Full opinion text
CARDAMONE, Circuit Judge: This appeal, arising originally from the disposal of hazardous substances at two Connecticut landfills, Beacon Heights and Laurel Park, requires us to determine whether defendants accused of generating and transporting hazardous substances deposited at the two landfill sites might be liable under the Comprehensive Environmental Response, Compensation, and Liability Act (CERCLA) to contribute to the costs of cleaning up the sites. Under CERCLA, several classes of responsible parties are liable for most costs incurred in responding to and remediating sites where hazardous substances are found. We must, in addition, decide if certain parties remain susceptible to suit by the United States and the State of Connecticut for response costs incurred by these governments. Plaintiffs-appellants the United States, the State of Connecticut, the Beacon Heights Coalition, and the Laurel Park Coalition appeal from a judgment entered May 2, 1995 by the United States District Court for the District of Connecticut (Dorsey, C.J.). The court granted judgment on the pleadings against the United States and Connecticut and granted summary judgment against the Beacon Heights Coalition and the Laurel Park Coalition and in favor of nearly 100 defendants alleged in plaintiffs’ complaints to be potentially responsible parties in an action under CERCLA, 42 U.S.C. §§ 9601-9675, as amended by the Superfund Amendments and Reauthorization Act of 1986, Pub.L. No. 99-499, 100 Stat. 1613 (1986). The Beacon Heights and Laurel Park Coalitions are groups of industrial waste generators that sought contribution from third parties after having settled their liability with the United States, the State of Connecticut, and the Murthas and affiliated entities — the owners/operators of the two landfill sites. The Laurel Park Coalition sought initially to add 1151 other potentially responsible parties to the litigation. The district court ground away at this number slowly, yet it “ground exceeding small.” Insisting that these plaintiffs only implead those parties against whom they had a claim that was both legally and factually substantiated, the district court reduced to 41 the 1151 third party defendants that these plaintiffs moved to add, thus eliminating over 1000 potential parties from the suit. Later, with motions for summary judgment before it, the district court construed some of CERCLA’s basic provisions in a manner inconsistent with our precedents and, ruling that the coalitions had failed to advance sufficient proof, thereby was able to grant summary judgment to nearly all of the roughly 100 defendants, dismissing plaintiffs’ complaints against them. The district court also found that both governments had been fully reimbursed by the industrial coalitions and the landfills’ owners and therefore granted summary judgment against these governments. BACKGROUND We assume the reader’s familiarity with our previous decision in this case, B.F. Goodrich Co. v. Murtha, 958 F.2d 1192 (2d Cir.1992) (Murtha I), and with the published decisions of the district court relevant to this appeal. See B.F. Goodrich Co. v. Murtha, 815 F.Supp. 539 (D.Conn.1993) (Murtha II); B.F. Goodrich Co. v. Murtha, 840 F.Supp. 180 (D.Conn.1993) (Murtha III); B.F. Goodrich Co. v. Murtha, 855 F.Supp. 545 (D.Conn.1994) (Murtha IV). We detail only those background facts necessary to an understanding of this appeal. Those facts relevant to the discussion of a specific issue and necessary to its disposition will be included in the analysis. A. Initial Litigation The Beacon Heights and Laurel Park landfills have both been designated as Superfund sites by the Environmental Protection Agency (EPA). Terrance and Harold Murtha, and several corporations controlled by them (collectively, Murtha), owned and operated Laurel Park from 1961 until 1987, and owned and operated Beacon Heights from 1970 to 1987. In 1987 Murtha was sued in separate actions by the EPA, the State of Connecticut’s Department of Environmental Protection (DEP), Uniroyal Chemical Company, Inc., and a coalition of corporations led by B.F. Goodrich Company. Murtha filed third party actions for contribution or indemnification against some 200 third party defendants. These parties were accused either of generating or of transporting hazardous substances to the landfills. Most of the defendants named in the Murtha third party suits were not named by the EPA as potentially responsible parties. Murtha I, 958 F.2d at 1196. In September 1987, when the Murtha litigation began, the EPA entered into a consent decree concerning the remediation of Beacon Heights with 33 potentially responsible parties who may have generated hazardous substances that were later deposited in that landfill. These parties, later known as the Beacon Heights Coalition, agreed to undertake the remediation of that site and to reimburse the government for its future oversight costs in excess of $500,000. Since this coalition did not provide for past costs or for those future response costs unrelated to oversight, the EPA sued several non-settling potentially responsible parties in a separate action to recover its past and future remediation costs. The EPA also filed two lawsuits concerning the costs of remediating the Laurel Park landfill. One led to a settlement with another industrial coalition of various potentially responsible parties, now known as the Laurel Park Coalition. In an August 1992 consent decree, the Laurel Park Coalition agreed to perform remediation at the site, pay for the government’s future oversight costs in excess of $200,000, and pay the EPA $500,000 for its past response costs. The EPA then sued, as it did in the Beacon Heights ease, the non-settling potentially responsible parties to recover past costs and future- response costs not addressed by the consent decree. The State of Connecticut, which is also involved in the Laurel Park cleanup, received over $1 million from the coalition for response costs and sued the same non-settling parties for unreimbursed costs. Murtha also settled its litigation with the United States, with Connecticut, and with the two coalitions (coalitions, plaintiffs, or appellants). The parties embodied their agreement in a consent decree (Murtha consent decree) which also was entered by the district court in August 1992. Murtha resolved its liability at both landfill sites for $5,375,-000; this amount was apportioned among several parties. The Beacon Heights Coalition received $1,875,000 and $322,500 was given to the EPA for unreimbursed costs relating to Beacon Heights. Murtha assigned its third party claims to the Beacon Heights Coalition. The EPA also received $625,000 for unreimbursed costs at Laurel Park, as did the State of Connecticut, with the remaining $1,975,000 set aside for remediation at that site. Although the Laurel Park Coalition had not been formed when the Murtha consent decree was negotiated, the parties agreed that should an industrial coalition form within 18 months of the entry of Murtha’s- consent decree, the coalition would receive the funds. Because this condition was met, the Laurel Park Coalition received the $1.975 million set aside by Murtha. See Murtha III, 840 F.Supp. at 183. Many of the third party defendants were municipalities and municipal corporations that generated and collected municipal solid waste. They claimed that CERCLA exempted municipal waste from its coverage and that an exemption for household waste in the Resource Conservation and Recovery Act (RCRA), 42 U.S.C. §§ 6901-6992k, was incorporated into CERCLA. The district court rejected these arguments, B.F. Goodrich Co. v. Murtha, 754 F.Supp. 960 (D.Conn.1991), and we affirmed, Murtha I, 958 F.2d at 1197. The municipal generators therefore remained in the case as named defendants and potentially responsible parties. B. Summary Judgment Decisions The Beacon Heights Coalition was already armed with Murtha’s assignment of its third party claims against generators and transporters of hazardous substances. The Laurel Park Coalition then moved to add to the action as third party defendants 1151 potentially responsible parties. The district court granted the motion to the extent of allowing 41 parties to be joined, but denied it as to the 1110 other potential parties. Murtha II, 815 F.Supp. at 546-61. In April 1992, approximately six weeks after we rejected the municipal defendants’ claim of CERCLA exemption, the district court decided sua sponte to permit parties to file motions for summary judgment. Nearly all of the original third party defendants filed such motions. Both coalitions opposed the motions and in turn cross-moved for the same relief. Murtha III, 840 F.Supp. at 183. With dispositive motions before the trial court, the two coalitions became increasingly concerned about their ability to present evidence in support of their contentions that the third party defendants had generated or transported hazardous substances to the landfills. Throughout the proceedings in the trial court, discovery had been limited (independently and through a special master). Uniform discovery orders and case management orders effectively restricted discovery to a set of court-drafted interrogatories. The special master assured the coalitions on several occasions that additional discovery would be permitted. After initially denying the coalitions’ request for expanded discovery, the district court in December 1992 allowed some limited deposition discovery. Summary judgment motions were submitted to the court and decided in a series of orders, beginning in December 1993. See Murtha III, 840 F.Supp. at 188-191. Summary judgment was eventually granted to every third party defendant-appellee that had moved for such relief, id., and the district court subsequently dismissed sua sponte nearly all the remaining third party defendants, including many that did not move for that relief. Of the 88 defendants sued by the coalitions who had not entered into settlement agreements, all but four were dismissed. Judgment on the pleadings was entered against the United States and the State of Connecticut in their suits against non-settling potentially responsible parties. Murtha IV, 855 F.Supp. at 549. The district court reasoned that the EPA and DEP had presented no figures for incurred response costs that were not covered by the $3,500,000 net Murtha settlement, id. at 546, and therefore dismissed both governments’ claims for additional reimbursement. It also granted summary judgment in favor of the remaining three of the eight defendants at Beacon Heights sued by the EPA. From an amended judgment entered May 2, 1995, the United States, the State of Connecticut, and both coalitions appeal. We affirm in part, reverse in part, and remand for further proceedings. DISCUSSION The coalitions point to a number of reasons calling for reversal. They contend the district court granted summary judgment to many of the defendants based on a misapplication of substantive CERCLA law, and that instead their cross-motions for summary judgment should have been granted. They also maintain that procedural errors were committed that require reversal. The trial court’s initial decision, Murtha II, 815 F.Supp. 539, laid down several broad legal rules, which the court applied, occasionally by reference, in deciding subsequent cases. See Murtha III, 840 F.Supp. at 183 (“Many comments on the evidence in [Murtha II ] are pertinent to the present motions and will not be repeated here.”). We address first the coalitions’ assertion that the district court incorrectly interpreted CERC-LA; second, we consider the coalitions’ procedural contentions; third, we review de novo the grants of summary judgment in 85 individual cases (with reference to an appendix incorporated in tins opinion); and, finally, we address the appeals of the United States and the State of Connecticut. I The Law of CERCLA A. Basic Principles Our construction of CERCLA does not begin with this case. Rather, we are guided by previous writings on this subject in this and other Circuits. CERCLA is a “broad remedial statute,” Murtha I, 958 F.2d at 1197, and was enacted with the purpose of “[fjirst, assuring that those responsible for any damage, environmental harm, or injury from chemical poisons bear the costs of their actions.” S.Rep. 848, 96th Cong., 2d Sess. 13 (1980) (Senate Report), reprinted in 1 Senate Comm, on Env’t and Pub.Works, Legislative History of the Comprehensive Environmental Response, Compensation and Liability Act of 1980 (Superfund), at 305, 320 (1983) (Leg.Hist.). As a remedial statute, CERC-LA should be construed liberally to give effect to its purposes. Schiavone v. Pearce, 79 F.3d 248, 253 (2d Cir.1996). These include facilitating efficient responses to environmental harm, holding responsible parties liable for the costs of the cleanup, see Murtha I, 958 F.2d at 1198, and encouraging settlements that reduce the inefficient expenditure of public funds on lengthy litigation, see In Re Cuyahoga Equip. Corp., 980 F.2d 110, 119 (2d Cir.1992). CERCLA makes four classes of persons liable—(1) present owners and operators of facilities that accepted hazardous substances, (2)past owners and operators of such facilities, (3) generators of hazardous substances, and (4) certain transporters of hazardous substances. 42 U.S.C. § 9607(a). As we have noted, “[t]he Act’s broad reach extends liability to all those contributing—from generation through disposal to—the problems caused by hazardous substances.” Murtha 1, 958 F.2d at 1198. Responsible parties are hable for a broad range of remediation expenses, including ah costs of removal of the substances not inconsistent with the national contingency plan, other necessary response costs, damages for injury to natural resources, and the cost of health assessments. See 42 U.S.C. § 9607(a)(4). The Act imposes strict liability. Strict liability is intended to make sure that those who “benefit financially from a commercial activity” internalize the environmental costs of the activity as a cost of doing business. Senate Report at 13, Leg.Hist. at 320; see New York v. Shore Realty Corp., 759 F.2d 1032, 1042 (2d Cir.1985) (discussing legislative history). Liability under the Act is joint and several, unless potentially responsible parties can prove that the harm is divisible. See United States v. Alcan Aluminum Corp., 990 F.2d 711, 721-22 (2d Cir.1993); Murtha I, 958 F.2d at 1198. A CERCLA plaintiff establishes a prima facie case by proving that (1) the defendant is within one of the four categories of responsible parties enumerated in § 9607(a); (2) the landfill site is a facility as defined in § 9601(9); (3) there is a release or threatened release of hazardous substances at the facility; (4) the plaintiff incurred costs responding to the release or threatened release; and (5) the costs and response actions conform to the national contingency plan. 42 U.S.C. § 9607(a); Alcan, 990 F.2d at 719-20; Murtha I, 958 F.2d at 1198. Once a plaintiff makes a prima facie showing, a defendant may avoid liability only if it establishes by a preponderance of the evidence that the release or threatened release was caused by an act of God, an act of war, certain acts or omissions of third parties other than those with whom the defendant has a contractual relationship, or a combination of these reasons. 42 U.S.C. § 9607(b). If the plaintiff can establish each of the pri-ma facie elements on undisputed facts, and the defendant is unable to demonstrate by a preponderance of the evidence that an affirmative defense applies, summary judgment on liability is appropriate. Alcan, 990 F.2d at 720. Significantly, it is “not required that the [plaintiff] show that a specific defendant’s waste caused incurrence of clean-up costs.” Id. at 721. Because CERCLA imposes strict liability, there is no causation requirement. “Hazardous substance” is expansively defined to include (A) any substance designated pursuant to section 1321(b)(2)(A) of Title 33, (B) any element, compound, mixture, solution, or substance designated pursuant to section 9602 of this title, (C) any hazardous waste having the characteristics identified under or listed pursuant to section 3001 of. the Solid Waste Disposal Act...., (D) any toxic pollutant listed under section 1317(a) of Title 33, (E) any hazardous air pollutant listed under section 112 of the Clean Air Act ..., and (F) any imminently hazardous chemical substance or mixture with respect to which the [EPA] Administrator has taken action pursuant to section 2606 of Title 15. 42 U.S.C. § 9601(14). Pursuant to subsection (B), the EPA has listed more than 700 hazardous substances in Table 302.4 of 40 C.F.R. § 302.4 (1995). We have noted that “[t]he breadth of § .9601(14) cannot be easily escaped, and we have expressly held that ‘[q]uantity or concentration is not a factor.’ ” Alcan, 990 F.2d at 720 (quoting Murtha I, 958 F.2d at 1200). In other words, the Act’s definition includes “even minimal amounts of pollution.” Alcan, 990 F.2d at 720. These propositions presently govern CERCLA litigation in this Circuit. If a district court grants summary judgment based on an incorrect understanding of CERCLA, we must reverse and remand. See Keywell Corp. v. Weinstein, 33 F.3d 159, 163 (2d Cir.1994) (reviewing decision granting summary judgment includes determining whether district court correctly applied the law). B. Analysis of District Court’s CERCLA Interpretation The coalitions aver that many third party defendants-appellees were granted summary judgment based on an erroneous application of CERCLA and the case law interpreting the statute. We address each claim of error. 1. Component Parts The district court held, in Murtha II, 815 F.Supp. at 546, that “[a]bsent a finding by EPA that a particular product warrants classification as a [hazardous substance] ..., it cannot be found that the product, notwithstanding its constituent elements, is an [hazardous substance] or [hazardous waste].” It later clarified that holding, by explaining that “mere presence of an element, a named [hazardous substance], as a constituent of a product does not render the product a [hazardous substance].” Murtha III, 840 F.Supp. at 184. In Murtha I we rejected the contention that a specific mixture or waste solution must be listed within § 9601(14) or incorporated by reference in order to fall within CERC-LA’s coverage. We explained that “[w]hen a mixture or waste solution contains hazardous substances, that mixture is itself hazardous for purposes of determining CERCLA liability. Liability under CERCLA depends only on the presence in any form of listed hazardous substances.” Murtha I, 958 F.2d at 1201; see also United States v. Alcan Aluminum Corp., 964 F.2d 252, 259-61 (3d Cir.1992) (rejecting quantitative requirement); Amoco Oil Co. v. Borden, Inc., 889 F.2d 664, 669 (5th Cir.1989) (same). It is enough that a mixture or waste solution contain a hazardous substance for that mixture to be deemed hazardous under CERCLA. The waste product itself “need not be listed by name — instead of its constituent components — to fall within the Act.” Murtha I, 958 F.2d at 1201; see also Louisiana-Pacific Corp. v. ASARCO Inc., 24 F.3d 1565, 1573 (9th Cir.1994) (holding that even if product not specifically listed as hazardous substance, if components include hazardous substances, product is regulated by CERCLA), cert. denied, — U.S. -, 115 S.Ct. 780, 130 L.Ed.2d 674 (1995); Eagle-Picher Indus. v. EPA, 759 F.2d 922, 930-31 (D.C.Cir.1985) (rejecting contention that because most mining wastes and fly ash contain some hazardous substances, mining wastes and fly ash must therefore be specifically listed as hazardous substances). Our understanding is also consistent with Congress’ plan. Legislative history reveals that “the release of any [listed hazardous or toxic substance] or any constituent of them invokes the notice requirements and response provisions and any costs of removal or remedial action or any damages are subject to the liability provisions of the bill.” Senate Report at 24-27, Leg.Hist. at 331 (emphasis added). In short, it makes no difference that the specific wastes disposed of by the appellees were not themselves listed as hazardous substances, because so long as their component parts were listed as hazardous substances there may be CERCLA liability. Appellees declare this approach will lead to CERCLA liability if a discarded object had any EPA listed hazardous substance in its chemical genealogy, whether or not the chemical component’s characteristics had been unalterably changed in the manufacturing process. Even if this objection is sound in theory, it is not relevant. The instant case involves not only discarded objects like pens or plastic containers that might include a hazardous substance in their chemical makeup. Appellees are accused of dumping waste that contained hazardous substances in separable, identifiable forms. According to an affidavit submitted by the coalitions, the cumulative waste stream discarded in the Mur-tha’s landfills included municipal sludge, incinerator ash, varnishes, paints, tar remover, caulking compounds, pesticides, glues, de-greasers, and automotive lubricants. Obviously, disposal of this sort of waste poses a greater threat of releasing hazardous substances than is presented by — to use an example offered by the appellees at oral argument — the disposal of a single plastic pen. In an attempt to avoid this conclusion, appellees invoke Massachusetts v. Blackstone Valley Elec. Co., 67 F.3d 981, 992-93 (1st Cir.1995). Blackstone held that the EPA should make an initial determination of whether ferric ferrocyanide (FFC) — in this case a blue-colored substance on wood chips — is a cyanide and therefore a hazardous substance. The First Circuit ruled that even if FFC contained cyanide as a chemical component, the cyanide might not exist in a form convertible to free cyanide and therefore might not be a CERCLA hazardous substance. Id. at 990-91. However, in this case, the First Circuit focused on whether FFC was a hazardous substance. It did not suggest that were FFC found to be a hazardous substance, liability might be avoided because “wood chips” — the product allegedly discarded by the polluter — were not in themselves hazardous substances within the Act’s definition. In short, the appellees’ reliance on Blackstone is misplaced. Thus, those dismissals in the present case that were predicated on the notion that a waste product is not covered by CERCLA unless specifically listed as a hazardous substance — even when its component parts are hazardous substances — misconstrued CERC-LA and must be reversed. 2. Releasability The trial court ruled further that “[t]he absence of evidence of breakdown of products that contain [hazardous substances] ... precludes a finding that disposal of that product constitutes a disposal of [hazardous substances].” Murtha II, 815 F.Supp. at 545. Later, it explained that statement by saying that “[liability does not lie if the material disposed of would release a [hazardous substance] only on the intervention of another force.”. Murtha III, 840 F.Supp. at 188. The coalitions believe that these pronouncements were wrong. As noted, to make out a prima facie CERCLA case, a plaintiff must show a “release, or a threatened release ... of a hazardous substance.” § 9607(a)(4). Accordingly, if a hazardous substance is only used in a non-releasable form in the manufacturing of a product, it may scientifically be impossible for the plaintiff to show a “threatened release.” Hence, the district court properly required plaintiffs to show a release or threatened release. But, when it required them to show an actual breakdown of products containing hazardous substances, and determined there was no liability if the hazardous substances would only be released by an intervening force, it acted in a manner contrary to precedent. In Alcan, we held that proof that a defendant’s waste did not release listed hazardous substances is only relevant to the issue of apportionment of damages, not to the issue of liability. Alcan, 990 F.2d at 722. Independent releasability is not required to establish liability; a defendant otherwise liable may show “nonreleasability” in order to mitigate its share of damages. It follows logically that a defendant who disposes of hazardous substances that are not independently releasable may still be held liable, even though that defendant may not be required to pay damages when the cost apportionment phase of the litigation is reached. In addition, the trial court’s reading of CERCLA is inconsistent with the Act’s language, which provides only the four already recited causation-related defenses. 42 U.S.C. § 9607(b). The canon of construction that says “expressio unius est exelusio alteri-us” cautions against creating additional exceptions to complex statutory enactments. Cf. Greene v. United States, 79 F.3d 1348, 1355 (2d Cir.1996) (applying this principle in the context of federal tax law). Moreover, if we required a plaintiff to show more than a release or threatened release, we essentially would be asking the plaintiffs to prove that a specific defendant’s hazardous substances caused the release of a hazardous substance. No causation is needed, however, to establish liability under CERCLA, see Alcan, 990 F.2d at 721, because it is, as stated, a strict liability statute. See Shore Realty, 759 F.2d at 1042; see also Alcan Aluminum, 964 F.2d at 266 (requiring only that plaintiff prove that hazardous substances were deposited at a site from which there was a release or threatened release that caused response costs to be incurred); Amoco, 889 F.2d at 670 n. 8 (“[I]u cases involving multiple sources of contamination, a plaintiff need not prove a specific causal link between costs incurred and an individual generator’s waste.”). None of the lower court cases upon which the district court relied as precedent for its view regarding releasability is persuasive. See, e.g., United States v. New Castle County, 769 F.Supp. 591, 597 (D.Del.1991); United States v. Serafini, 750 F.Supp. 168, 171 (M.D.Pa.1990). Although appellants must show a release or threatened release, in order to make out a prima facie case, the additional releasability requirement of product breakdown finds no support in CERCLA. 3. Negligible Amounts The coalitions further contend that the district court impermissibly required a showing that appellees disposed of hazardous substances above a threshold amount. It granted summary judgment to an appellee where the amounts of hazardous substances it disposed of were “minuscule,” Murtha III, 840 F.Supp. at 184, or “[njominal,” Murtha II, 815 F.Supp. at 545. Twice we have said that quantity “is not a factor” when determining CERCLA liability because had Congress wanted to distinguish liability on the basis of quantity, it would have so provided. Murtha I, 958 F.2d at 1200; see also Alcan, 990 F.2d at 720 (“The statute on its face applies to ‘any’ hazardous substance, and it does not impose quantitative requirements.”). The absence of threshold quantity requirements in CERCLA. leads logically to the conclusion that the Act’s “hazardous substance” definition includes even minimal amounts. See Alcan, 990 F.2d at 720. The dismissals on these grounds constitute legal error. 4. EPA Designation The trial judge implied, in a few cases, that the EPA’s decision not to designate a party as a potentially responsible party is relevant in determining CERCLA liability. See, e.g., Murtha III, 840 F.Supp. at 184 (“[Appellee] was not designated as a PRP by the EPA.”). In Murtha I, 958 F.2d at 1205, we concluded that EPA policy “merely indicates that the EPA presently does not intend to pursue enforcement actions against” unnamed parties. However, here the district court weighed the plaintiffs’ evidence of liability “against the backdrop of the EPA Municipal Policy that a [party] would not be charged as a PRP absent evidence that its waste deposited at the site in question contained-specific [hazardous substances].” Murtha III, 840 F.Supp. at 188. EPA enforcement decisions are not helpful in deciding whether a party is properly joined as a defendant in a given action. See Murtha I, 958 F.2d at 1205 (explaining that EPA prosecutes only largest contributors or those with the means to pay, leaving those defendants to seek contribution from other potentially liable polluters). This is consistent with the general understanding that “[a]n agency generally cannot act against each technical violation of the statute it is charged with enforcing.” Heckler v. Chaney, 470 U.S. 821, 831, 105 S.Ct. 1649, 1656, 84 L.Ed.2d 714 (1985). Congress has explicitly authorized private rights of action independent of EPA enforcement decisions. For these reasons, the agency’s decision not to initiate its own enforcement action is irrelevant in determining whether a defendant might be liable under CERCLA. Appellees maintain that no decision was made in any case in exclusive reliance on the EPA policy, and that the court’s reference to that policy was merely illustrative. It is difficult to see what the EPA policy illustrated, as we have ruled that the policy is legally irrelevant when determining CERCLA liability. To the extent the district court relied upon an EPA designation in deciding whether to grant summary judgment, it committed error. 5. Successor Liability Several decisions were made regarding the scope of successor liability. In the case of one appellee, it was found that although the successor continued its predecessor’s business “using its former assets, employees and customers,” summary judgment in favor of that appellee on liability was proper. Murtha III, 840 F.Supp. at 185. In another instance, no successor liability was found even when “[a] prior sole proprietorship was conducted under a comparable name” and the predecessor “sold the business and related assets to [a successor].” Id. at 191. A third case was dismissed despite a successor’s purchase of assets and continuation of the predecessor’s business. Id. at 190. The district court did not explicitly formulate a test for deciding when successor liability exists. See Murtha III, 840 F.Supp. at 185. And, although the question has been addressed in other circuits, we have yet to decide exactly what is required for successor liability under CERCLA, see New York v. N. Storonske Cooperage Co., 174 B.R. 366, 374 (N.D.N.Y.1994) (recognizing that we have not decided this issue). We turn to this question now. We consider first the Act’s language, recognizing at the outset that the Act has no provision explicitly governing corporate successor liability. See Smith Land & Improvement Corp. v. Celotex Corp., 851 F.2d 86, 91 (3d Cir.1988) (“CERCLA failed to address many important issues, including corporate successor liability.”), cert. denied, 488 U.S. 1029, 109 S.Ct. 837, 102 L.Ed.2d 969 (1989); see also United States v. Carolina Transformer Co., 978 F.2d 832, 837 (4th Cir.1992) (explaining that CERCLA is silent on this subject); Louisiana-Pacific Corp. v. Asarco, Inc., 909 F.2d 1260, 1263 (9th Cir.1990) (“Congress has not addressed the issue of successor liability under CERCLA”). Nonetheless, most courts have read CERCLA as making successor corporations liable, in certain circumstances, for their predecessors’ acts. Carolina Transformer, 978 F.2d at 837; Anspec Co. v. Johnson Controls, Inc., 922 F.2d 1240, 1245 (6th Cir.1991); Louisiana-Pacific, 909 F.2d at 1263; Smith Land, 851 F.2d at 91-92; N. Storonske, 174 B.R. at 374; In Re Acushnet River & New Bedford Harbor Proceedings Re Alleged PCB Pollution, 712 F.Supp. 1010, 1013 (D.Mass.1989) (following Smith Land). See United States v. Mexico Feed & Seed Co., 980 F.2d 478, 487 (8th Cir.1992) (successor corporations are “persons” liable under CERCLA). In our view CERCLA imposes successor liability. The plain language of the Act imposes liability on any “person” who fits within the four categories of liable parties. 42 U.S.C. § 9607(a). “Person” is defined to include, inter alia, a “corporation, association, partnership, consortium, joint venture, [or] commercial entity.” § 9601(21). Further, federal law provides a rule of construction that when the word “company” or “association” is used, it “embrace[s] the words ‘successors and assigns of such company or association.’ ” 1 U.S.C. § 5; see Anspec, 922 F.2d at 1247 (applying § 5 in analyzing CERCLA liability). Thus, CERCLA’s statutory language, properly read, provides for successor liability. Moreover, the Act’s broad remedial purpose would be sharply curtailed if the Act did not encompass successor liability. As the Third Circuit has observed, “[t]he concerns that have led to a corporation’s common law liability ... for the torts of its predecessor are equally applicable to the assessment of responsibility for clean-up costs under CERCLA.” Smith Land, 851 F.2d at 91. In other words, absent successor liability, a predecessor could benefit from the illegal disposal of hazardous substances and later evade responsibility for remediation simply by changing the form in which it does business, thereby subverting the Act’s purpose of holding responsible parties liable for cleanup costs. See Murtha I, 958 F.2d at 1198. We will not read § 9607(a) in a manner that would defeat Congress’ overall purpose, absent an express statement from the legislature that we do so. See Shore Realty, 759 F.2d at 1045. A number of courts have recognized the importance of national uniformity and applied traditional rules of successor liability, rather than the successor liability law of a given state. See Smith Land, 851 F.2d at 92 (instructing district court to consider national uniformity in resolving successor liability issues under CERCLA); see also Mexico Feed, 980 F.2d at 487 n. 9 (upholding the application of federal law); Louisiana-Pacific, 909 F.2d at 1263 & n. 2 (deeming appropriate a federal rule that relies on traditional common law notions because of the need for national uniformity). But see Anspec, 922 F.2d at 1248 (applying successor liability law of forum state); N. Storonske, 174 B.R. at 375 (applying a mixture of state and traditional norms). The traditional common law rule states that a corporation acquiring the assets of another corporation only takes on its liabilities if any of the following apply: the successor expressly or impliedly agrees to assume them; the transaction may be viewed as a de facto merger or consolidation; the successor is a “mere continuation” of the predecessor; or the transaction is fraudulent. Carolina Transformer, 978 F.2d at 838; see also Mexico Feed, 980 F.2d at 487; Louisiana-Pacific, 909 F.2d at 1263. Both sides agree that this is the standard. The coalitions contend that when determining whether there is a “mere continuation,” we should not use the common law test (sometimes referred to as the identity test), which requires the existence of a single corporation after the transfer of assets, with an identity of stock, stockholders, and directors between the successor and predecessor corporations. See Carolina Transformer, 978 F.2d at 838 (explaining traditional rule). Rather, they urge that we employ the “continuity of enterprise” approach, also called the “substantial continuity” rule. The Supreme Court has followed this approach in the labor law context by asking if the successor maintains the same business, with the same employees doing the.same jobs, under the same supervisors, working conditions, and production processes, and produces the same products for the' same customers. Fall River Dyeing & Finishing Corp. v. NLRB, 482 U.S. 27, 43, 107 S.Ct. 2225, 2236, 96 L.Ed.2d 22 (1987); see also Mozingo v. Correct Mfg. Corp., 752 F.2d 168, 175 (5th Cir.1985) (using a slightly different version of substantial continuity test). CERCLA is entitled to a construction that advances its primary goals. Because the substantial continuity test is more consistent with the Act’s goals, it is superior to the older and more inflexible “identity” rule, see Carolina Transformer, 978 F.2d at 838, 840 (endorsing this approach); see also Mexico Feed, 980 F.2d at 487-89 (same); United States v. Western Processing Co., 751 F.Supp. 902, 905 (W.D.Wash.1990) (looking to continuing business test). But see Sylvester Bros. Dev. Co. v. Burlington Northern R.R., 772 F.Supp. 443, 447-49 (D.Minn.1990) (declining to apply the “continuing business enterprise” test). We therefore adopt the substantial continuity test as the appropriate legal test for successor liability under CERCLA. Cf. Schiavone, 79 F.3d at 255 (recognizing direct operator liability because it is compatible with statutory language and consistent with CERCLA’s remedial scheme). The district court spent little time exploring this question. It noted the coalitions’ citation to Turner v. Bituminous Casualty Co., 397 Mich. 406, 244 N.W.2d 873, 883 (1976), an early ease defining the “continuity of the enterprise” or substantial continuity approach, yet refused to impose successor liability without “convincing authority.” Murtha III, 840 F.Supp. at 185. Because an approach quite similar to that used in Turner is the one we follow in determining successor liability, we must reverse those grants of summary judgment where the district court applied a different and more restrictive law of successor liability. 6. Transporter Site-Selection Summary judgment also was granted to several alleged transporters of hazardous substances for various reasons, including those we have already discussed. An additional reason was offered for the decision concerning the transporters — they were held not liable because “they were directed to the sites or transfer stations by others.” Id. Hence, we must address the legal question of whether site-selection is a required element for transporter liability. The Act imposes liability on any person who accepts or accepted any hazardous substances for transport to disposal or treatment facilities, incineration vessels or sites selected by such person, from which there is a release, or a threatened release which causes the incurrence of response costs, of a hazardous substance 42 U.S.C. § 9607(a)(4). This awkwardly worded provision requires that we decide whether “selected by such person” modifies the entire phrase “disposal or treatment facilities, incineration vessels or sites” or only the immediately antecedent word “sites.” If the latter reading is correct, there is no basis for imposing a site-selection requirement for transporter liability where the substances were, as here, disposed of at a disposal or treatment facility. The appellants offer two principal arguments in support of this reading. First, they point out that the Act elsewhere defines a facility as “any site ... where a hazardous substance has been deposited, stored, disposed of, or placed.” 42 U.S.C. § 9601(9). They aver that reading the “selected by such person” language to apply only to sites other than those where hazardous substances have been deposited, stored, disposed of, or placed, is therefore necessary to prevent the word “sites” from being redundant. See, e.g., Bailey v. United States, — U.S. -, -, 116 S.Ct. 501, 507, 133 L.Ed.2d 472 (1995) (construing statute to avoid making words redundant). Second, they urge that we keep in mind the Act’s legislative purpose of making all who profit from pollution pay for their conduct. Recently, a sister Circuit concluded that CERCLA includes a site-selection requirement for transporters. Tippins Inc. v. USX Corp., 37 F.3d 87, 94 (3d Cir.1994). While noting the canon of statutory construction that “absent a clear intention to the contrary, a modifier’s reference is to the closest noun,” id. at 93, the court recognized that this canon need not be followed in all instances. Id.; see also Nobelman v. American Sav. Bank, 508 U.S. 324, 330-31, 113 S.Ct. 2106, 2110-11, 124 L.Ed.2d 228 (1993) (reliance on the “rule of the last antecedent” is not “compelled”). Here, following the canon and adopting the appellants’ rule would create an odd paradox: A transporter who brought hazardous substances to a disposal facility would be strictly liable, while a transporter who brought hazardous substances to some site other than a facility would not be liable unless the transporter itself selected the site. Tippins, 37 F.3d at 93-94. Thus, the coalitions’ reading, while “quite sensible as a matter of grammar,” Nobelman, 508 U.S. at 330, 113 S.Ct. at 2111, would “lead to a curious result.” Tippins, 37 F.3d at 93. The Third Circuit’s conclusion is consistent with most other courts that have considered the subject. See id. (collecting cases); see also United States v. Hardage, 985 F.2d 1427, 1435 (10th Cir.1993) (requiring site selection for transporter liability); United States v. Petersen Sand & Gravel, Inc., 806 F.Supp. 1346, 1356 (N.D.Ill.1992) (same); Alcatel Info. Sys. v. Arizona, 778 F.Supp. 1092, 1096 (D.Ariz.1991) (same); United States v. Western Processing Co., 756 F.Supp. 1416, 1420 (W.D.Wash.1991) (same). But see Prisco v. New York, 902 F.Supp. 374, 388 (S.D.N.Y.1995) (stating test for transporter liability without noting a site selection requirement); Jersey City Redevelopment Auth. v. PPG Indus., No. CIV A. 85-2014, 1987 WL 54410, at *5 (D.N.J. Sept. 3, 1987) (finding that “selected by the transporter” modifies only “sites”), aff'd without written opinion, 866 F.2d 1410, 1411 (3d Cir.1988). We are persuaded by the reasoning of Tippins and hold that only transporters who select the disposal location are liable under CERCLA. As a consequence, we agree with the district court that unless the transporters selected the facility or disposal site, they cannot be found liable under the Act. However, the meaning of the word “selected” is hardly clear. Tippins, 37 F.3d at 94, determined that a transporter is properly liable under the Act when it ultimately selects the disposal facility or when its active participation in the decision amounted to “substantial input into which facility was ultimately chosen.” This view accords with the Act’s language and purpose. First, a transporter can be said to select a facility or site when it helps the generator choose a disposal facility by recommending a facility or set of facilities. Second, a transporter who plays an active role in choosing the disposal facility is obviously more culpable and responsible for the resultant harm than a transporter who simply does the bidding of a given generator; the latter has a more attenuated connection to the harm. An “active participation” standard recognizes the often significant role played by transporters in choosing the disposal site. Hence, in cases of transporter liability, we must vacate the grants of summary judgment and remand for further consideration in light of the above discussion. II Procedural Issues We address next appellants’ procedural arguments that are common to a great many of the cases before us. A. Summary Judgment Standard We have recognized in CERCLA’s context that summary judgment is a “powerful legal tool[ ]” that can “avoid lengthy and perhaps needless litigation.” Alcan, 990 F.2d at 720. This tool is one well adapted to encouraging settlement and speeding remediation. See In Re Cuyahoga Equip. Corp., 980 F.2d at 119 (CERCLA designed to encourage settlement to reduce litigation costs). Its utility in CERCLA litigation is not a license to use it when material facts are genuinely disputed. Just as there is no heightened pleading standard in CERCLA cases, see Warwick Admin. Group v. Avon Prods., Inc., 820 F.Supp. 116, 121 (S.D.N.Y.1993); cf. Leatherman v. Tarrant County Narcotics Intelligence & Coordination Unit, 507 U.S. 163, 168, 113 S.Ct. 1160, 1163, 122 L.Ed.2d 517 (1993) (heightened pleading requirement exists only in cases involving fraud or mistake), the showing required to survive summary judgment also remains the same. We now briefly describe the standard for summary judgment under Fed.R.Civ.P. 56. A party moving for summary judgment has the burden of showing the absence of a genuine issue of material fact. Adickes v. S.H. Kress & Co., 398 U.S. 144, 157, 90 S.Ct. 1598, 1608, 26 L.Ed.2d 142 (1970). When a court is confronted with facts that permit several different conclusions, all inferences from the underlying facts must be drawn in the non-movant’s favor. Quaratino v. Tiffany & Co., 71 F.3d 58, 64 (2d Cir.1995). The trial court must bear in mind that “[cjredibility determinations, the weighing of the evidence, and the drawing of legitimate inferences from the facts are jury functions, not those of a judge.” Anderson v. Liberty Lobby, Inc., 477 U.S. 242, 255, 106 S.Ct. 2505, 2513-14, 91 L.Ed.2d 202 (1986). If the movant demonstrates an absence of material issues of fact, a limited burden of production shifts to the non-movant, which must “demonstrate more than ‘some metaphysical doubt as to the material facts, ... [and] must' come forward with “specific facts showing that there is a genuine issue for trial.” ’ ” Aslanidis v. United States Lines, Inc., 7 F.3d 1067, 1072 (2d Cir.1993) (citations and emphasis omitted). Summary judgment, then is granted only when “there is an absence of evidence to support the nonmoving party’s case,” Celotex Corp. v. Catrett, 477 U.S. 317, 325, 106 S.Ct. 2548, 2553-54, 91 L.Ed.2d 265 (1986), or, in other words, only if “no rational jury could find in favor of the nonmoving party because the evidence to support its case is so slight.” Gallo v. Prudential Residential Servs., 22 F.3d 1219, 1224 (2d Cir.1994). The trial court’s task is “carefully limited to discerning whether there are any genuine issues of material fact to be tried, not to deciding them. Its duty, in short, is confined at this point to issue-finding; it does not extend to issue-resolution.” Id. B. Sua Sponte Summary Judgment The district court granted summary judgment sua sponte in a number of cases. A sua sponte grant of summary judgment may be appropriate so long as the losing party was notified prior to its grant and given the opportunity to produce whatever proof it has in opposition. Celotex, 477 U.S. at 326, 106 S.Ct. at 2554; see also FLLI Moretti Cereali S.P.A. v. Continental Grain Co., 563 F.2d 563, 565 (2d Cir.1977) (power to grant summary judgment sua sponte “must be exercised with great caution and with care taken to give the parties -an opportunity to present materials in opposition”). To ensure the parties have notice and an opportunity to be heard, we have established a policy of adhering closely to the ten day notice period specified by Rule 56(c). See Scottish Air Int'l, Inc. v. British Caledonian Group, PLC, 945 F.2d 53, 55 (2d Cir.1991) (per curiam). Notice furthers two important interests. First, it gives the losing party an opportunity to develop the record to demonstrate that genuine material issues are in dispute or that determinative legal questions should be resolved in its favor, and thereby prevents the losing party from suffering procedural prejudice. Second, the parties’ development of a complete record facilitates de novo appellate review. See Hispanics For Fair and Equitable Reapportionment (H-FERA) v. Griffin, 958 F.2d 24, 25 (2d Cir.1992) (per cu-riam). When the non-moving party asserts that it cannot adequately respond to its opponent’s motion for summary judgment in fewer than-ten days, it should-be given the full ten days unless there is good reason to believe that the assertion is a sham. See Gutwein v. Roche Labs., 739 F.2d 93, 96 (2d Cir.1984). Here, the coalitions insist they were denied notice and an opportunity to be heard before the district court granted summary judgment sua sponte against them in a number of cases. Our review of the underlying procedural history indicates that such was the case. On December 20, 1993, the trial court ordered the coalitions to file a list of third parties that the coalitions intended to pursue. Murtha III, 840 F.Supp. at 191. Although it directed the coalitions to “be sure that [its prior] rulings .... are fully considered and accommodated,” id., this is too vague an admonition to be deemed equivalent to notice. When the coalitions asked for additional time, their request was denied and the prior order was reaffirmed. Three weeks later, on January 10, 1994, the coalitions were again ordered to review their files and “list” those third party claims each intends to pursue as valid. The district court further explained that “[a]ny third party claim not so listed by January 21, 1994, will be dismissed as not intended to be prosecuted further.” The district court acknowledged — in an order dated February 5, 1994 — that appellants eventually prepared and presented such lists but, rather than simply dismissing those claims the coalitions did not intend to pursue, the district court dismissed a great many others sua sponte. In a ease involving some 3000 filings, the coalitions were no doubt dedicating most of their effort between December 20, 1993 and February 5, 1994 to sifting through their records and compiling the list requested by the trial court. The coalitions were not asked to produce supporting papers or legal arguments to undergird their liability claims — they were only directed to submit a list within 11 days of the January 10 order. It is hardly surprising that the coalitions simply submitted the lists and did little else. A more prudent course for the district court would have been to give the coalitions ten days’ notice to show good cause why summary judgment should not be granted against' them. Such a course also would have given the district court the benefit of •the arguments and evidence of both sides and would have provided a more complete record on appeal. See Herzog & Straus v. GRT Corp., 553 F.2d 789, 792 & n. 5 (2d Cir.1977) (notice and opportunity to make arguments are vital to making an informed judgment); cf. Volvo N. Am. Corp. v. Men’s Int’l Professional Tennis Council, 857 F.2d 55, 65 (2d Cir.1988) (ruling erroneous a decision to grant motion to dismiss on ground neither raised nor briefed by parties). As the record currently stands, we have no knowledge of the precise reasons for the district court’s decision to grant summary judgment sua sponte in those cases where it took that course of action. It said only that there was “want of substantiation,” a truism that necessarily applies to all grants of summary judgment and which adds nothing to our understanding of the trial court’s decision. Because the coalitions’ claims are not demonstrably frivolous, the summary judgment dismissals, made sua sponte without ten days’ notice, were error. C. Discovery Limitations In the course of this litigation, the district court placed certain restrictions on discovery. For over one year — from March 6,1991 until April 29, 1992 — it did not allow discovery outside of that permitted by the uniform discovery orders. Showings of good cause for further discovery were not entertained until April 29, 1992, when the court solicited motions for summary judgment. Several weeks later, it promised the coalitions additional discovery for them to respond to the forthcoming summary judgment motions. The special master repeated the promise in August, but refused to authorize discovery until the motions for summary judgment were submitted and coordinated. In December, appellants were given 60 days within which to conduct limited deposition diseov-ery. They contend that these discovery limitations were so severe as to be an abuse of the trial court’s discretion and to constitute reversible error. A trial court necessarily has wide discretion in managing pre-trial discovery, see Cruden v. Bank of New York, 957 F.2d 961, 972 (2d Cir.1992), and we will not disturb its orders absent a clear abuse of its discretion, id.; Robertson v. National Basketball Ass’n, 622 F.2d 34, 35-36 (2d Cir. 1980). Reversal of a judgment because of an improper order denying or curtailing discovery is “unusual.” 8 Charles Alan Wright, Arthur R. Miller & Richard L. Marcus, Federal Practice and Procedure § 2006, at 92 (2d ed. 1994). We find no abuse of discretion in the district court’s management of discovery. In complex multi-party litigation, a trial court often needs to use its broad authority to control discovery. Some of the measures adopted here, such as time limits, schedules for discovery, and limitations on deposition discovery, have been specifically recommended as acceptable options. See Manual For Complex Litigation (Third) § 21.422 (1995); H.R.Rep. No. 253(1), 99th Cong., 2d Sess. 80 (1985), reprinted in 1986 U.S.C.C.A.N. 2835, 2862 (supporting the enactment of amendments to CERCLA and urging district courts to “carefully manage” CERCLA cases “to insure that the litigation is conducted in an expeditious manner” through, inter alia, the “exercise of strict judicial control over multi-party proceedings ... as well as use of the procedures set forth in the Manual For Complex Litigation”). The district court sought to guard against what it thought might be a wildfire of unwarranted discovery, recognizing that unduly extensive use of discovery could make the litigation expense disproportionate to the cost of cleaning up the landfills. In order to have orderly and reasonable discovery, the trial judge required that a high standard of good cause be met before discovery beyond that provided in the uniform orders would be permitted. Its orders were tailored appropriately to meet this need. Appellants also strongly challenge the denial of their requests for additional discovery under Fed.R.Civ.P. 56(f). Rule 56(f) allows a party faced with a motion for summary judgment to request additional discovery, and the Supreme Court has suggested that such a request be granted when “the nonmoving party has not had an opportunity to make full discovery.” Celotex, 477 U.S. at 326, 106 S.Ct. at 2554. Here too we will reverse only if there was an abuse of discretion. Paddington Partners v. Bouchard, 34 F.3d 1132, 1137 (2d Cir.1994). Doubtless, more discovery might have been helpful, but in a case this complex the information falling within the broad purview of federal discovery is almost without limit. In this light it is hard to fault the district court for placing limitations on discovery. Given the circumstances of this case, we see no abuse of discretion either in its discovery orders or its decision rejecting the appellants’ Rule 56(f) motions. D. Expert Opinion A final contested issue common to most of the claims involved in the instant case concerns the district court’s evaluation of the scientific expert opinion of Dr. Kirk- W. Brown, a Professor of Soil and Crop Sciences at Texas A & M University. The coalitions relied on Dr. Brown’s affidavit as crucial evidence in opposing the appellees’ summary judgment motions. Before discussing the district court’s assessment of Dr. Brown’s affidavit, a brief review of the contents of the affidavit is necessary. 1. Expert’s Credentials and Methodology Dr. Brown holds a masters degree from Cornell University and a doctorate in agronomy from the University of Nebraska. He has written more than 150 scholarly, scientific articles and a book dealing with the disposal and treatment of hazardous waste substances. His professional qualifications make him an expert witness, and the district court so found. See Murtha II, 815 F.Supp. at 543. Dr. Brown submitted an affidavit in support of the coalitions’ position that appellees were potentially responsible parties at both landfill sites. In preparing this sworn statement, he reviewed numerous reports including, most significantly, six reports relating to the Laurel Park landfill and three reports discussing the Beacon Heights landfill. After examining these reports, the expert was of the opinion that “there [were] CERCLA hazardous substances present in each and every leachate sample from these landfills.” Dr. Brown did not personally visit the landfills or dig up any shovelfuls of waste, but he was not required to do so. He studied EPA reports concerning municipal and industrial landfills in the United States and he set forth the types of hazardous substances found in waste streams generated by households, commercial establishments, institutions, and industries. He used these reports to generate lists of the wastes and CERCLA hazardous substances routinely discarded by typical households, commercial establishments, institutions, and industries during the time period relevant to this case. After comparing the concentrations of hazardous substances in the leachates from the two landfills with the concentrations generally found in municipal solid waste landfills in the United States, he concluded, in his expert opinion, that the waste streams disposed of in the two lándfills were similar to the landfills he had studied and that they contained CERCLA hazardous substances. Further, Dr. Brown read the statements that the various appellees made in their deposition testimony and read other documents relevant to the kind of waste disposed of by each appellee. His comments varied for each appellee. Based on the available information, he meticulously linked the types of wastes, including identification of specific products discarded by each defendant, with specific CERCLA hazardous substances. His analysis relied on a series of studies and reports that showed which known CERCLA hazardous substances are in specific waste products. This allowed him to correlate the specific wastes disposed of by each appellee to specific hazardous substances. For example, when a party admitted to disposing of tire products, Dr. Brown explained which hazardous substances are found in normal tire products. Although he could not swear with absolute certainty that the tire products disposed by a specific ap-pellee were similar to normal tire products, such absolute certainty is not required to oppose successfully a summary judgment motion; rather, at the summary judgment stage, the trial judge must decide only whether a reasonable jury could find liability by a preponderance of the evidence. Further, when the appellees failed to submit any evidence to distinguish the products of which they disposed from those addressed by Dr. Brown in his affidavit, Dr. Brown’s probabilistic evidence stands largely uncontradicted. Moreover, each and every type of product disposed of need not contain hazardous substances. Any amount of such substances is sufficient to create liability.' 2. Admissibility of Brown Affidavit Although the district court agreed that Dr. Brown was an expert, and that his evidence could be admissible against certain parties, it apparently did not place much weight on the affidavit. It ruled that the expert’s sworn statement was not controlling for non-municipal generators, id. at 544, and that it could not be applied to the municipal generators because no adequate foundation for its aceep-' tance had been laid. Murtha III, 840 F.Supp. at 189. More specifically, the district court added, Dr. Brown’s “references to facts, without specification, in various documents relied on, are insufficient to make his opinions relevant.” Id. It therefore rejected the affidavit as lacking “probative merit” because it did not rely on “direct evidence.” Id. at 184. In addition, the court expressly weighed the affidavit against the backdrop of the EPA Municipal Policy, according to which the EPA made certain enforcement decisions. Id. at 188. Our view of this affidavit and its probative worth differs dramatically from that of the district court. This is partly because it is difficult for us to imagine an expert with more experience and knowledge in the hazardous substances field than Dr. Brown