Section 313 of the Emergency Planning and Community Right-To-Know Act (“EPCRA” or the “Act”), 42 U.S.C. § 11023, 100 Stat. 1613, 1741 (Oct. 17, 1986), is captioned “Toxic chemical release forms.” The Act calls for the creation of a Toxic Release Inventory List (“TRI”). EPA acknowledged in the district court that “Congress intended that the TRI would contain only toxic chemicals.” EPA Memorandum in Support of Cross-Motion for Summary Judgment at 20 (June 6, 2003) (“EPA’s Summary Judgment Memorandum”) (internal quotation marks omitted). Congress used the phrase “toxic chemical” 38 times in § 313. A naive observer might think that the section’s sole *403subject is toxic chemicals. He would be right. A naive observer might also think it obvious that that was so. He would be wrong.
In 1996, the American Chemical Council petitioned EPA to delete methyl ethyl ketone (“MEK”) from the TRI, see Methyl Ethyl Ketone; Toxic Chemical Release Reporting; Community Right-To-Know, 63 Fed.Reg. 15195, 15196 (March 30, 1998) (“MEK Petition Denial ”), arguing principally that MEK is not a toxic chemical as that term is used in the Act. Joint Appendix (“J.A.”) 96. EPA denied the petition, MEK Petition Denial, 63 Fed.Reg. at 15195, 15199, and the court below granted EPA’s motion for summary judgment, holding that the decision not to delist MEK was reasonable, based on a permissible construction of the statute, and neither arbitrary nor capricious. See American Chemistry Council v. Whitman, 309 F.Supp.2d 111, 116 (D.D.C.2004). We review a grant of summary judgment de novo. Huls America Inc. v. Browner, 83 F.3d 445, 450 (D.C.Cir.1996). Because we find EPA’s decision is based on an impermissible construction of the statute, we vacate the decision of the district court and remand so that the district court can direct EPA to delist MEK.
Hs * H: * * *
MEK is a clear, colorless, low-boiling, highly volatile, and highly flammable liquid. MEK Petition Denial, 63 Fed.Reg. at 15196. A highly effective solvent, it is released in the United States in substantial quantities — with nearly 80 million pounds released into the air in 1994, and another 100,000 and 50,000 pounds into water and onto land, respectively. See id. at 15198. Although MEK can cause “chronic developmental toxicity at moderately high to high doses,” id. at 15198/1, it has “low acute and chronic (systemic) toxicity in that effects occur only at high doses,” id. at 15199/1. MEK was on the initial TRI list of 309 chemicals and 20 chemical categories specified by Congress, see 42 U.S.C. § 11023(c); 52 Fed.Reg. 3479 (Feb. 4,1987), which had consolidated two pre-existing state lists of hazardous chemicals, see 59 Fed.Reg. 1788, 1788 (Jan. 12,1994).
The Emergency Protection and Community Righ1>-To-Know Act was intended to provide “communities with information on potential chemical hazards within their boundaries” and to facilitate awareness and planning for accidental releases. See Huls America, Inc., 83 F.3d at 446 (citing H.R.Rep. No. 253, 99th Cong., 2d Sess., pt. 1 at 60). The Act establishes state emergency response commissions and local emergency planning committees, 42 U.S.C. § 11001, and requires certain facilities that manufacture, process, or use chemicals on the TRI to provide an estimate of the amount of the chemical present at the facility and the annual quantity of the chemical entering the environment. See 42 U.S.C. § 11023(g)(1); see also 40 C.F.R. § 372.25 (reports required for listed chemical if a facility uses at least 10,000 pounds or manufactures or possesses at least 25,000 pounds annually). Such facilities report this information to EPA, which then makes the information available to the public. 42 U.S.C. § 11023(h), (j).
The Act provides for listing if
(A) The chemical is known to cause or can reasonably be anticipated to cause significant adverse acute human health effects at concentration levels that are reasonably likely to exist beyond facility site boundaries as a result of continuous, or frequently recurring, releases.
(B) The chemical is known to cause or can reasonably be anticipated to cause in humans—
(i) cancer or teratogenic effects, or
(ii) serious or irreversible—
(I) reproductive dysfunctions,
*404(II) neurological disorders,
(III) heritable genetic mutations, or
(IV) other chronic health effects.
(C) The chemical is known to cause or can reasonably be anticipated to cause, because of—
(i) its toxicity,
(ii) its toxicity and persistence in the environment, or
(iii) its toxicity and tendency to bioaccumulate in the environment,
a significant adverse effect on the environment of sufficient seriousness, in the judgment of the Administrator, to warrant reporting under this section.
§ 313(d)(2), 42 U.S.C. § 11023(d)(2).
Any person may petition the Administrator to delete a chemical from the list, see 42 U.S.C. § 11023(e)(1), and the Administrator may delete a chemical if there isn’t sufficient evidence to establish any of the statutory criteria, 42 U.S.C. § 11023(d)(3). Although § 11023(e)(1) mentions only subsections (A) and (B) of § 11023(d)(2), the parties apparently agree that it allows petitions to delist a chemicál that fails to satisfy subsection (C), and we assume that to be the case for present purposes.
In its petition to EPA, the Council argued that MEK doesn’t satisfy any of the three listing criteria, and EPA agreed as to the “acute human health effects” criterion of § 313(d)(2)(A). But EPA found that MEK did meet the “chronic health effects” requirement of § 313(d)(2)(B), and the “significant adverse effect on the environment” requirement of § 313(d)(2)(C). In doing so EPA rested entirely on the proposition that MEK, as a volatile organic chemical (“VOC”), “contribute[s] to the formation of tropospheric ozone[,] which is known to cause significant adverse effects to human health and the environment.” See MEK Petition Denial, 63 Fed.Reg. at 15199/2; see also Appellees’ Br. at 9 (“EPA’s decision was expressly based on MEK’s role as a precursor to ozone”).
There is no dispute here either that ozone itself is a toxic chemical under the criteria of § 313(d)(2), or that MEK to some degree contributes to the creation of ozone. See 60 Fed.Reg. 31643 (June 16, 1995). That leaves the issue before us— whether this contribution to the creation of a concededly toxic chemical is adequate to support listing on the TRI.
4* 4» 4* 4' 4* 4*
The Council’s principal contention is that the TRI is only intended to include “toxic” chemicals, meaning chemicals that cause harm through exposure to the chemical. See Appellant’s Br. at 22-23, 29-30. EPA responds that, properly interpreted, § 313(d)(2) provides for listing chemicals that result in harm not by exposure to the listed chemical, but by exposure to some downstream chemical that the listed chemical in some way fosters. In the context of § 313(d)(2)(B), EPA speaks of “indirect causation” and “indirect effects.” For § 313(d)(2)(C), EPA speaks of “indirect toxicity.” But these are all linguistic variants on the same theme, rejecting the idea that the Act is confined to toxic chemicals in the ordinary sense of the term.
With respect to § 313(d)(2)(B), the chronic health effects criterion, EPA first suggests that the statute does not require that the listed chemical be toxic at all. Rather, EPA insists the section permits the listing of any chemical that is “known to cause or can reasonably be anticipated to cause” serious chronic health effects, as EPA interprets that phrase. EPA views the phrase “cause or can reasonably be anticipated to cause” as ambiguous, and interprets it to incorporate “indirect effects.” It claims deference for this interpretation under Chevron U.S.A., Inc. v. Natural Resources Defense Council, Inc., 467 U.S. 837, 104 S.Ct. 2778, 81 L.Ed.2d *405694 (1984). Before reaching the question of what notion of causation is permissible, however, we must address whether there is a prerequisite that a chemical be toxic, and if so, what that term means.
The agency attempts to make something of the fact that the term “toxicity” appears explicitly in § 313(d)(2)(C), but not in § 313(d)(2)(B). See Appellees’ Br. at 20-21. It is true of course, that when Congress “includes particular language in one section of a statute but omits it in another section of the same Act, it is generally presumed that Congress acts intentionally and purposely in the separate inclusion or exclusion.” Barnhart v. Sigmon Coal Co., 534 U.S. 438, 452, 122 S.Ct. 941, 151 L.Ed.2d 908 (2002) (quoting Russello v. United States, 464 U.S. 16, 23, 104 S.Ct. 296, 78 L.Ed.2d 17 (1983)); Troy Corp. v. Browner, 120 F.3d 277, 286 (D.C.Cir.1997). But that principle has little purchase when, as here, the overall scope of a statute is clearly limited by a requirement that is not explicitly mentioned in every subsection. The basic question is whether the overall text, structure, and purpose of the statute allow EPA to list non-toxic chemicals. See Nat’l Rifle Ass’n of America v. Reno, 216 F.3d 122, 127 (D.C.Cir.2000). The answer is that they do not.
The term toxicity or toxic chemical is used 38 times in the statute, and while the frequency of usage says nothing about the term’s content, it does strongly suggest that the Toxics Release Inventory is intended to include chemicals and chemical compounds that are, in fact, toxic. Moreover, while EPA sometimes suggests that the only requirement for listing under § 313(d)(2)(B) is that the chemical “cause or can reasonably be anticipated to cause” the stated harm, see MEK Petition Denial, 63 Fed.Reg. at 15199/3, there is at least some question whether EPA actually subscribes to this view. Even in this case EPA observed that “Congress intended that the TRI would contain only toxic chemicals.” See EPA’s Summary Judgment Memorandum at 20 (internal quotation marks omitted).
The overall structure of § 313(d)(2) conveys a similar impression. Each of the subsections seems to address a different form of toxicity. Section 313(d)(2)(A) corresponds roughly to the notion of acute toxicity, and § 313(d)(2)(B) to chronic toxicity. Congress was apparently focused on precisely this distinction. Subsection (B) originated in an amendment offered by its proponent explicitly to assure reporting for chemicals “which are not only acutely toxic, as [then] provided for in the bill, but also those chemicals that pose serious chronic health risks.” See 5 Legislative History of the Superfund Amendments and Reauthorization Act of 1986 at 4193 (1990) (emphasis added) (explaining purpose); see also id. at 4353-54 (vote adopting amendment). As the proponent, Rep. Edgar, said, “[i]f we are monitoring emissions of chemicals that can cause death or sickness in hours, why not do the same for those which may lead to injury over longer exposure periods? ” Id. at 4194 (emphasis added). Although the ultimately enacted language more closely resembles the Senate version of the legislation, the conference report evinces the same focus on toxicity in its acute and its chronic form. See H.R. Conf. Rep. No. 99-962, 1986 U.S.C.C.A.N. at 3276, 3387 (Oct. 3, 1986) (“Subsection (d) of the conference agreement requires reporting on listed toxic chemicals that cause, or reasonably can be anticipated to cause, significant adverse acute human health effects, various chronic human health effects, and significant adverse effects on the environment.”) (emphasis added). Section 313(d)(2)(C) extends coverage to chemicals that escape classification under (A) or (B) but nonetheless, through their toxicity (alone or in combination with persistence in the envi*406ronment or tendency to bioaccumulate), have severe environmental effects. Each subsection obviously captures a different type of resulting harm. Read together, they plainly focus on toxic chemicals.
EPA argues that even if § 313(d)(2)(B) contains an implicit requirement of toxicity, that term should be understood differently from it's use in ordinary parlance. See Appellees’ Br. at 22. This argument takes two forms. To start with, EPA urges that the term “toxic” should not be given its ordinary meaning because “toxic chemical” is defined in the statute’s formal definitions to mean “a substance on the list described in section 313(c)” — that is, a substance on the TRI, as revised by the Administrator. See § 329(10), 42 U.S.C. § 11049(10); see also Appellees’ Br. at 22. There is no doubt that “[w]hen a statute includes an explicit definition, we must follow that definition, even if it varies from that term’s ordinary meaning.” Stenberg v. Carhart, 530 U.S. 914, 942, 120 S.Ct. 2597, 147 L.Ed.2d 743 (2000). But this argument is simply a repetition of the argument that § 313(d)(2)’s “cause or can reasonably be anticipated to cause” formula dispenses with any requirement of toxicity.
More substantively, EPA argues that even if there is a requirement of toxicity, “EPA would not have been required to-accept the Council’s proposed definition; for ... there is no single, universally accepted definition of toxicity.” Appellees’ Br. at 25 (internal quotation marks omitted). EPA suggests that toxicity should be defined as “the potential to cause harm to biological or living systems.” Id. at 33. The Council argues that a chemical is toxic if “it causes illness or injury when ingested, inhaled, or otherwise absorbed into the body.” Appellant’s Br. at 29. Virtually all' definitions other than EPA’s litigating position closely approximate the Council’s idea, and contain either an implicit or explicit requirement of harm resulting from exposure. Compare Hawley’s Condensed Chemical Dictionary 1117 (13th ed.1997) (“ability of a substance to cause damage to living tissue, impairment of the central nervous system, severe illness or, in extreme cases, death when ingested, inhaled, or absorbed by the skin”) (emphasis added), with Webster’s Neio World Dictionary (3d college ed., 4th prtg.1988) 1415, 1043 (defining toxic as “of, affected, by, or caused by a toxin, or poison” and defining poison as “a substance causing illness or death when eaten, drunk, or absorbed even in relatively small quantities”) (emphasis added). EPA acknowledges that there “is arguably support for the ACC’s proposed definition.” Appellees’ Br. at 32. We think the case is a lot stronger than that. After all, EPA’s own Toxicology Handbook (3d prtg.1988) defines “toxicant” as a “harmful substance or agent that may injure an exposed organism.” J.A. 154 (emphasis added). Even in its decision here, EPA characterized as “fundamental” the question “whether, regardless of the number of intervening steps, there is a natural and continuous line, unbroken by any intervening causes, between exposure to the chemical and the toxic effect.” MEK Petition Denial, 63 Fed.Reg. at 15199/3 (emphasis added).
EPA’s currently preferred definition of toxicity would apparently qualify materials that no scientist or educated lay person would term toxic. For example, the release of large volumes of any liquid, be it juice, milk, or gasoline, has “the potential to cause harm to biological or living systems” — for example by causing death by drowning or crop destruction by flooding— but it would be nonsense to suggest that all liquids are “toxic chemicals” for purposes of listing on the TRI. And while EPA expressly disavows its intention to list all VOCs, MEK Petition Denial, 63 Fed.Reg. at 15199, the agency readily admits that under the view it advances here *407it would be free to list any and all VOCs because of their potential contribution to the formation of ozone, id. In fact, at oral argument, counsel for EPA conceded that EPA’s interpretation would allow water to be listed on the TRI. Oral Arg. at approx. 30:00-25. Congress does not generally hide elephants in mouseholes, Whitman v. American Trucking Associations, 531 U.S. 457, 468, 121 S.Ct. 903, 149 L.Ed.2d 1 (2001), and we think it utterly improbable .that by creating a list of several hundred toxic chemicals, with authority for add-ons, Congress intended to allow EPA to list all VOCs, orange juice, and water. EPA’s preferred definition is clearly inconsistent with the statute.
To fit MEK under § 313(d)(2)(C), EPA offers additional arguments. See MEK Petition Denial, 63 Fed.Reg. at 15199. One is simply a replay of EPA’s strenuous effort to remove any limiting notion of toxicity from the statute, with MEK’s indirect contribution to ozone’s harms supposedly amounting to “toxicity.” EPA speaks of “indirect toxicity” in the context of § 313(d)(2)(C) and of “indirect effects” or “indirect causation” in the context of § 313(d)(2)(B). But this is merely an effort to fit the same square peg into different circular holes. EPA’s other variant depends on eliminating “its” from § 313(d)(2)(C) (“because of its [MEK’s] toxicity”), so that MEK becomes a toxic chemical because of the toxicity of ozone, to whose creation MEK contributes. This ruthless surgery is equally unjustified.
Although EPA argues that the statute should be liberally construed to effect the purpose of the statute, its own proposed removal of virtually any constraints on the discretion of the Administrator would hardly serve that purpose. Congress intended the Act to facilitate community information, awareness, and planning for the release of hazardous chemicals from nearby facilities. See Huls America, Inc., 83 F.3d at 446. But the sort of community response seemingly anticipated — precautions against relatively local toxic releases — is hardly advanced by including a chemical that, when mixed with other chemicals thousands of feet above the point of release, tends to generate a third chemical, which in turn may result in adverse effects on humans and the environment in regions hundreds of miles from the initial time and place of the release. See EPA, Air Quality Criteria and Related Photochemical Oxidants ch. 3, 1-A, 36-37 (July 1996) (describing role of VOCs in creation of ozone).
In sum, we hold that §§ 313(d)(2)(B) and (C) allow only for the listing of toxic chemicals, substantially as the term toxic is used in ordinary parlance. At a minimum, the chemical must cause harm via exposure. Because EPA’s own analysis demonstrates that MEK fails this test, EPA’s denial of the Council’s petition to delist was improper. As a result, we vacate the decision of the district court and remand so that it can direct EPA to delete MEK from the TRI.
So ordered.