FLETCHER, Circuit Judge:
The Natural Resources Defense Council (NRDC) petitions for review of a final rule issued by the Environmental Protection Agency (EPA). The rule provides that with regard to some, but not all, of the polluted waters listed pursuant to section 304(Z) of the Clean Water Act, 33 U.S.C. § 1314(Z), the states must identify the factories and other “point sources” responsible for discharging toxic pollutants into those waters and must develop strategies to control the pollution from those sources in an expedited manner. 40 C.F.R. §§ 123.-46, 130.10. The NRDC argues that with regard to
all
of the listed waters, the states must identify “point source” toxic polluters and must develop strategies to control all the sources identified.
We grant the petition with respect to the claim that identification of toxic polluters must be made for all listed waters and remand for EPA to reconsider the question of individual control strategies.
I.
STATUTORY BACKGROUND
The Water Quality Act of 1987 (WQA), Pub.L. No. 100-4, 101 Stat. 7, amended the Clean Water Act (CWA), 33 U.S.C. §§ 1251
et seq.,
adding a number of new provisions, including section 304(Z), 33 U.S.C. § 1314(Z), which is the focus of this peti
tion. Section 304(i) refers to other provisions in the Clean Water Act; its proper construction requires a familiarity with the history, the structure, and, alas, the jargon of the federal water pollution laws.
A.
Prior to 1972, Congress attempted to control water pollution by focusing regulatory efforts on achieving “water quality standards,” standards set by the states specifying the tolerable degree of pollution for particular waters.
See EPA v. State Water Resources Control Board,
426 U.S. 200, 202-03, 96 S.Ct. 2022, 2023-24, 48 L.Ed.2d 578 (1976). This scheme had two important flaws. First, the mechanism of enforcement was cumbersome. Regulators had to work backward from an overpollut-ed body of water and determine which entities were responsible; proving cause and effect was not always easy. Second, the scheme failed to provide adequate incentives to individual entities to pollute less; an entity’s dumping pollutants into a stream was ignored if the stream met the standards.
Id.
The scheme focused on “the tolerable effects rather than the preventable causes” of pollution.
Id.
In 1972, Congress passed the Clean Water Act, which made important amendments to the water pollution laws. The amendments placed certain limits on what an individual firm could discharge, regardless of whether the stream into which it was dumping was overpolluted at the time. Firms were required to use progressively more advanced technology; by 1977 they were to use the “best practicable control technology,” CWA § 301(b)(1)(A), 33 U.S.C. § 1311(b)(1)(A), and by 1987 at the latest they were to use the more demanding “best available technology” to limit the discharge of pollutants. CWA § 301(b)(2)(A), 33 U.S.C. § 1311(b)(2)(A); CWA § 402, 33 U.S.C. § 1342. With regard to toxic pollutants listed pursuant to CWA § 307, 33 U.S.C. § 1317,
compliance with the “best available technology” was required by 1984. CWA § 301(b)(2)(A). The limits on discharges were to be effectuated by a system of permits, the National Pollution Discharge Elimination System (NPDES). Without a permit, no person could “discharge ... any pollutant.” CWA § 301(a), 33 U.S.C. § 1311(a). Section 301(a)’s ban on the discharge of pollutants sounded bolder than it really was. The term “discharge of any pollutant” was a statutorily defined term meaning, “any addition of any pollutant to navigable waters
from any point
source.” CWA § 502(12), 33 U.S.C. § 1362(12). The Clean Water Act defined a “point source” as a discrete location from which pollutants could be discharged, such as a pipe or drain from a factory. CWA § 502(14), 33 U.S.C. § 1362(14).
The Act thus banned only discharges from point sources. The discharge of pollutants from nonpoint sources — for example, the runoff of pesticides from farmlands — was not directly prohibited.
The Act focused on point source polluters presumably because they could be identified and regulated more easily than nonpoint source polluters.
Congress, in passing the Clean Water Act, thus shifted the focus of the water pollution laws away from the enforcement of water quality standards and toward the
enforcement of technological standards. But Congress recognized that even if all the firms discharging pollutants into a certain stream segment were using the best available technology, the stream still might not be clean enough to meet the water quality standards set by the states. To deal with this problem, Congress supplemented the “technology-based” limitations with “water-quality-based” limitations.
See
CWA §§ 302, 303, 33 U.S.C. § 1312, 1313.
The water quality standard for a particular stream segment was to be determined in the following manner. First, the state in which the stream segment was located was to designate the uses to which it wished to put the segment. The designations that the states had made prior to the 1972 Clean Water Act were deemed to be the initial designations under that Act; however, states were thereafter to review their designations at least once every three years. CWA § 303(c)(1), 33 U.S.C. § 1313(c)(1). Pursuant to the statute’s policy that the designation of uses “enhance” the quality of water, CWA § 303(c)(2), 33 U.S.C. § 1313(c)(2), EPA enacted regulations setting limits on the states’ ability to downgrade previously designated uses. If a state wished to redesignate a use so that the new use did not require water clean enough to meet the statutory goal of fisha-ble, swimmable water,
see
CWA § 101(a)(2), 33 U.S.C. § 1251(a)(2), it had to conduct a “use attainability analysis” as a condition to federal approval of the redesig-nated use. CWA § 303(c)(3), 33 U.S.C. § 1313(c)(3); 40 C.F.R. §§ 131.10®, 131.-3(g) (1989). If the result of the “use attainability analysis” was that it was feasible to attain fishable, swimmable waters, EPA would reject the redesignated use.
Second, the state was to determine the “criteria” for each segment — the maximum concentrations of pollutants that could occur without jeopardizing the use. These criteria could be either numerical (e.g. 5 milligrams per liter) or narrative (e.g. no toxics in toxic amounts). The criteria, like the uses, were subject to federal review. The EPA was to reject criteria that did not protect the designated use or that were not based on a “sound scientific rationale.” 40 C.F.R. § 131.11 (1989).
Under sections 301(b)(1)(C) and 402(a)(1), 33 U.S.C. §§ 1311(b)(1)(C), 1342(a)(1), NPDES permit writers were to impose, along with the technology-based limitations, any more stringent limitations on discharges necessary to meet the water quality standards. Although ostensibly they were supposed to impose these more stringent limitations, in practice they often did not.
One explanation for this failure is that the criteria listed by the states, particularly for toxic pollutants, were often vague narrative or descriptive criteria as opposed to specific numerical criteria. These descriptive criteria were difficult to translate into enforceable limits on discharges from individual polluters. As one commentator put it:
The descriptive criteria, in particular, call for both expert testimony and a receptive forum to transform, let us say, a general obligation to maintain ‘recreational’ uses into a specific obligation to reduce loadings of phosphorus or nitrogen from a particular source. The decision requires, among other things, judgments about the degree of algal bloom that interferes with ‘recreational’ uses such as swimming or boating, estimates of loadings from all contributing point and nonpoint sources, assumptions about degrees of control elsewhere, and predictions of how a water segment will respond to a hoped-for change of parameters.
Rodgers, 2 Environmental Law § 4.16 at 250-51 (1986). The Clean Water Act dealt with the difficulty of these decisions and judgments in various ways, for example by calling for the publication by the EPA of criteria documents spelling out causes and effects of various pollutant loads,
see
CWA § 304(a), 33 U.S.C. § 1314(a), and by requiring states to set total maximum daily loads for certain pollutants (but, notably, not for toxic pollutants) CWA § 303(d)(1), 33 U.S.C. § 1313(d)(1); however, the complexity of these decisions and judgments led many a permit writer to avoid making them altogether. Rodgers, § 4.18 at 283-84.
B.
In 1987 Congress reexamined the water pollution laws. It found that the requirement that individual polluters use the best available technology was not sufficient to solve the pollution problem, particularly the problem of toxic pollutants; a renewed emphasis on water quality-based standards was necessary. Congress enacted a number of new provisions. Three are relevant for our purposes.
CWA section 319, 33 U.S.C. § 1329, requires states to submit for federal approval nonpoint source reports and management programs by August 4, 1988, identifying specific nonpoint sources of pollution and setting forth a plan for implementing the “best management practices” to control such sources by 1992. Section 319 does not require states to penalize nonpoint source polluters who fail to adopt best management practices; rather it provides for grants to encourage the adoption of such practices.
CWA section 303(c)(2)(B), 33 U.S.C. § 1313(c)(2)(B), requires states to adopt “specific numerical criteria” for toxics for which the EPA has published criteria pursuant to section 304(a), 33 U.S.C. § 1314(a). Those criteria are to be adopted when the state first reviews its water quality standards following the enactment of the 1987 amendments. The requirement of numerical criteria for toxics makes it easier for permit writers to incorporate the water quality standards into NPDES permits. Permit writers thus no longer have an excuse for failing to impose water-quality-based limitations on permit holders.
In addition to requiring the adoption of numerical criteria, Congress enacted new CWA section 304(i), 33 U.S.C. § 1314(i), the section directly at issue in this petition. Certain aspects of section 304(Z) are not in dispute. We briefly explain these background aspects to bring the disputed issues into sharper focus.
Section 304(Z) provides:
“(Z) Individual Control Strategies for Toxic Pollutants.
(1) State List of Navigable Waters and Development of Strategies. Not later than 2 years after the date of the enactment of this subsection [February 4, 1987], each State shall submit to the Administrator for review, approval, and implementation under this subsection—
(A) a list of those waters within the State which after the application of effluent limitations required under section 301(b)(2) of this Act cannot reasonably be anticipated to attain or maintain (i) water quality standards for such waters reviewed, revised, or adopted in accordance with section 303(c)(2)(B) of this Act, due to toxic pollutants, or (ii) that water quality which shall assure protection of public health, public water supplies, agricultural and industrial uses, and the protection and propagation of a balanced population of shellfish, fish and wildlife, and allow recreational activities in and on the water;
(B) a list of all navigable waters in such State for which the State does not expect the applicable standard under section 303 of this Act will be achieved after the requirements of sections 301(b), 306, and 307(b) are met, due entirely or substantially to discharges from point sources of any toxic pollutants listed pursuant to section 307(a);
(C) for each segment of the navigable waters included on such lists, a determination of the specific point sources discharging any such toxic pollutant which is believed to be preventing or impairing such water quality and the amount of each such toxic pollutant discharged by each such source; and
(D) for each such segment, an individual control strategy which the State determines will produce a reduction in the discharge of toxic pollutants from point sources identified by the State under this paragraph through the establishment of effluent limitations under section 402 of this Act and water quality standards under section 303(c)(2)(B) of this Act, which reduction is sufficient, in combination with existing controls on point and non-point sources of pollution, to achieve the applicable water quality standard as soon as possible, but not later than 3 years after the date of the establishment of such strategy.
(2) Approval or Disapproval. Not later than 120 days after the last day of the 2-year period referred to in paragraph (1), the Administrator shall approve or disapprove the control strategies submitted under paragraph (1) by any State.
(3) Administrator’s Action. If a State fails to submit control strategies in accordance with paragraph (1) or the Administrator does not approve the control strategies submitted by such State in accordance with paragraph (1), then, not later than 1 year after the last day of the period referred to in paragraph (2), the Administrator, in cooperation with such State and after notice and opportunity for public comment, shall implement the requirements of paragraph (1) in such State. In the implementation of such requirements, the Administrator shall, at a minimum, consider for listing under this subsection any navigable waters for which any person submits a petition to the Administrator for listing not later than 120 days after such last day.
Section 304(Z) requires the preparation of three lists. The list required by section 304(i)(l)(B) (hereinafter the “B list”) is the narrowest of the three lists. It consists only of waters that are not expected to meet water quality standards, even after the application of the technology-based limitations, due entirely or substantially to toxic pollution from
point sources.
The list required by section 304(i )(l)(A)(i) (hereinafter the “A(i) list”) is broader; it includes most of the waters on the B list
plus waters expected not to meet water quality standards due to pollution attributable entirely or almost entirely to toxic pollution from
nonpoint sources.
The list required by section 304(Z )(l)(A)(ii) (hereinafter the “A(ii) list”) is the broadest. It includes all the waters on the other two lists plus any waters which, after the implementation of technology-based controls, are not expected to meet the water quality goals of the Act; since the goals of the Act are sometimes higher than the state standards, the A(ii) list includes even some waters expected to comply fully with applicable water quality standards.
The effect of the individual control strategies is simply to expedite the imposition of water-quality-based limitations on polluters — limitations which otherwise would have had to be imposed when the polluters’ NPDES permits expired. NPDES permits are issued for periods of no more than five years, although administrative delays can extend
de facto
the duration of the permits.
The EPA has promulgated a number of regulations interpreting the statute; two are particularly important for our purposes. The first regulation, codified at 40 C.F.R. § 130.10(d), interprets sections 304(i )(1)(A), (B), and (C).
The first two
subsections of that regulation, subsections 130.10(d)(1) and (2), simply track the language of subsections 304(i)(l)(A) and (B) respectively. The next subsection of the regulation, section 130.10(d)(3), does
not,
however, track subsection 304(i )(1)(C); rather it changes the word “lists” in the statute to “list.” By referring to list in the singular, it excludes the A(i) and A(ii) lists from the requirement of section 304(i )(1)(C) that point sources of toxic pollution and the amount of pollution discharged for each source be identified. Only the B list is subject to the identification requirement.
The second regulation, codified at 40 C.F.R. § 123.46,
interprets section 304(i )(1)(D). It provides that individual control strategies (ICS’s) must be prepared only for the point sources identified in section 304(i )(1)(C), as interpreted, of course, by the first regulation. Thus, under the regulations, ICS’s are required only in connection with waters on the B list. The A(i) and A(ii) lists are not to be consulted in determining which segments require ICS’s.
The NRDC petitioned for review of these regulations, arguing they are inconsistent with the statute. We have jurisdiction under CWA § 509(b)(1)(E), 33 U.S.C. § 1369(b)(1)(E).
II.
DISCUSSION
We review two separate regulations. The first one, 40 C.F.R. § 130.10(d)(3), interprets section 304(i )(1)(C). The second, 40 C.F.R. § 123.46, interprets section 304(Z )(1)(D). We consider each regulation in turn.
EPA argues that its interpretation of section 304(Z)(1)(C) is entitled to special deference on review. In
Chevron U.S.A. v. Natural Resources Defense Council,
467 U.S. 837, 843, 104 S.Ct. 2778, 2781, 81 L.Ed.2d 694 (1984), the Supreme Court held:
[Wjhen a court reviews an agency’s construction of the statute which it administers, it is confronted with two questions. First, always, is the question whether Congress has spoken directly to the precise question at issue. If the intent of Congress is clear, that is the end of the matter; for the court, as well as the agency, must give effect to the unambiguously expressed intent of Congress.... [I]f the statute is silent or ambiguous with respect to the specific issue, the question for the court is whether the agency’s answer is based on a permissible construction of the statute.
In this case, Congress has spoken directly, in unambiguous terms, to the question whether subsection 304(Z )(1)(C) requires the identification of point sources discharging toxics into the waters identified on all three lists. By using the plural “lists,” Congress foreclosed EPA from restricting the scope of paragraph C to waters on the B list. Since the language of paragraph C is unambiguous, there is no need to resort to extrinsic sources to interpret the statute.
Cf. Green v. Commissioner,
707 F.2d 404, 405 (9th Cir.1983).
The EPA makes two arguments as to why paragraph C is ambiguous. First, it points to the caption of § 304(Z)(1), which says, “State List of Navigable Waters and Development of Strategies.” EPA argues that the use of the singular “List” in the caption creates an ambiguity, thus triggering Chevron’s requirement of deference. While words in the title of a statute or the heading of a section can shed light on the meaning of an ambiguous word or phrase in the text of a statute, they cannot
create
an ambiguity where none otherwise would exist.
Brotherhood of Railroad Trainmen v. B. & O. Railroad,
331 U.S. 519, 528-29, 67 S.Ct. 1387, 1391-92, 91 L.Ed. 1646 (1947). Since the text is not ambiguous, the caption does not aid our interpretation.
EPA attempts to distinguish
Trainmen
by arguing that it is not offering the caption as evidence of ambiguity but rather as evidence that Congress made a drafting error. But, other than the caption, EPA offers no evidence indicating that the text is the product of a drafting error. When the caption is the only evidence of a drafting error, there is no good reason to assume that it is the text and not the caption that is erroneous. The ordinary presumption is that Congress’ drafting of the text is deliberate.
United States v. Motamedi,
767 F.2d 1403, 1406 (9th Cir.1985). Moreover in this case, even absent that presumption, there is good reason to believe that the use of the plural “lists” is not a drafting error. The conference committee which drafted section 304(i) fused elements of House and Senate bills containing similar provisions. The House version of section 304(Z) required only one list to be prepared, corresponding to the B list. Cong.Research Service,
A Legislative History of the Water Quality Act of 1987,
Comm. Print No. 1, 100th Cong., 2d Sess. at 1186. Since only one list had to be prepared, the House equivalent to section 304(Z)(1)(C) required the identification of point sources for the waters on “such list.”
Id.
The Senate version of section 304(Z) required two lists, corresponding to the A(i) and A(ii) lists.
Id.
at 1557. The conferees determined that all three lists should be prepared. This change, in turn, necessitated changing the House version’s phrase, “such list,” since leaving that phrase intact would have created an ambiguity. Changing the phrase to “such lists” made clear that identification of point sources on all three lists was required. In these circumstances, it can hardly be argued that the change to the plural was inadvertent.
The EPA’s next argument for its interpretation of section 304(Z)(1)(C) is more complex. The EPA starts not with the language of paragraph C, but with the language of paragraph D. EPA asserts several propositions concerning paragraph D: that in referring to “effluent limitations under section 402 of this Act,” paragraph D necessarily refers only to limitations imposed on
-point sources,
since it is only point sources that are subject to the NPDES permitting process described in section 402; that paragraph D requires individual control strategies which will cause a reduction in toxic pollutants “sufficient, in combination with existing controls on point and nonpoint sources of pollution, to achieve the applicable water quality standard ... not later than three years after the date of the establishment of such strategy;” and that since ICS’s are not required under paragraph D when they cannot be expected to achieve water quality standards within the specified time, Congress must have intended ICS’s to be required only for polluters discharging into streams whose failure to meet standards could be cured by eliminating discharges of toxics from point sources. From these propositions, the validity of which we do not now decide, EPA then arrives at two controversial conclusions: that ICS’s are required only for polluters discharging into the streams on the B list, streams whose failure to meet water quality standards is due “entirely or substantially to discharges from point sources;” and that because only B list waters require ICS’s under paragraph D, only those waters are subject to the identification requirement of paragraph C. EPA argues that the prepositional phrase introducing paragraph D, “for each such segment,” compels these conclusions
because the phrase must refer only to segments that will require ICS’s.
We disagree. Even if EPA’s interpretation of paragraph D is proper, which we do not decide here, its interpretation of paragraph C cannot stand. In using the phrase, “for each such segment,” Congress was simply requiring the states to consult each segment before determining whether an ICS on that segment could achieve water quality standards within the relevant period of time. We acknowledge that if the use of the plural “lists” in paragraph C were treated as a drafting error, the phrase “for each segment” would make paragraphs B, C, and D flow together more smoothly. But a statute is not ambiguous simply because an agency can suggest a change in wording that would make the statute more elegant. Since paragraph C as drafted neither is ambiguous in its terms nor is incoherent when considered together with the other provisions of the statute to which it relates, we do not accord EPA’s regulation redrafting paragraph C any special deference on review.
Cf Chevron.
Reviewing the regulation
de novo,
we conclude that EPA reached the wrong interpretation of paragraph C because it started with a faulty premise. EPA assumed that paragraph D, requiring individual control strategies for certain waters, was the only significant provision of section 304(i )(1) and that paragraphs B and C of that section had to be read as having one purpose and one purpose only—to effectuate paragraph D. This assumption has two flaws, one obvious and one more subtle. The obvious flaw is that the assumption utterly fails to account for the presence of paragraph A—especially paragraph A(ii). If Congress was interested only in individual control strategies for toxic pollutants, why would it have wanted a list of waters whose failure to meet the goals of the Act was not necessarily traceable to toxic pollutants? One readily can infer from the presence of paragraph A that Congress wanted certain information not necessarily because it would affect the ICS program but because it might subsequently be useful in formulating other statutory or regulatory programs. There are other provisions in the Clean Water Act which require the gathering of information but which do not necessarily require immediate action on the basis of the information.
See, e.g.,
CWA § 305, 33 U.S.C. § 1315; CWA § 303(d), 33 U.S.C. § 1313(d).
The more subtle flaw in EPA’s assumption is that it does not account for the purposes that paragraph C might serve. Like paragraph A, paragraph C will produce useful information. It requires the identification of point sources discharging toxic pollutants and the determination of the amount of such pollutants discharged. Such information may prove useful to regulators even if every point source identified does not require an ICS.
In sum, we hold that EPA erred in assuming that paragraph D and paragraph C of subsection 304(i) must perfectly interlock. Since the provisions do not serve the identical purpose, there was no need to distort paragraph C in order to make it connect better with paragraph D.
B.
Our determination that EPA erred in interpreting section 304(i )(1)(C) does not settle the issue on which the parties have focused most of their attention: which waters are subject to the ICS’s required by section 304(i )(1)(D)?
EPA’s position is that only waters on the B list,
i.e.,
waters for which the state does not expect water quality standards to be achieved “due entirely or substantially to point sources of any toxic pollutants” are
subject to the ICS requirement. EPA interprets “entirely or substantially” broadly:
If a water meets either of the two conditions listed below the water must be listed [under paragraph B] on the grounds that the applicable standard is not achieved or expected to be achieved due entirely or substantially to discharges from point sources.
(i) Existing or additional water quality-based limits on one or more point sources would result in the achievement of an applicable water quality standard for a toxic pollutant; or
(ii) The discharge of a toxic pollutant from one or more point sources, regard-' less of any nonpoint source contribution of the same pollutant, is sufficient to cause or is expected to cause an excursion above the applicable water quality standard for the toxic pollutant.
40 C.F.R. § 130.10(d)(5) (1989).
Consider the hypothetical situation of a stream that can absorb a load of 100 pounds per day of a particular toxic substance without violating water quality standards. In June of 1992 (the date by which ICS’s are required to achieve their purpose) after existing controls are implemented, it is expected that 105 pounds per day of the toxic will flow through the segment. If six of the expected 105 pounds are to come from point sources, the point source contribution is considered “substantial” under EPA’s regulation and ICS’s are required for the point sources. If three of the 105 pounds are expected to come from point sources, the contribution is considered “insubstantial” and no ICS’s are required. The reason why the six pound contribution is considered substantial is that a reduction of six pounds would be “sufficient, in combination with existing
controls on point and nonpoint sources of pollution, to achieve the applicable water quality standard.” The language just quoted comes, of course, from paragraph D, which specifies when ICS’s are required. EPA thus derived its interpretation of paragraph B essentially by beginning with its interpretation of paragraph D and working backward. That is the same method EPA used to interpret paragraph C. Since we are remanding to the agency to have it promulgate new regulations under paragraph C, we do not decide whether EPA’s current interpretation of paragraph D is too restrictive. (It is not too inclusive.
) Rather we invite EPA to reconsider its interpretation on remand. In the meantime, until EPA promulgates new regulations, the program shall continue.
III.
CONCLUSION
We grant the petition for review in part and remand. On remand, EPA must, pur
suant to CWA § 304(i )(1)(C), amend its regulations to require the states to identify all point sources discharging any toxic pollutant which is believed to be preventing or impairing the water quality of any stream segment listed under CWA §§ 304(i)(l)(A) and (B) and to indicate the amount of the toxic pollutant discharged by each source. EPA shall also reconsider its interpretation of CWA § 304(Z)(1)(D).
REMANDED.