Commonwealth of Massachusetts v. Blackstone Valley Electric Company

LYNCH, Circuit Judge. The Commonwealth of Massachusetts seeks to recover response costs under CERCLA and Mass.Gen.L. ch. 21E from Blackstone Valley Electric Co. (“BVE”) for the removal of ferric ferroeyanide (“FFC”) from a waste site in North Attleboro, Massachusetts. The Commonwealth’s ability to recover its response costs, said to be $5.8 million, turns largely on the question of whether FFC is a “hazardous substance” within the meaning of CERCLA. The broader concern raised by this case is identifying who should decide that question and by what process. We hold that neither CERCLA nor the existing EPA regulations clearly establish whether FFC is a hazardous substance, and that the district court erred in trying to resolve the question on the Commonwealth’s motion for summary judgment, in the face of warring expert affidavits, where there is no textual plain meaning to resolve the issue. Invoking the doctrine of primary jurisdiction, we hold that the EPA should, as Congress intended, address the question in the first instance. Accordingly, we vacate the grant of partial summary judgment and order referral to the EPA for an administrative determination. In so doing we reject the EPA’s argument as amicus curiae in this court that it has effectively answered the question of whether FFC is a CERCLA “hazardous substance” by adopting standard testing protocols for effluent discharge regulations promulgated under the Clean Water Act. I. Factual Background Like many other environmental cases, the story of this case starts in the last century. Before the construction of the natural gas pipeline system, gas for consumer use in heating, lighting, and cooking was often manufactured from coal at localized facilities. According to one 1985 study commissioned by the EPA, there were some 1500 such manufactured gas plants in operation throughout this country between 1889 and 1950. The cleanup of the waste byproducts of the manufacturing process, which often were buried on site or deposited in landfills, has been a source of modem environmental litigation. See, e.g., John Boyd Co. v. Boston Gas Co., 1992 WL 212231, *1 (D.Mass. Aug. 18, 1992), aff'd, 992 F.2d 401, 403-04 (1st Cir.1993); Interstate Power Co. v. Kansas City Power & Light Co., 992 F.2d 804, 805-06 (8th Cir.1993). An important step in the gas manufacturing process was the purification of the gas obtained from the coal. One typical purification method involved pumping the untreated gas through “purifier boxes” containing wood chips coated with iron oxide. As the untreated gas passed through the boxes, it reacted chemically with the coated wood chips, causing unwanted substances to be filtered out. The byproducts of the purifying chemical reactions would build up on the wood chips. One such byproduct was a blue substance called ferric ferroeyanide (more commonly, Prussian Blue). Eventually, the spent wood chips — still bearing the byproducts of the chemical purification process — would typically be incinerated or buried. In the early 1980’s, blue-colored wood chips and soil were discovered in a landfill near a residential area in North Attleboro, Massachusetts. Between July 1984 and May 1986, the Massachusetts Department of Environmental Protection (“DEP”) excavated the site to remove the blue soil and wood chips. The blue substance on the wood chips was identified as ferric ferroeyanide, and the wood chips were identified as “purifier box waste” from the coal gasification process. The DEP determined that the FFC-eoated chips at the site had been transported there from a gas facility that had been operated by BVE’s direct corporate predecessor in Rhode Island from 1920-1961. The Commonwealth subsequently sued BVE as a generator of the *984 FFC to recover its cleanup costs, pursuant to the relevant provisions of CERCLA and the analogous Massachusetts state statutes. 1 . II. The Statutory Framework A. The “Hazardous Substance” List CERCLA provides state and federal governmental authorities with broad power to clean up waste sites, and then to seek recovery of response costs from responsible parties. 42 U.S.C. §§ 9604, 9607; see generally Dedham, Water Co. v. Cumberland Farms Dairy, Inc., 889 F.2d 1146, 1150 (1st Cir.1989). One of the predicates to CERCLA liability is the release or threatened release of a “hazardous substance” at the site. A “hazardous substance” is defined in CERCLA, 42 U.S.C. § 9601(14), by incorporation of certain lists of substances, wastes, and pollutants identified in a number of other environmental statutes, including the Clean Water Act (“CWA”), 33 U.S.C. § 1251 et seq. 2 CERCLA requires the Administrator of the EPA to promulgate and revise regulations designating as additional “hazardous substances” any substances which, “when released into the environment may present substantial danger to the public health or welfare or the environment....” 42 U.S.C. § 9602(a). The EPA has codified a consolidated list of hazardous substances subsuming all of the statutory lists incorporated by CERCLA, at 40 C.F.R. § 302.4, Table 302.4 (“Table 302.4”). 3 The substance FFC is not specifically named in any of the statutory lists of substances incorporated by CERCLA and hence does not appear in Table 302.4. The EPA has never taken official action pursuant to its authority under 42 U.S.C. § 9602(a) specifically to add FFC to the CERCLA hazardous substance list. Table 302.4 does list, however, a broad category of compounds — “cyanides” — which, in turn, the Commonwealth claims, does encompass FFC. B. “Cyanides ” The category “cyanides” in Table 302.4 has its origins in the CWA. The EPA was required to promulgate, within a short period following the CWA’s enactment, a list of “any toxic pollutant or combination of such pollutants” to be subject to regulation under the statute. See CWA, Pub.L. No. 92-500, § 307(a)(1), 86 Stat. 816, 856, 1972 U.S.C.C.A.N. 951, 1000. Pursuant to this directive, an ad hoc EPA work group developed a proposed list of 65 toxic pollutants. After public notice and comment, this list (the “CWA list”) was adopted by Congress, see 33 U.S.C. § 1317(a), published by the EPA, see 43 Fed.Reg. 4108-09 (Jan. 31, 1978), and codified, see 40 C.F.R. § 401.15. In addition to identifying various specific, discrete chemical compounds (e.g., “benzene,” “2,4-dichlorophenol”), the CWA list also identifies several groups of compounds associated with particular elements (e.g., “arsenic and compounds,” “zinc and "compounds”), and classes of more generally denominated compounds (e.g., “nitrosamines,” “chlorinated ethanes”). One of the latter such classes of compounds on the list is “cyanides.” The dispute in this case has centered on whether the term “cyanides” in the CWA list (and incorporated into Table 302.4) includes FFC, thereby bringing FFC within the scope of CERCLA’s definition of “hazardous substance.” III. Proceedings in the District Court After discovery, the Commonwealth moved for partial summary judgment as to liability against BVE, claiming that, as a matter of law, FFC is a “hazardous substance” within the meaning of CERCLA. The Commonwealth argued FFC falls within the “plain meaning” of the term “cyanides” in Table *985 302.4. 4 After a hearing on the Commonwealth’s motion, the district court directed the parties to “focus only on the meaning of the term [‘cyanides’] as it is understood in the general scientific community.” Commonwealth of Mass. v. BVE, Civ. No. 87-1799-T, Memorandum at 5 (D.Mass. May 23, 1990). Accordingly, BVE filed expert affidavits attesting that the plain meaning of “cyanides” does not include the substance FFC, and the Commonwealth filed expert affidavits attesting that it does. Additionally, the Commonwealth attempted to solicit the EPA’s involvement in the case. Before filing its motion, the Commonwealth had asked the EPA to participate in the case as amicus curiae, but the EPA had refused. After the summary judgment hearing, the Commonwealth asked the EPA to provide an affidavit stating that the EPA’s own definition of “cyanides” encompasses FFC. The EPA again declined the Commonwealth’s invitation. Instead, the EPA wrote a letter to the Massachusetts Attorney General’s office, signed by Stephen D. Luftig, the Director of EPA’s Emergency Response Division (the “Luftig Letter”). The letter purported to describe the EPA’s administrative view of the status of FFC vis-a-vis the CERCLA/CWA category of “cyanides.” The Commonwealth provided this letter to the district court as additional support for its motion. The district court granted the Commonwealth’s motion for partial summary judgment. Commonwealth of Massachusetts v. Blackstone Valley Electric Co., 777 F.Supp. 1036 (D.Mass.1991). The district court made no mention of the Luftig Letter in its decision. It relied instead upon two sentences of text concerning chemical testing procedures for cyanides contained in a reference publication called “Standard Methods.” See American Public Health Ass’n et al., Standard Methods for the Examination of Water and Wastewater (18th ed. 1992). One of the Commonwealth’s experts had averred that Standard Methods is a “ ‘universally accepted environmental chemistry lab testing manual in the general scientific community.’ ” 777 F.Supp. at 1038 n. 3 (quoting expert affidavit). Based on its reading of that publication, 5 the district court concluded that FFC was properly classified as a “complex cyanide,” that “[t]he plain meaning of cyanides includes complexes such as ferric ferrocyan-ide,” and that FFC was therefore a hazardous substance within the meaning of CERC-HA. Id. at 1039. The district court rejected BVE’s arguments and expert affidavits supporting a contrary result and added: “Blackstone’s argument, essentially, is that FFC should not be on the list. This is a contention that Blackstone should present to the EPA, not to this court.” Id. 6 We disagree with the district court’s conclusions about the “plain meaning” of “cya-nides.” We have considerable sympathy, however, for its sentiment that BVE’s arguments about the status of FFC are best suited for presentation to the EPA. IV. Discussion We review the district court’s summary judgment order de novo. See Vasapolli v. Rostoff, 39 F.3d 27, 32 (1st Cir.1994). Our *986 review of the district court’s interpretation of the relevant statutory framework also is plenary. See Estey v. Commissioner, Maine Dep’t of Human Services, 21 F.3d 1198, 1201 (1st Cir.1994). In assessing whether the Commonwealth is entitled to judgment as a matter of law, we must regard the record and draw all inferences in a manner favorable to BVE. Only if, viewed in that light, the record discloses no genuine issue of material fact will we uphold the grant of summary judgment. FDIC v. Bay Street Dev. Co., 32 F.3d 636, 639 (1st Cir.1994). A. Absence of Plain Meaning The Commonwealth and BVE both argued to the district court that the term “cyanides” has a “plain meaning,” but were sharply at odds as to whether that meaning encompasses FFC. The district court endorsed the plain meaning approach, see 777 F.Supp. at 1038, and agreed with the Commonwealth that the plain meaning of “cyanides” includes FFC. Id. at 1039. We find that the district court’s reliance on the plain meaning approach was misplaced on the facts here. Of course when the words of a statutory provision are clear, the provision’s plain meaning must govern its application, unless a palpably unreasonable outcome would result. See, e.g., Hogan v. Bangor & Aroostook Railroad Co., 61 F.3d 1034, 1037 (1st Cir.1995); Pritzker v. Yari, 42 F.3d 53, 67-68 (1st Cir.1994). Yet, as the qualifications that are a part of the plain meaning rule suggest, that rule does not provide a panacea for every problem of statutory construction. Words can be ambiguous, often materially so. See Greenwood Trust Co. v. Commonwealth of Mass., 971 F.2d 818, 825 (1st Cir.1992) (“[T]he plain-meaning doctrine is not a pedagogical absolute.”), cert. denied, — U.S. -, 113 S.Ct. 974, 122 L.Ed.2d 129 (1993). When ambiguity is identified, a dispute about a statute’s or regulation’s proper construction cannot be resolved simply by placing the gloss of “plain meaning” on one competing interpretation. See, e.g., In re Jarvis, 53 F.3d 416, 419 (1st Cir.1995) (finding plain meaning inquiry inapposite where relevant language was indeterminate); United States v. O’Neil, 11 F.3d 292, 294-96 (1st Cir.1993) (same, where term “revoke” was ambiguous in relevant context); Isaac v. Harvard Univ., 769 F.2d 817, 820 (1st Cir.1985) (same, where terms “proceedings” and “terminated” were ambiguous as used); cf. Allen v. Adage, Inc., 967 F.2d 695, 700 (1st Cir.1992) (finding term “reduction-in-force” to be ambiguous and therefore “unplain” in context of an ERISA plan). Here, both BVE and the Commonwealth argue that the plain meaning of “cyanides” can be ascertained by consulting “the scientific community.” To this end, they have filled the record with competing expert affidavits setting forth contradictory views (each ostensibly authoritative) as to whether FFC is a member of the category “cyanides.” But no “plain” meaning of the term “cyanides” can be identified from among these conflicting expert affidavits. It is true that, as a general rule of construction, when a statute contains “technical words or terms of art, ‘it [is] proper to explain them by reference to the art or science to which they [are] appropriate.’ ” Corning Glass Works v. Brennan, 417 U.S. 188, 201, 94 S.Ct. 2223, 2231, 41 L.Ed.2d 1 (1974) (bracketed alterations in original) (quoting Greenleaf v. Goodrich, 101 U.S. (11 Otto) 278, 284, 25 L.Ed. 845 (1880)). Assuming that the “scientific community” is the appropriate body by reference to which the meaning of “cyanides” should be determined, the basic indeterminacy nonetheless remains. The “scientific community” is not a monolithic entity that has spoken here in a single authoritative voice. As one of BVE’s experts stated, members of different disciplines within the scientific community at large are apt to take sharply contrasting approaches and to give conflicting answers to the question whether FFC can properly be classified as one of the “cyanides.” Thus, the Commonwealth’s key expert, an analytical chemist, states confidently that “[t]here can be no dispute ... that cyanides and all other chemical substances are defined based on chemical reactivity [emphasis added],” and thus understands “cyanides” to include “all those chemical compounds containing the negatively charged cyanide ion, CN-” and *987 that “can yield the free cyanide ion” in laboratory conditions. Then, stating that “there is no doubt that the CN moiety in iron cyanide complexes is uni-negative” and that FFC does release the free cyanide ion when boiled in concentrated sulfuric acid, he concludes that FFC is properly categorized as one of the “cyanides.” On the other hand, BVE’s expert (who was one of the consultants to the EPA who helped devise the CWA list) asserts just as confidently that most scientists other than analytical chemists “would define ‘cyanides’ as substances that are toxic due to the CN group.” Stating that FFC is not toxic and does not release “toxi-eologically significant doses of [free] cyanide under environmental conditions [emphasis added],” BVE’s expert concludes that FFC is not properly classified as one of the “cya-nides” within the meaning of CERCLA. The term “cyanides” as it appears in Table 302.4 is, we believe, ambiguous in the context of this ease. The term suffers from an ambiguity that might be classified as a “categorical indeterminacy.” See Clark D. Cunningham et al., Plain Meaning and Hard Cases, 103 Yale L.J. 1561, 1585 (1994) (reviewing Lawrence M. Solan, The Language of Judges (1993)). At least on the record before us, the category “cyanides” does not admit of crisply defined boundaries, and resolution of the disagreement about whether FFC falls within those fuzzy boundaries requires a value-laden choice from among competing interpretive assumptions, a choice that cannot be made through mere inspection of the term’s normal or ordinary usage. Mindful that we must view the record in the light most favorable to BYE, this indeterminacy cannot be resolved by designating the Commonwealth’s rendition of the meaning of “cyanides” as “plain.” From the viewpoint of a federal court presented with facially credible expert affidavits that directly contradict each other on the issue, the question whether “cyanides” in Table 302.4 encompasses FFC for purposes of CERCLA liability cannot be answered as a matter of plain meaning. 7 B. Legislative and Regulatory History Having found considerable ambiguity in the word “cyanides,” we turn to whether the history of the CWA provides a clearer understanding. The legislative history of the statute contains no express congressional guidance as to the scope of the term. The regulatory history of the CWA toxic pollutant list, however, does provide substantial reason for skepticism about the Commonwealth’s and the EPA’s claim that “cyanides” encompasses FFC. The list of substances and classes of substances currently codified at 40 C.F.R. § 401.15 (and incorporated into Table 302.4) was developed by the EPA pursuant to Congress’ directive to produce a list of toxic pollutants to be subject to regulation under the CWA. See CWA, Pub.L. No. 92-500, § 307(a)(1), 86 Stat. 816, 1972 U.S.C.C.A.N. 951, 1000. Congress defined “toxic pollutants” as those “pollutants, or combination of pollutants” that were believed to “cause death, disease, behavioral abnormalities, cancer, genetic mutations, physiological malfunctions (including malfunctions in reproduction) or physical deformations” in organisms or their offspring. 33 U.S.C. § 1362(13). Congress expressly instructed the EPA in devising the list to “take into account the toxicity of the pollutant, its persistence, degradability, the usual or potential presence of the affected organisms in any waters, the importance of the affected organisms and the nature and extent of the effect of the toxic pollutant on such organisms.” CWA, supra, § 307(a)(1), 86 Stat. at 856, 1972 U.S.C.C.A.N. at 1000. After public notice and an initial period for public comment, the EPA published a proposed list of toxic pollutants for regulation under the CWA. 38 Fed.Reg. 24342 (Sep. 7, 1973). This original proposed list did not include the category “cyanides.” Instead, it listed “cyanide and all cyanide compounds.” 38 Fed.Reg. at 24344. The EPA commentary accompanying the proposed list noted that “[c]yanide is on the list because of its high order of toxicity to aquatic life.” Id. *988 The text then acknowledged that some commentators had “objected to inclusion of ‘all cyanide compounds’” and had “argued that only compounds which dissociate in water to form toxic concentrations of cyanide ion or hydrogen cyanide should be included.” Id. Thus, there was some public concern that the listing of “all cyanide compounds” was over-inclusive, and that only a subset of “cyanide compounds” — compounds which, in environmental conditions would produce toxic results — should be included on the CWA list. The EPA’s commentary stated that the “proposed effluent standards will take these comments into account_” Id. (emphasis added). On the final CWA list of toxic pollutants (as adopted by Congress), the category “cyanide and all cyanide compounds” was replaced with the category “cyanides.” See 40 C.F.R. § 401.15. Viewed in the light most favorable to BVE, the evidence indicates that FFC is highly stable, insoluble in water, and completely non-toxic to human and aquatic life. Against the backdrop of (1) the definition of “toxic pollutant” contained in the CWA; (2) the congressional directive that required the EPA to “take into account the toxicity” of pollutants in producing the CWA list; (3) the EPA’s own comment that “cyanide is on the list because of its high order of toxicity”; and (4) the objections that appear to have precipitated the change from “cyanide and all cyanide compounds” to “cyanides”, there would seem substantial reason to doubt that FFC, claimed to be a non-toxic substance, could properly be deemed to fall within the category “cyanides.” The history of the CWA list tends to support BVE’s claim that the category “cyanides” was never contemplated to encompass substances such as FFC for purposes of the CWA or CERCLA. This leaves the question of whether the EPA has adopted elsewhere an official agency interpretation that clearly includes or excludes FFC as a CERCLA hazardous substance. We find that no such agency interpretation has been established. C. Absence of A Regulatory Definition In its amicus brief, the EPA distances itself from the parties’ and district court’s “plain meaning” approach. It argues that the initial decision whether FFC is a “hazardous substance” is one that must be left to the EPA. We agree. It further argues, however, that although no definition of “cyanides” can be found in the regulations identifying CERCLA hazardous substances, it is spelled out elsewhere in the applicable regulatory framework, and that it encompasses FFC. On this score, we are not persuaded, for a number of reasons. The regulatory text itself does not support the argument; the argument leads to results which are overbroad and defy common sense; the EPA has itself taken inconsistent positions; the position is articulated solely and for the first time in a litigation posture; and policy reasons dictate against the approach proposed by the EPA. 1. Absence of Rules Specifically Concerning FFC The EPA has clearly not acted pursuant to its. authority under CERCLA, 42 U.S.C. § 9602(a), nor under the CWA, 33 U.S.C. § 1317(a)(1), to promulgate a rule specifically listing FFC as a “hazardous substance” (CERCLA) or a “toxic pollutant” (CWA). Neither the Commonwealth nor amicus contends otherwise. The EPA has also never issued a rule specifically for the purpose of defining the scope of the term “cyanides.” The EPA has in the past resorted to its rulemaking authority to provide clear guidance to the public as to the scope of at least six other substances or classes of substances listed as CWA toxic pollutants, see 40 C.F.R. § 129.4, but it never has done so with respect to the term “cyanides.” Indeed, the Luftig Letter submitted by the Commonwealth to the district court acknowledges that “[t]he term ‘cyanides’ is not specifically defined in the CWA or, as far as we can determine, in the legislative history....” 2. The “Total Cyanide” Test The EPA’s central focus in its amicus brief is on certain regulations establishing a testing protocol for the analysis of cyanide in effluent discharges under the CWA. The EPA contends that this test procedure provides the legal definition of “cyanides,” and *989 that FFC falls within this definition. We conclude that this contention is not supported by the relevant regulatory framework, and that the testing protocol referred to cannot properly be interpreted to provide the definition of “cyanides” under CERCLA. Independent of its authority to designate hazardous substances and toxic pollutants under CERCLA and the CWA, the EPA also has authority and responsibility under the CWA to “promulgate guidelines establishing test procedures for the analysis of pollutants that shall include the factors which must be provided in any [CWA compliance certification or permit application].” 33 U.S.C. § 1314(h). Pursuant to this delegation, the EPA has issued regulations incorporating test procedures for measuring the level of certain “parameters” in a water or waste sample. See 40 C.F.R. § 136.3. One of the listed parameters is “cyanide-total, mg/L.” Id. The regulation indicates that the procedures to be used in measuring “cyanide-total” in effluent discharges are those described in the Standard Methods reference publication. Amicus places overriding significance upon one of these procedures, called the “total cyanide” test. See Standard Methods, supra, at 4-20, 4-23. This procedure involves boiling the sample to be tested in concentrated sulfuric acid. Compounds that contain the CN group in their chemical composition, it is said, will release detectable amounts of free cyanide when subjected to the procedure. Amicus claims that, under the EPA’s regulations, any substance that releases cyanide upon being boiled in sulfuric acid under the “total cyanide” test qualifies as one of the “cyanides” for purposes of CERCLA liability. It further claims that because FFC releases some cyanide when subjected to the “total cyanide” test, it is necessarily one of the “cyanides” within the meaning of CERCLA. We do not think that this conclusion follows. The EPA’s own regulations do not use the test procedures identified at 40 C.F.R. § 136.3 (including the Standard Methods protocol) to define the scope of CWA- or CERCLA-designated categories of toxic pollutants or hazardous substances. The regulations never state that any substance that releases cyanide under the “total cyanide” test qualifies as one of the “cyanides” for purposes of CERCLA Rather, the regulations themselves say something quite different. The regulation that specifically governs the applicability of the identified test procedures states that the procedures are intended to “perform the measurements ” required in connection with (a) Clean Water Act permit applications, (b) discharge reports, and (c) certain compliance certifications issued by states. 40 C.F.R. § 136.1 (emphasis added). None of these three expressly designated uses for the test procedures is applicable here, and none has anything to do with providing a definition of any class of pollutants. The regulations intend the “total cyanide” procedure to serve only the purpose of measuring the total CN 8 by weight in the chemical composition of a given waste sample, not to define which chemical substances count as a member of the category “cyanides.” Thus, even if FFC can be measured for “total cyanide” composition under the Standard Methods procedure, 9 it surely does not follow as a matter of law or logic that FFC is one of the “cyanides” for purposes of CERCLA liability. A further difficulty with amicus’ attempt to define “cyanides” by reference to the total cyanide test is that such a definition may lead to nonsensical results. One of BVE’s experts observes that there are many everyday substances that contain the CN group in their chemical composition {e.g., vitamin B-12, the synthetic fiber Orion, and a number of common medicines such as Lomotil), and some or all of these substances, like FFC, would also release cyanide when subjected to *990 the conditions of the total cyanide test. 10 Yet no one, including the EPA, would categorize vitamin B-12, for example, as one of the “cyanides” within the meaning of the CWA or CERCLA. BVE’s experts have averred that FFC’s chemical structure and composition are much more similar to substances like vitamin B-12 than to toxic substances like potassium cyanide. Assuming BVE is correct, as we must here, a rule that defined “cyanides” to include all substances that release any cyanide when subjected to the total cyanide test would appear to be untenably overinclusive. 11 The unsettled nature of the status of FFC vis-a-vis the category “cyanides” is further demonstrated by at least one documented situation in which the EPA has appeared to take official action at odds with the position articulated in its amicus brief. This situation, discussed in some detail in the Luftig Letter submitted to the district court, involved the EPA’s handling in 1985 of ferro-cyanide wastes generated at a facility operated by the Mearl corporation. Mearl had filed a petition before the EPA to exclude its wastewater treatment sludge from regulation under RCRA. See 50 Fed.Reg. 7882, 7888-90 (Feb. 26, 1985). Although a test for total cyanide indicated positive results, Mearl argued to the EPA that all cyanide in the waste was “in the insoluble, non-toxic form of ferric ferrocyanide.” 50 Fed.Reg. at 7889. After public hearing and comment, the EPA granted Mearl’s petition to exclude the waste from RCRA regulation, stating that “the waste does not exhibit any of the characteristics of hazardous waste.” 50 Fed.Reg. 48886, 48890 (Nov. 27, 1985). The EPA further stated that “the cyanide present [in the sludge, in the form of FFC] will not convert to free cyanide [in environmental conditions] and therefore is not of regulatory concern with respect to ground-water or atmospheric exposure routes.” 50 Fed.Reg. at 48890 (emphasis added). While technically, the decision to exclude Mearl’s wastewater sludge from RCRA regulation was limited to Mearl’s own facility, and did not directly affect the status of FFC under the CWA or CERCLA, it is difficult to ignore the EPA’s statement that the FFC in the Mearl sludge was “not of regulatory concern” because it would not convert to free cyanide under environmental conditions. Here, too, the record supports the conclusion that the FFC found at the Attleboro site may pose no threat of releasing free cyanide under normal environmental conditions. 12 At a minimum, the EPA’s action with respect to the Mearl petition provides some support for BVE’s position in this litigation. We conclude that the EPA rules promulgated under the CWA that identify test procedures for the measurement of wastewater parameters, including the total cyanide test, do not set forth an agency definition of “cya-nides” for purposes of the CWA’s list of toxic pollutants or CERCLA’s list of hazardous substances. Thus, even assuming that FFC *991 releases cyanide when subjected to the Standard Methods test for measuring “total cyanide,” it does not follow as a matter of law that FFC is one of the “cyanides” for purposes of CERCLA liability. 3. Agency Deference The varying positions stated in the EPA’s amicus brief and in the Luftig Letter concerning the EPA’s purported definition of “cyanides” are not entitled to deference under the principles of 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). It is apparent that the argument for defining “cyanides” by reference to the total cyanide test has been tailored to and articulated specifically for purposes of this particular litigation. 13 As such, that position need not be given any special weight. See Martin v. Occupational Safety & Health Rev. Comm’n, 499 U.S. 144, 156-57, 111 S.Ct. 1171, 1178-79, 113 L.Ed.2d 117 (1991) (agency’s litigating position, in the nature of “post hoc rationalization” rather than the result of the official exercise of rulemaking authority, is entitled to no Chevron deference); see also Director, Office of Workers’ Compensation Programs, U.S. Dep’t of Labor v. General Dynamics Corp., 980 F.2d 74, 79 (1st Cir.1992); Brewster v. Sullivan, 972 F.2d 898, 901 (8th Cir.1992). 4. Policy Considerations We are also troubled by the EPA’s approach here as a matter of policy. A complicated regulatory regime like CERCLA or the CWA cannot function effectively unless citizens are given fair notice of their obligations. Congress delegated to the EPA the continuing task of defining which substances are “hazardous substances” to which CERCLA liability can attach. The EPA does not argue here that the term “cyanides” has a plain meaning that would enable a person to answer the question of whether FFC falls within that category. Instead, it suggests that the patchwork of regulations relating to the measurement of effluent discharges can be adapted to the task at hand. We can thus determine the status of FFC for CERCLA liability purposes, says the EPA, by boiling the FFC in concentrated sulfuric acid. Yet the EPA points to no regulation or other source — except its amicus submission to this court — that tells the public that boiling a substance in concentrated sulfuric acid is the way to determine whether it legally qualifies as one of the “cyanides.” That is not fair notice to the public and is not what Congress contemplated when it granted the EPA power to promulgate regulations to define and supplement CERCLA’s list of hazardous substances. D. Primary Jurisdiction Because there exists no basis for concluding as a matter of law that FFC falls within the scope of the term “cyanides,” the district court’s order granting partial summary judgment in favor of the Commonwealth must be vacated. We are left, then, to decide whether the proper disposition of this appeal is to remand the case to the district court for trial, or to prescribe some other avenue for appropriate factfinding with respect to “cyanides” and FFC. We conclude that the proper course is a referral to the EPA under the doctrine of primary jurisdiction. Having found that the term “cyanides” is ambiguous, that EPA’s regulatory framework does not adequately define the term, that the legislative and regulatory history of the term “cyanides” does not establish the Commonwealth’s position, and that the position advocated by amicus is not entitled to *992 deference, we are left with virtually no legislative or administrative guidance for determining whether, on the record before us, FFC is one of the “cyanides.” Congress delegated to the EPA, not to the courts, the authority to administer the CWA toxic pollutant list and the CERCLA list of hazardous substances. This case seems clearly to call for referral to the EPA under the “primary jurisdiction” doctrine, for an appropriate administrative determination of whether FFC falls within the category “cyanides.” Cf. Chastain v. AT & T Co., 351 F.Supp. 1320, 1323 (D.D.C.1972) (invoking primary jurisdiction doctrine and referring ease to the relevant agency, where the court was “unwilling and unable to assume the initial responsibility of evaluating the highly technical questions raised by the parties”). The Supreme Court has stated that “[n]o fixed formula exists for applying the doctrine of primary jurisdiction.” United States v. Western Pacific Railroad Co., 352 U.S. 59, 64, 77 S.Ct. 161, 165, 1 L.Ed.2d 126 (1956). Broadly speaking, the doctrine, informed by principles of deference to agency decisionmaking, gives effect to the eminently sensible notion that “in eases raising issues of fact not within the conventional experience of judges or cases requiring the exercise of administrative discretion, agencies created by Congress for regulating the subject matter should not be passed over.” Id. (quoting Far East Conference v. United States, 342 U.S. 570, 574-75, 72 S.Ct. 492, 494-95, 96 L.Ed. 576 (1952)); see generally II Kenneth C. Davis & Richard J. Pierce, Jr., Administrative Law Treatise § 14.1, at 271-80 (3d ed. 1994). The doctrine is intended to “serve[ ] as a means of coordinating administrative and judicial machinery,” and to “promote uniformity and take advantage of agencies’ special expertise.” Mashpee Tribe v. New Seabury Corp., 592 F.2d 575, 580 (1st Cir.1979). This court has said that there are three factors that guide the decision whether or not to defer a matter to an agency under the primary jurisdiction doctrine: (1) whether the agency determination l[ies] at the heart of the task assigned the agency by Congress; (2) whether agency expertise [i]s required to unravel intricate, technical facts; and (3) whether, though perhaps not determinative, the agency determination would materially aid the court. Id. at 580-81 (citing Chicago Mercantile Exchange v. Deaktor, 414 U.S. 113, 114-15, 94 S.Ct. 466, 466-67, 38 L.Ed.2d 344 (1973)). All three of these factors are plainly satisfied here. The determination whether FFC is a hazardous substance is specifically within the scope of the EPA’s delegated authority; the EPA’s expertise is required to sift through and properly weigh all of the arguments for and against including FFC within the category “cyanides”; and official action by the EPA on this issue would indisputably assist the court in determining BVE’s liability to the Commonwealth under CERCLA. 14 The judicial machinery is ill-suited to fashioning a workable rule for determining whether the substance FFC by virtue of its chemical, structural, functional, or other qualities, falls within the properly conceived definition of “cyanides.” That determination is much better left to the EPA. Referral to the EPA under the doctrine of primary jurisdiction will also serve the interest of national uniformity in regulation. The question of whether FFC is a CERCLA hazardous substance is of more than local concern. As noted earlier, FFC is a common byproduct of the gas manufacturing process that was prevalent in prior decades at some 1500 different facilities across the country. Moreover, a determination as to whether FFC is one of the “cyanides” would undoubtedly have significant implications — beyond our purview — for similar substances whose status under CERCLA currently remains unclear. Rather than leave this matter to *993 the risk of inconsistent outcomes before particular courts in different parts of the country, we believe it better to have the EPA resolve the issue nationwide. 15 Accordingly, we conclude that this ease should be referred to the EPA for an administrative determination of whether FFC is one of the “cyanides” within the meaning of 40 C.F.R. § 401.15 and Table 302.4. We further conclude that the district court proceedings in this case shall be stayed and that the court shall retain jurisdiction over this case pending an appropriate determination of the relevant issues by the EPA. 16 See Reiter v. Cooper, — U.S. -, -, 113 S.Ct. 1213, 1220, 122 L.Ed.2d 604 (1993) (explaining that court has discretion to retain jurisdiction pending administrative referral or to dismiss the case without prejudice). The district court’s order granting the motion for partial summary judgment is vacated. The case is remanded to the district court for primary jurisdiction reference to the EPA The district court shall refer the matter to the EPA to determine whether FFC qualifies as one of the “cyanides” within the meaning of 40 C.F.R. § 401.15 and 40 C.F.R. § 302.4, Table 302.4. No costs are awarded. Footnotes [1] The only issue presented here is the CERCLA one. [2] CERCLA’s definition of "hazardous substance" also incorporates the pollutants listed in the Solid Waste Disposal Act, as amended by the Resource Conservation and Control Act, 42 U.S.C. § 6921 et seq., the Clean Air Act, 42 U.S.C. § 7401 et seq., and the Toxic Substances Control Act, 15 U.S.C. § 2601 et seq. The parties agree that only the CWA list is pertinent here. [3] The Massachusetts analogue to CERCLA defines “hazardous material” to include all “hazardous substances” under CERCLA. See Mass. Gen.L. ch. 21E, § 2. [4] To avoid confusion, we observe that neither party attaches controlling significance to the fact that the common name of the substance at issue — ferric ferrocyanide — contains the word "cyanide.” In fact, according to modem chemical nomenclature conventions, the proper name for FFC is “iron(III) hexacyanoferrate(II).” The appearance of the word “cyanide” within the name "ferric ferrocyanide” does not factor into the interpretation. [5] The district court focused on the following paragraph: Cyanide refers to all of the CN groups in cyanide compounds that can be determined as the cyanide ion, CN-, by the methods used. The cyanide compounds in which cyanide can be obtained as CN- are classed as simple and complex cyanides. 777 F.Supp. at 1038 (quoting Standard Methods, supra, at 4-18). [6] Later, based in part on its grant of partial summary judgment against BVE on the FFC issue, the court entered summary judgment in favor of the Commonwealth on the issue of BVE's liability as a generator under CERCLA, 42 U.S.C. § 9607(a)(3). See Commonwealth of Mass. v. Blackstone Valley Electric Co., 808 F.Supp. 912, 914-16 (D.Mass.1992). BVE has not appealed from the latter order. [7] We also note that the EPA has not argued in its amicus brief that the plain meaning of cyanides in Table 302.4 includes FFC. [8] "CN” is the chemical formula of the cyanide molecule. [9] BVE’s expert asserts that because of the properties of FFC, the total cyanide test cannot actually give an accurate quantification of the total CN composition by weight in a given sample of FFC. The Commonwealth's expert appears to agree on this point, but states that because FFC does nevertheless yield some cyanide when subjected to the procedure, FFC is one of the “cya-nides.” [10] The Commonwealth's expert appears to deny that vitamin B-12 would release cyanide in a total cyanide test. For purposes of deciding the Commonwealth's summary judgment motion, we credit BVE’s position on this factual issue, as we must. The EPA in its amicus brief does not attempt to dispute BVE's factual assertion. [11] Indeed, the problem of overbreadth is what appears to have prompted commentators to object to the EPA's original inclusion of "cyanide and all cyanide compounds” on the proposed CWA list, and what prompted the change to "cyanides.” See 38 Fed.Reg. at 24344. Yet, using the "total cyanide" test to define "cya-nides" as amicus proposes would, in effect, make the category “cyanides” equivalent to the rejected formulation, "all cyanide compounds.” [12] The Commonwealth’s experts (and its counsel at oral argument) have suggested that FFC could degrade and release cyanide gas when exposed to sunlight. BVE's expert has attested to a directly contrary conclusion. Additionally, the Commonwealth's counsel asserted at oral argument that some free cyanide was found at the Attleboro site, suggesting that the cyanide had dissociated from the FFC under environmental conditions. However, as far as the record discloses, only trace amounts of free cyanide were found at the site, i.e., measuring less than 1 part per million. By comparison, the generally recognized safety threshold for free cyanide in workroom air is 10 parts per million. For purposes of evaluating the Commonwealth’s summary judgment motion, we must assume that FFC does not degrade when exposed to sunlight, and that no more than background levels of free cyanide were detectable at the Attleboro waste site. [13] The Luftig Letter, while written by an EPA official, does not set forth an entrenched EPA view. The letter does not articulate a definition as such of the term "cyanides” and is tellingly circumspect in its discussion of the EPA's purported position on whether FFC falls within that category. Instead of stating outright that the EPA adheres to an established definition of "cya-nides" that encompasses FFC, it makes only the far weaker statement that "the manner in which EPA addresses cyanides under the Clean Water Act indicates that the term does include ferric ferrocyanide.” Furthermore, while the letter says that the EPA uses the total cyanide test described in Standard Methods, it never states that the EPA has defined "cyanides" by reference to that test, offering the more limited assertion that "[t]he[se] testing procedures provide confirmation that ferric ferrocyanide is a 'cyanide' [emphasis added].” [14] We acknowledge the general principle that a primary jurisdiction reference to an agency is usually inappropriate in an enforcement action brought by the agency. See ICC v. B & T Transp. Co., 613 F.2d 1182, 1187 (1st Cir.1980) (stating that the primary jurisdiction doctrine does not apply where the agency brings suit, because the agency's position on the matter to be litigated will be clear). This action, however, was brought by the Commonwealth, not by the EPA. As noted above, the EPA's position on the definition of "cyanides” is far from clear. [15] The EPA’s determination would, of course, be subject to judicial review and thus would not be immune from challenge if arbitrary, unreasonable, clearly contrary to the statute’s intended effect, or otherwise unlawful. See ABF Freight Sys., Inc. v. NLRB, - U.S. -, -, 114 S.Ct. 835, 839, 127 L.Ed.2d 152 (1994); Brown v. Secretary of HHS, 46 F.3d 102, 106 (1st Cir.1995). [16] In so doing, we note that BVE has placed the $5.8 million at stake here in an interest bearing escrow account. The Commonwealth’s interests will be protected during the stay. When we asked the Commonwealth at oral argument if any additional protections would be required should the EPA’s primary jurisdiction be invoked, the Commonwealth sought nothing further.