 era.
The EPA does not itself argue that the language of 33 U.S.C. § 1311(b)(1)(B) — that “there shall be achieved ... effluent limitations based upon secondary treatment”— gives it discretion to set effluent limitations lower than those deemed attainable through the application of secondary treatment. The majority’s resolution, to the extent it finds discretion for the EPA’s decision from the term “based upon,” see Maj. Op. at 1043 is premised on its own construction of the statute, not the EPA’s. That runs counter to the logic of Chevron deference and consequently to a core principle of judicial review of agency action. “If the basis stated by the agency for its decision is insufficient, we may not supply another that the agency itself has not chosen to rely on.” American Meat Inst. v. EPA 526 F.2d 442, 453 (7th Cir.1975) (citing SEC v. Chenery Corp., 332 U.S. 194, 196, 67 S.Ct. 1575, 1577, 91 L.Ed. 1995 (1947) (“[T]he court is powerless to affirm the administrative action by substituting what it considers to be a more adequate or proper basis. To do so would propel the court into the domain which Congress has set aside exclusively for the administrative agency.”)). If Congress has implicitly or explicitly left gaps in a statutory scheme, Chevron requires us to defer to reasonable efforts on the part of the agency to fill those gaps through policy and rule-making, see 467 U.S. at 843^44, 104 S.Ct. at 2781-83, and is explicitly concerned with the agency’s construction of congressional language to fill those gaps, id. Such gap-filling can only be upheld if the agency’s own rationale for its actions — including its construction of the statute — is proper.1
Reviewing the Agency’s denial of Maier’s petition, its briefs, as well as the material accompanying its earlier promulgation and amendment of secondary treatment regula*1047tions, I can find no suggestion from the Administrator that were she to find reductions of a particular pollutant attainable by means of secondary treatment, she would not need to promulgate a generally-applicable effluent limitation for that pollutant. Indeed, the Agency may hold a contrary view of its § 1311(b)(1)(B) discretion. In responding to Maier’s petition, the Administrator states that were technologies to control NOD considered to be secondary treatment, “[a]ny such revised secondary treatment requirements would be universally applicable to all POTWs pursuant to section 301(b)(1)(B) [33 U.S.C. § 1311(b)(1)(B)].” A.R. at 123-24; see also EPA Br. at 26 (‘Were NOD limitations to be made part of ‘secondary treatment,’ they would apply to all POTWs regardless of local conditions.”).
Moreover, the relevant regulatory history strongly suggests that the Agency would not interpret the “based upon” language in § 1311(b)(1)(B) to give it discretion to depart from reductions attainable by the technology described in 33 U.S.C. § 1314(d)(1). Effluent limitations on POTWs are set pursuant to 40 C.F.R. § 133.102. This regulation, which the Agency refers to as the Secondary Treatment Information regulation, consistently cites both § 1311(b)(1)(B) and § 1314(d)(1) as its statutory authority, see, e.g., 41 Fed.Reg. 37222 (1976), and has never suggested that reductions deemed attainable via secondary treatment need not be translated directly into applicable effluent limitations. Rather, the Agency has implicitly viewed the Secondary Treatment Information regulation as simultaneously satisfying both its information publication obligations under § 1314(d)(1) and its limitation promulgation obligations under § 1311(b)(1)(B). See, e.g., 42 Fed.Reg. 54664 (1977) (“The Secondary Treatment Information regulation contains effluent limitations in terms of biological oxygen demand, suspended solids and pH which must be achieved by municipal wastewater treatment plants ... in accordance with section 301(b)(1)(B) of the ... FWPCA. The Secondary Treatment Information regulation was promulgated pursuant to section 304(d)(1) of the FWPCA.”).
In fact, the Agency appears to regard the Secondary Treatment Information regulation as simultaneously defining secondary treatment and establishing the effluent limitations applicable to POTWs. See 41 Fed.Reg. 37222 (1976) (“Secondary treatment (as defined in 40 C.F.R. 133) is the minimum level of treatment required for all publicly-owned treatment works.”); 49 Fed.Reg. 36987 (1984) (“The secondary treatment regulation defines ‘secondary treatment’ as attaining an average effluent quality for both biochemical oxygen demand, five-day (BOD 5) and SS of 30 mg/1 in a period of 30 consecutive days, an average effluent quality of 45 mg/1 for the same pollutants in a period of 7 consecutive days, and 85 percent removal of the same pollutants in a period of 30 consecutive days.”). Agency practice has thus never recognized a disjunction between its obligation to publish attainable reductions under § 1314(d)(1) and to promulgate effluent limitations under § 1311(b)(1)(B). Yet the majority’s “based upon” analysis would create this disjunction and effectively attribute it to the Agency’s discretion.
The Agency claims “considerable discretion ... to define ‘secondary treatment.’ ” EPA Br. at 27. Exercising this definitional discretion, the Agency asserts that controls on NOD and nutrients “simply should not be required as part of ‘secondary treatment.’ ” Id. at 25. Were the Administrator responding to Maier’s petition in a regulatory vacuum, we might be required to defer to this agency definition of secondary treatment. But that is not the case. The secondary treatment regulations have always set controls on biological oxygen demand (BOD), see 38 Fed.Reg. 10642 (1973) (defining minimum level of BOD reduction attainable through application of secondary treatment), and such “gap-filling” appears entirely consistent with the applicable legislative history.2 Moreover, as noted above, BOD controls, in conjunction with those imposed on certain other pollu*1048tants such as suspended solids, have been administratively regarded as defining secondary treatment.
The Agency recognizes that NOD is one of two components of BOD, the other being carbonaceous BOD (or “CBOD”). See, e.g., 48 Fed.Reg. 52272, 52274 (1983). Maier’s petition therefore requests the Agency to apply specific controls to a pollutant whose restriction falls broadly within the administrative and legislative understanding of secondary treatment. Of course, given the EPA’s statutorily-conferred discretion to achieve “effluent limitations based upon secondary treatment,” § 1311(b)(1)(B), the EPA may not be obliged to impose secondary treatment-based controls on NOD. But having included the control of oxygen-depleting compounds within the general definition of secondary treatment, it is incumbent upon the EPA to explain its refusal to promulgate NOD and nutrient limitations.
In the past, the Administrator has principally explained the refusal to treat NOD controls as part of secondary treatment as proceeding from the impracticality of such controls. See, e.g., 49 Fed.Reg. 36986, 36988 (1984); 48 Fed.Reg. 52272, 52273 (1983) (citing supporting documentation for 1973 regulations). Such a decision, if adequately supported by the record, is well within the Administrator’s rulemaking discretion. In denying Maier’s petition, the EPA now points to two factors. First, the Agency reiterates that secondary treatment is concerned with the removal of carbonaceous organic material. This argument begs the question. If Maier’s petition questions the EPA’s earlier conclusion as to non-attainability, the Agency should explain its continued reliance on its previous explanation. The EPA has not done so, instead insisting that any new information on attainability submitted by Maier “does not establish (or even suggest) why control of nitrogen or phosphorus should be considered secondary treatment.” A.R. at 125.
That error alone would not require us to remand to the Agency were the second factor relied on by the Administrator to deny Mai-er’s petition more persuasive. It is not. The Agency’s second defense of its secondary treatment regulations is to point to its policy preference for quality-based controls rather than generally applicable limitations, at least for pollutants that do not have a uniform impact on receiving bodies of water. See A.R. at 113-14,123; see also EPA Br. at 19-20. The EPA may yet have good reasons for refusing to regulate NOD via generally-applicable effluent limitations on POTWs, but a policy preference for quality-based measures over generally-applicable technology-based measures is not one of them. Such a preference improperly construes the CWA.
Before 1972, the stated purpose of the Federal Water Pollution Control Act (“FWPCA”) was “to enhance the quality and value of our water resources and to establish a national policy for the prevention, control, and abatement of water pollution.” 33 U.S.C. § 1151(a) (1970) (superseded by Pub.L. 92-500, § 2, 88 Stat. 816 (1972)). To this end, the pre-1972 legislation employed ambient water quality standards as the primary mechanism for water pollution control. See EPA v. California ex rel. State Water Resources Control Bd., 426 U.S. 200, 202, 96 S.Ct. 2022, 2023, 48 L.Ed.2d 578 (1976). The 1972 Amendments to the FWPCA, popularly known as the Clean Water Act, deliberately ended this approach. Prompted by the Senate Committee on Public Works’ review of the FWPCA program, and its conclusion that “the national effort to abate and control water pollution has been inadequate in every vital respect,” S.Rep. 92-414, at 7, reprinted in 1972 U.S.C.C.A.N. 3668, 3674, Congress declared as the new national goal of the program that “the discharge of pollutants into the navigable waters be eliminated,” 33 U.S.C. § 1251(a)(1). Consistent with this end, the CWA substituted technology-based, generally-applicable effluent limitations for water quality-based regulatory approaches. See State Water Resources Control Bd., 426 U.S. at 204, 96 S.Ct. at 2024 (“Such direct restrictions on discharges facilitate enforcement by making it unnecessary to work backward from an overpolluted body of water to determine which point sources are responsible and which must be abated.”). The leg*1049islative history of the Act is replete with references to the need for this substitution.3
The EPA’s denial of Maier’s petition effects an entirely opposite substitution. In order for an administrative construction that runs counter to basic policies underlying the relevant statutory scheme to be reasonable under the second step of Chevron, the implementing agency must point to some language in the statute to justify its policy conclusion — here, that the POTW regulatory regime can legitimately depart from the core public policy of the CWA.4 The Administrator has not done so. Her passing reliance on § 1311(b)(1)(C), which allows the Administrator to set “more stringent limitation[s]” to meet water quality standards, is misplaced. At most, that provision allows the Administrator to set quality-based limits for pollutants that cannot be attainably reduced by secondary treatment, or to set supplementary quality-based limits for pollutants already regulated by a floor of generally-applicable limitations based on secondary treatment. It cannot reasonably be read as general discretion to redefine secondary treatment to cover only those pollutants that are — in the view of the Administrator — more appropriately regulated via generally-applicable regulations rather than case-by-case quality-based limits. That interpretation makes a mockery of the primacy accorded technology-based regulation by the plain language and legislative history of the CWA.
In fact, Congress has itself confirmed that POTWs are not exempted from this core policy. In 1977, Congress enacted 33 U.S.C. § 1311(h), which permits the Administrator, on a case-by-ease basis, to relax secondary treatment requirements for POTWs releasing pollutants into marine waters. See 33 U.S.C. § 1311(h). If the EPA has the discretion relied on by the Administrator in the present case, this provision was (and is) entirely unnecessary. The Administrator could simply declare that the biological treatment of pollutants by POTWs that release into marine waters is not “secondary treatment” because she has made the policy choice to address such discharges solely through individual NPDES permit requirements. In passing § 1311(h), Congress effectively stated that such discretion was not open to the Administrator. Cf. Bridger Coal Co. v. Director, Office of Workers’ Compensation Programs, 927 F.2d 1150, 1153 (10th Cir.1991) (statute should be interpreted to give meaning and effect to each provision). It is not our place to offer discretion to the Agency where Congress has not.5
*1050On a number of occasions, the Agency has itself confirmed that “effluent limitations based upon secondary treatment” cannot be fixed by reference to quality-based considerations. See, e.g., 38 Fed.Reg. 22298 (1973) (POTW effluent limitation regulation “is to be based on the capabilities of secondary treatment technology and not ambient water quality”); 41 Fed.Reg. 30786, 30788 (1976) (same). In denying Maier’s petition, the Administrator alludes to this constraint, see A.R. at 117 (“[T]he definition of secondary treatment is to be technology-based rather than water quality based”), then ignores it without explanation.
The denial of Maier’s petition must be “based on a consideration of the relevant factors.” Citizens to Preserve Overton Park, Inc. v. Volpe, 401 U.S. 402, 416, 91 S.Ct. 814, 824, 28 L.Ed.2d 136 (1971). Here, the EPA’s denial is based on one factor that is illegitimate — its “reasoned” policy preference for quality-based over generally-applicable, technology-based restrictions — and another that is legitimate but unsubstantiated — the nonattainability of NOD reductions.6 I would remand the petition to the Agency for reconsideration in light of the correct legal principles. See American Horse Protection Ass’n v. Lyng, 812 F.2d 1, 7-8 (D.C.Cir.1987).
. The majority argues that I fail to distinguish between "the source of agency discretion, which we must determine in the first instance under Chevron, and the basis for the agency’s exercise of its discretion, for which ... we may not supply our own rationale.” Maj. Op. at 1043 n. 17. The majority’s distinction is untenably semantic because an agency’s exercise of discretion under Chevron must be based on its claimed statutory source of discretion. In reviewing the former, a court is bound to review the latter. The majority would have the reviewing court independently root through the statute on its own cognizance looking for gaps that the agency’s policymaking might permissibly back-fill. What the majority has done is to identify an ambiguous portion of the statute, i.e. the "based upon” language of § 1311(b)(1)(B), and defended the agency's failure to promulgate generally-applicable NOD effluent limitations on the basis of the court's own construction of that ambiguous term. Quite aside from the fact that the EPA has implicitly disavowed this particular construction of the statute, Chevron contains absolutely no authorization for approving administrative constructions in this manner. The most that the majority can conceivably make of the "based upon” language of § 1311(b)(1)(B) is that the administrative construction of "secondary treatment” does not violate that particular statutory provision. This view is probably unsustainable in light of the EPA's previous interpretation of § 1311(b)(1)(B). But even assuming its validity, the majority’s view fails to show that the administrative construction of "secondary treatment” does not fall afoul of some other provision of the CWA.
. “Secondary treatment as considered in the context of a publicly-owned treatment works is generally concerned with suspended solids and biologically degradable, oxygen demanding materials (BOD)." H.Rep. No. 92-911, 92d Cong., 2d Sess., at 101 (1971), quoted in Proposed Rule, 48 Fed.Reg. 52272, 52273 (1983).
. The Senate Report accompanying the CWA notes: "The application of Phase I technology to industrial point sources is based upon the control technologies for those sources and to publicly owned sewage treatment works is based upon secondary treatment. It is not based upon ambient water quality considerations." S.Rep. 92-414, at 43, reprinted in 1972 U.S.C.C.A.N. 3668, 3710 (emphasis added).
. Contrary to the majority’s assertion, placing this obligation on the Administrator does not “tum[] the Chevron test on its head.” Maj. Op. at 1045. Chevron authorizes the Administrator to fill legislative gaps, but only when done in compliance with her statutory policymaking discretion. Chevron, 467 U.S. at 843-44. Here, the EPA has without justification chosen to fill a gap by means of a policy that contravenes the most fundamental tenet of the CWA. See Maislin Indus., U.S. v. Primary Steel, 497 U.S. 116, 134-35, 110 S.Ct. 2759, 2770, 111 L.Ed.2d 94 (1990) (agency "does not have the power to adopt a policy that directly conflicts with its governing statute”). Without some language suggesting that POTWs are exempt from the force of this basic statutory imperative, the EPA's regulatory inaction must be regarded as "manifestly contraiy to the statute,” and accordingly invalid under Chevron. 467 U.S. at 844, 104 S.Ct. at 2782.
. The majority states that it is not authorizing the EPA's exercise of general policy discretion to substitute quality-based restrictions for generally-applicable, technology-based effluent limitations, but is instead restricting the EPA's discretion to cases in which it advances a "reasoned explanation” for this substitution. See Maj. Op. at 1045. I am not so sure. Nowhere does the CWA suggest that its clear technology-first imperative is subject to cancellation by the agency’s "reasoned explanations." Nor does the majority explain why in the absence of statutory authorization, the EPA is free to ignore that imperative on the basis of its own "reasoned explanations.” See Director, Office of Workers' Comp. v. Newport News, 514 U.S. 122, 136, 115 S.Ct. 1278, 1288, 131 L.Ed.2d 160 (1995) ("Every statute proposes, not only to achieve certain ends, but also to achieve them by particular means____ The withholding of agency authority is as significant as the granting of it, and we have no right to play favorites between the two."). Finally, the majority fails to explain why the qualitative variability of pollutant discharges constitutes such a “reasoned ex*1050planation," or what other types of agency explanation would or would not allow for a similar departure from the basic public policy of the CWA. Without some limiting principle, it is hard not to conclude that the majority is essentially deferring to the EPA's policy preference for quality-based standards.
. The majority states that I view technological feasibility as the "only criterion” that the EPA can use to define secondary treatment. Maj. Op. at 1042. That is incorrect. Our review is appropriately confined to the reasons given by the EPA for its denial of Maier's petition. In my view, the only argument the EPA offers that is not "manifestly contrary to the statute,” is one based on unsubstantiated claims of technological feasibility. This should not be contorted to mean that the EPA’s only possible basis for defining secondary treatment is technological feasibility.