MOORE, Circuit Judge.

Plaintiff-appellant Kalamazoo River Study Group (“KRSG”) is an unincorporated association of paper manufacturers seeking to recover costs incurred in the investigation and remediation of contamination by polychlorinated biphenyls (PCBs) in the Kalamazoo River. KRSG brought suit against eight corporate defendants, including defendants-appellees Me-nasha Corporation and Eaton Corporation, pursuant to the contribution provision of the Comprehensive Environmental Response, Compensation, and Liability Act (“CERCLA”), 42 U.S.C. § 9601 et seq. The district court granted summary judgment in favor of Menasha and in favor of Eaton with regard to its Kalamazoo and Marshall facilities, and it ruled in favor of Eaton with regard to its Battle Creek facility after a bench trial. Because the district court applied an incorrect Lability standard for CERCLA contribution actions, we REVERSE the judgments and REMAND for further proceedings.

I. BACKGROUND

We set forth the relevant facts underlying this dispute in a previous opinion:

In 1990, after nearly 20 years of investigating polychorinated [sic] biphenyl (PCB) contamination in the Kalamazoo River, the Michigan Department of Natural Resources (“MDNR”) determined that a three-mile portion of Portage Creek from Cork Street to the Kalamazoo River, and a 35-mile portion of the Kalamazoo River from this confluence downstream to the Allegan City Dam (the “Site”) were heavily concentrated with PCBs. Consequently, the MDNR listed the Site as an environmental contamination site under the Michigan Environmental Response Act, and the United States Environmental Protection Agency (“EPA”) listed the Site on the National Priority List (“NPL”) as a Superfund Site pursuant to CERCLA § 105. The MDNR and the EPA authorized the MDNR to conduct an Endan-germeni/Risk Assessment (E/RA) for the Site.

Following the E/RA, MDNR identified three paper mills — HM Holdings, Inc., Georgia-Pacific Corporation and Simpson Plainwell Paper Company— with facilities located on or near the Kalamazoo River as potentially responsible parties (“PRPs”) for the PCB contamination. All three companies entered into an Administrative Order by Consent (“AOC”) which required them to perform a Remedial Investigation and Feasibility Study (“RI/FS”) at the Site. Subsequently, James River Company, which voluntarily agreed to pay a portion of the RI/FS costs, joined with the other three companies to form an unincorporated association called the Kalamazoo River Study Group (“KRSG”).

Kalamazoo River Study Group v. Rockwell Int’l Corp., 171 F.3d 1065, 1066-67 *651 (6th Cir.1999) (footnotes and citations omitted). 1 Under the RI/FS, KRSG is required to extend its investigation upstream and downstream of the NPL site to include a ninety-five-mile stretch of the Kalamazoo River.

Although none of KRSG’s member corporations have admitted liability nor have they been adjudged legally liable for the contamination at the NPL site, KRSG has incurred significant costs for its performance of the RI/FS at the NPL site and will continue to incur substantial costs in connection with those activities as well as remediation of the NPL site. KRSG filed this action in the United States District Court for the Western District of Michigan in December of 1995, seeking to recover its response costs from eight other corporations 2 that allegedly contributed to the contamination at the NPL site. KRSG’s suit was based on CERCLA, the Michigan Natural Resources and Environmental Protection Act, Mich. Comp. Laws Ann. § 824.20101 et seq., and various common law theories. In its first case management order, the district court bifurcated the trial of this case into two phases: liability and allocation of damages. All of the district court’s rulings thus far have gone to liability. Only the district court’s judgments respecting two of the defendants — Eaton and Menasha — are at issue in this appeal.

On March 6, 1998, the district court granted summary judgment on the issue of liability to defendant Menasha. See Kalamazoo River Study Group v. Rockwell Int’l, 3 F.Supp.2d 799, 814 (W.D.Mich.1998). It was in this opinion that the district court articulated the liability standard that is challenged by KRSG in this appeal. The parties disputed whether CERCLA requires a plaintiff to show causation in the sense that a particular defendant’s release caused the plaintiff to incur costs. After examining the issue, the district court concluded that in a contribution action, the court should ask whether a particular defendant’s release was of sufficient significance to justify holding it liable for the response costs. Applying this “threshold of significance standard,” id. at 807, the court concluded that defendant Menasha was entitled to summary judgment.

In an opinion dated June 30, 1998, the court applied the threshold of significance standard in ruling on cross-motions for summary judgment as to liability filed by KRSG and Eaton. The court found a material question of fact with regard to Eaton’s liability for its facility at Battle Creek. However, the court granted summary judgment in favor of Eaton with regard to its facilities at Marshall and Kalamazoo.

From August 10 to August 17, 1998, the district court presided over a bench trial on the issue of Eaton’s liability with regard to the Battle Creek facility. In an opinion dated December 7, 1998, the district court entered judgment in favor of Eaton, explaining that “[bjecause the concentrations of PCBs upstream of Plaintiffs members are low, their incidence is sporadic, and they have not been located close to the Eaton facility, the Court concludes that Plaintiff KRSG has not met its burden of demonstrating that any PCBs released from Eaton’s Battle Creek facility have added to the PCB contamination of the Kalamazoo River.” J.A. at 346 (D.Ct.Op. 12/7/98).

On February 3, 1999, the district court granted KRSG’s motion to certify its March 6, June 30, and December 7 orders with respect to Eaton and Menasha as final and appealable pursuant to Federal Rule of Civil Procedure 54(b). We there *652 fore have jurisdiction pursuant to 28 U.S.C. § 1291.

II. ANALYSIS

The district court granted summary judgment on the issue of liability to Mena-sha, and to Eaton with regard to its Kalamazoo and Marshall facilities. We review de novo a district court’s grant of summary judgment. See Ercegovich v. Goodyear Tire & Rubber Co., 154 F.3d 344, 349 (6th Cir.1998). Summary judgment is proper only when there is no dispute as to a material question of fact and one party is entitled to a judgment as a matter of law. Fed.R.Civ.P. 56(c). Viewing all facts and inferences drawn therefrom in the light most favorable to the nonmovant, we must determine whether the evidence presented is such that a reasonable jury could find for that party. See Matsushita Elec. Indus. Co. v. Zenith Radio Corp., 475 U.S. 574, 587, 106 S.Ct. 1348, 89 L.Ed.2d 538 (1986).

The district court granted judgment in favor of Eaton with regard to its Battle Creek facility after a bench trial on the issue of liability. We review the legal conclusions of the district court de novo, but we review the district court’s factual findings following a bench trial for clear error. See Schroyer v. Frankel, 197 F.3d 1170, 1173 (6th Cir.1999).

A. The Proper Standard for CERCLA Contribution Actions

CERCLA is a statute designed “to facilitate the prompt cleanup of hazardous waste sites by placing the ultimate financial responsibility for cleanup on those responsible for hazardous wastes.” KRSG, 171 F.3d at 1068 (quotation omitted); see also B.F. Goodrich v. Betkoski, 99 F.3d 505, 514 (2d Cir.1996) (noting that the purposes of CERCLA include “facilitating efficient responses to environmental harm, holding responsible parties liable for the costs of the cleanup, and encouraging settlements that reduce the inefficient expenditure of public funds on lengthy litigation” (citation omitted)), cert. denied, 524 U.S. 926, 118 S.Ct. 2318, 141 L.Ed.2d 694 (1998).

Once a site has been cleaned up, CERCLA provides two causes of action for a party to recover the costs incurred as a result of the cleanup effort. The first is a recovery action governed by § 107(a), 42 U.S.C. § 9607(a). That section provides:

Notwithstanding any other provision or rule of law, and subject only to the defenses set forth in subsection (b) of this section—

(1) the owner and operator of a vessel or a facility,

(2) any person who at the time of disposal of any hazardous substance owned or operated any facility at which such hazardous substances were disposed of,

(3) any person who by contract, agreement, or otherwise arranged for disposal or treatment, or arranged with a transporter for transport for disposal or treatment, of hazardous substances owned or possessed by such person, by any other party or entity, at any facility or incineration vessel owned or operated by another party or entity and containing such hazardous substances, and

(4) 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, shall be liable for—

(A) all costs of removal or remedial action incurred by the United States Government or a State or an Indian tribe not inconsistent with the national contingency plan;

(B) any other necessary costs of response incurred by any other per *653 son consistent with the national contingency plan;

(C) damages for injury to, destruction of, or loss of natural resources, including the reasonable costs of assessing such injury, destruction, or loss resulting from such a release; and

(D) the costs of any health assessment or health effects study carried out under section 9604© of this title.

42 U.S.C. § 9607(a). “In order to establish a prima facie case for cost recovery under § 107(a), a plaintiff must prove four elements: (1) the site is a ‘facility’; (2) a release or threatened release of hazardous substance has occurred; (3) the release has caused the plaintiff to incur ‘necessary costs of response’; and (4) the defendant falls within one of the four categories of PRPs.” Centerior Serv. Co. v. Acme Scrap Iron & Metal Corp., 153 F.3d 344, 347-48 (6th Cir.1998). Liability under § 107(a) is generally joint and several on any defendant regardless of fault. See id. at 348.

In 1986 Congress amended CERCLA by enacting the Superfund Amendments and Reauthorization Act of 1986 (“SARA”), which explicitly provides for a right of contribution for private parties to recover costs associated with hazardous waste cleanup. See id. Section 113(f) states in relevant part:

Any person may seek contribution from any other person who is liable or potentially liable under section 9607(a) of this title, during or following any civil action under section 9606 of this title or under section 9607(a) of this title. Such claims shall be brought in accordance with this section and the Federal Rules of Civil Procedure, and shall be governed by Federal law. In resolving contribution claims, the court may allocate response costs among liable parties using such equitable factors as the court determines are appropriate. Nothing in this subsection shall diminish the right of any person to bring an action for contribution in the absence of a civil action under section 9606 of this title or section 9607 of this title.

42 U.S.C. § 9613(f)(1). The text of this provision instructs that parties seeking contribution under § 113 must look to § 107 to establish the basis and elements of the liability of the defendants. See Centerior, 153 F.3d at 350. Unlike with § 107, however, liability under § 113 is not joint and several, but several only, see id. at 348; 3 the provision grants the district court discretion to allocate response costs among liable parties.

1. The District Court’s March 6 Opinion

At the summary judgment stage, the district court was called upon to resolve a fundamental disagreement between the parties regarding KRSG’s burden of proof on the issue of causation. The defendants argued that a CERCLA plaintiff was required to prove that a particular defendant caused the plaintiff to incur response costs. KRSG disagreed, explaining that the “causation” element of the statute required a plaintiff to prove only that it incurred response costs as.a result of a release, not as a result of each particular defendant’s release. KRSG contended that a defendant should be held liable if a plaintiff proved that a release caused the incurrence of *654 response costs and that the defendant fell into one of the four categories of responsible parties, regardless of the amount of hazardous substances that the particular defendant discharged or the harm it caused. The district court explained:

This Court agrees that in a multi-generator context such as this, Plaintiff cannot be required to trace or fingerprint the waste from each PRP. The Court will not require the Plaintiff to prove that a particular defendant caused Plaintiff to incur response costs. On the other hand, this Court does not agree with Plaintiffs contention that a defendant’s release of any quantity, even the slightest, of hazardous substances is enough to support a judgment against that defendant for an undetermined extent in a contribution action brought by an admittedly responsible party.

KRSG, 3 F.Supp.2d at 804.

The district court went on to draw a distinction between cost recovery actions brought by innocent parties pursuant to § 107 and contribution actions brought by one PRP against another PRP pursuant to § 113. The court explained that the two provisions have different statutes of limitations and that there is also a difference as to whether liability will be several or joint. See id. at 805. Most importantly, the court noted that contribution actions, unlike cost recovery actions, are governed by the equitable contribution principles set forth in § 113(f): “These equitable contribution principles permit the court to consider whether or to what degree the defendant caused the response costs in a § 113(f) contribution action.” Id. In support of this contention, the district court cited two cases from the Fifth Circuit, but it relied primarily on a decision from the United States District Court for the District of Massachusetts.

In Acushnet Co. v. Coaters, Inc., 948 F.Supp. 128 (D.Mass.1996), aff'd, 191 F.3d 69 (1st Cir.1999), the district court reasoned that § 113(f) authorizes an equitable approach to determining liability:

The “equitable factors” mandate is appropriately interpreted as at least authorizing, if not mandating for this case, making comparisons between plaintiffs in a contribution action and defendants whom they allege to be potentially responsible parties, by considering evidence before the court (and jury, if one is used) for the purpose of determining whether the nature and extent of their respective ties to the hazards to persons, property, and the environment make it fair and reasonable to order that a defendant, or defendants grouped together for good reason, reimburse the plaintiffs in some amount or some share of “contribution”, allocated on an equitable basis reasoned from evidence concerning all the “equitable factors” for the application of which in this case some eviden-tiary basis exists.

Id. at 135-36. Articulating what has come to be known as the “threshold of significance standard,” the district court in Acushnet held that “plaintiffs must proffer sufficient evidence as to a particular defendant to satisfy a minimum standard of significance of that defendant’s responsibility as a source of one or more hazardous substances deposited at the site.” Id. at 136. The Acushnet decision, which has been criticized by commentators as a deviation, from prior CERCLA case law, 4 was plainly driven by the court’s concern for the expense that CERCLA litigation imposes on defendants who have been only small contributors to the environmental harm. See id. at 136 (“[I]t is inequitable to put the defendant to the burden of defending itself when the predictable outcome as to the claim for allocation of an equitable share of contribution, in the *655 event of an outcome as favorable to plaintiffs as can be reasoned from evidence, would be a share so small that the public interest in remediation of hazardous waste sites would have been disserved because of the commitment of public and private resources to litigation over that alleged share of contribution.”).

In the instant case, the district court found the Acushnet court’s reasoning persuasive. After noting that it was not aware of any Sixth Circuit case law discussing causation under CERCLA, the district court held that in a contribution action a court should apply a test “that asks whether a particular defendant’s responsibility was of sufficient significance to justify the response costs.” KRSG, 3 F.Supp.2d at 807. Therefore, the court explained, “[i]n reviewing the cross-motions for summary judgment, this Court will apply the threshold of significance standard: is the evidence of defendant’s release of sufficient significance to justify holding defendant liable for response costs?” Id.

2. Analysis of District Court’s Opinion

We begin our analysis of the district court’s opinion by examining the liability standard for cost recovery actions under § 107. It is clear from the text, structure, and legislative history of § 107 that the provision does not require a plaintiff to show that a particular defendant caused either the release or the incurrence of response costs in order to prove liability. See United States v. Township of Brighton, 153 F.3d 307, 329 (6th Cir.1998) (Moore, J., concurring in the result) (“In the absence of a proximate causation requirement; CERCLA imposes liability on a generator of hazardous waste at a particular facility even though that generator’s acts may not directly have caused or contributed to the contamination, or even where their waste may have comprised only a small portion of the waste present at the site.” (citations omitted)). First, the text of § 107 imposes no such causation requirement on its face. Rather, the text requires only that a plaintiff prove “that the defendant’s hazardous substances were deposited at the site from which there was a release and that the release caused the incurrence of response costs.” United States v. Alcan Aluminum Corp., 964 F.2d 252, 266 (3d Cir.1992) (“Alcan I”). Second, the structure of § 107 suggests this conclusion. As the Second Circuit has observed, “[ijnterpreting section 9607(a)(1) as including a causation requirement makes superfluous the affirmative defenses provided in section 9607(b), each of which carves out from liability an exception based on causation.” New York v. Shore Realty Corp., 759 F.2d 1032, 1044 (2d Cir.1985). “Finally, the legislative history supports the absence of a causation requirement, as the final version of the bill ultimately passed by Congress deleted the requirement that liability be imposed only on those who ‘caused or contributed to the release or threatened release’ contained in the earlier version passed by the House of Representatives.” Township of Brighton, 153 F.3d at 328 (Moore, J., concurring in the result) (quotation omitted). Moreover, other circuits have held that § 107 does not require a plaintiff to show any direct causal link between the waste each defendant sent to the site and the environmental harm. See, e.g., United States v. Alcan Aluminum Corp., 990 F.2d 711, 721 (2d Cir.1993) (“Alcan II”) (“What is not required [by § 107] is that the government show that a specific defendant’s waste caused incurrence of clean-up costs.”); Amoco Oil Co. v. Borden, 889 F.2d 664, 670 n. 8 (5th Cir.1989) (“[I]n cases involving multiple sources of contamination, a plaintiff need not prove a specific causal link between costs incurred and an individual generator’s waste.”); Alcan I, 964 F.2d at 266 (“Decisions rejecting a causation requirement between the defendant’s waste and the release or the incurrence of response costs are well-reasoned, consistent with the plain language of the statute and consistent with the legislative history *656 of CERCLA.”); United States v. Monsanto Co., 858 F.2d 160, 169 (4th Cir.1988) (explaining that a CERCLA plaintiff need not prove a nexus between a defendant’s waste and the resulting environmental harm), cert. denied, 490 U.S. 1106, 109 S.Ct. 3156, 104 L.Ed.2d 1019 (1989).

The district court in the instant case suggested, however, that CERCLA contribution actions may be subjected to a different liability standard than cost recovery actions brought solely under § 107. Rather than holding that causation is an element of liability under § 107, therefore, the district court seems to have “fashioned the ‘threshold of significance’ standard to impose a causation requirement only on parties seeking contribution under Section 113(f).” Br. of Appellant at 37.

The district court’s imposition of a requirement that plaintiffs in contribution actions show causation in order to establish a defendant’s liability was erroneous. The Lability standard for contribution claims is the same as the standard for cost recovery claims. This conclusion is clear from the plain language of § 113, which states that any person may seek contribution “from any other person who is liable or 'potentially liable under section 9607(a) of this title.” 42 U.S.C. § 9613(f)(1) (emphasis added). As we have explained, “parties seeking contribution under § 113(f) must look to § 107 to establish the basis and elements of the liability of the defendants.” Centerior, 153 F.3d at 350; see also Redwing Carriers, Inc. v. Saraland Apartments, 94 F.3d 1489, 1496 (11th Cir.1996) (“Whether [a plaintiff] brings its claims under § 107(a) or § 113(f) does not matter insofar as establishing the [defendant’s] liability. The elements of a claim under both sections are the same.” (footnote omitted)). Because causation is not an element of liability under § 107, and because § 107 defines the liability standard applicable in actions brought pursuant to § 113, then a § 113 plaintiff need not prove causation in order to establish a defendant’s liability. 5

It is true that the equitable contribution principles of § 113(f) permit a district court to consider causation. However, consideration of causation is proper only in allocating response costs, not in determining liability. See 42 U.S.C. § 9613(f)(1) (“In resolving contribution claims, the court may allocate response costs among liable parties using such equitable factors as the court determines are appropriate.” (emphasis added)). As the Eighth Circuit has noted: “Recovery of response costs by a private party under CERCLA is a two-step process. Initially, a plaintiff must prove that the defendant is *657 liable under CERCLA. Once that is accomplished, the defendant’s share of liability is apportioned in an equitable manner.” Control Data Corp. v. S.C.S.C. Corp., 53 F.3d 930, 934 (8th Cir.1995) (footnote omitted); see also Tosco Corp. v. Koch Indus., Inc., 216 F.3d 886, 891 (10th Cir.2000) (“The plaintiff in a CERCLA response cost recovery action involving multiple potentially responsible persons need not prove a specific causal link between costs incurred and an individual responsible person’s waste. To establish liability under § 9613(f), it is sufficient for the plaintiff to establish a connection between a particular defendant and the incurred response costs vis a vis the defendant’s identification as a responsible person as defined in § 9607(a).” (citation omitted)).

Indeed, the introduction of a causation element into the liability standard for contribution actions could have the effect of thwarting CERCLA’s central purpose— facilitating the prompt cleanup of hazardous waste. CERCLA’s scheme of strict liability for responsible parties in conjunction with the availability of contribution actions employing the same liability standard and contribution protection for parties who have settled with the government serves to encourage parties to clean up the site quickly and then litigate later to sort out the specifics of who should pay. As commentators have explained, this format provides “[t]he inevitable handful of PRPs selected by the government to remediate a site or reimburse the government’s cleanup costs [with] powerful incentives to invest their resources in spreading the costs to other PRPs rather than fighting a losing battle against the government.” Sro-lovic & Esterman, supra n. 4, at 469-70. The district court’s imposition of a threshold causation requirement for parties seeking contribution under § 113 could discourage parties from voluntary cleanup efforts and from settlement. See Goodheart & McGuire, supra n. 4, at 332 (“Instead of a presumption of shared liability and a focus on respective fair shares of the cleanup costs, contribution plaintiffs now face the prospect of being required to establish that a particular defendant in fact contributed at least a minimally significant share of the wastes at issue — a burden which courts have routinely explained is far too great to impose on government plaintiffs in cost recovery actions. As a result, some PRPs may find it undesirable to pay as much as they previously would have to settle governmental claims, and may even eschew settlement altogether.”).

In the instant case, it has been suggested that the district court’s decision could be affirmed on the basis that the court’s “threshold of significance standard,” rather than reflecting a determination as to liability, may simply have been a determination based on the exercise of its equitable powers under § 113(f) that, even if the defendants were liable, they should nevertheless bear a zero allocation of costs. On review of the Acushnet decision, the First Circuit, although disapproving of the district court’s decision “[t]o the extent that [it] may be interpreted to incorporate into CERCLA a causation standard that would require a polluter’s waste to meet a minimum quantitative threshold,” nevertheless affirmed on the different ground that the result was justified under the equitable allocation principles of § 113(f). Acushnet Co. v. Mohasco Corp., 191 F.3d 69, 72 (1st Cir.1999). The First Circuit explained: ‘While the judge was not making specific allocations, it is plain to us he was holding that, in light of the equitable factors he would apply should he make explicit findings, plaintiffs’ evidence showed too little pollution to justify compelling defendants to take on any meaningful share of the response costs. We read him to say that if he had to make an allocation for [these three defendants], the evidence dictated that each of their shares for response costs would be zero.” Id. at 81. Unlike in Acushnet, however, this was only the liability phase of a bifurcated trial. Because, as we have explained, a district court may not consider causation in making a liability determination, any exercise of its equitable *658 powers at this first stage of a bifurcated trial would have been improper. 6

B. The District Court’s Determination of Liability with Regard to Each Defendant

The district court erroneously required KRSG to show, as part of its prima facie case of liability, that each defendant’s discharge of PCBs was causally linked to KRSG’s incurrence of response costs. Under the correct standard, which includes no specific causation requirement, a plaintiff must prove: (1) that the site is a “facility”; (2) that a release of hazardous substances has occurred; (3) that the release caused it to incur “necessary costs of response”; and (4) that each defendant is a PRP. See Centerior, 153 F.3d at 347-48.

KRSG has proceeded on a theory that Menasha and Eaton are responsible parties pursuant to 42 U.S.C. § 9607(a)(3), which includes “any person who by contract, agreement, or otherwise arranged for disposal or treatment, or arranged with a transporter for transport for disposal or treatment, of hazardous substances owned or possessed by such person, by any other party or entity, at any facility or incineration vessel owned or operated by another party or entity and containing such hazardous substances.” KRSG contends that the two defendants fall into this category because each discharged hazardous substances (PCBs) into the Kalamazoo River (the site). J.A. at 1320 (Pl.’s Mot. for Summ. J. Against Def. Menasha); J.A. at 2559 (Pl.’s Mot. for Partial Summ. J. Against Def. Eaton); see New Jersey Turnpike Auth. v. PPG Indus., Inc., 197 F.3d 96, 104 (3d Cir.1999) (“In order to prove a case where a CERCLA plaintiff asserts that a PRP has ‘arranged’ for the transportation or disposal of hazardous substances, our prior case law is clear that such a plaintiff ‘must simply prove that the defendant’s hazardous substances were deposited at the site from which there was a release and that the release caused the incurrence of response costs.’ ” (quoting Alcan I, 964 F.2d at 266)); see also Alcan II, 990 F.2d at 721 (“CERCLA § 9607(a) imposes strict liability on ‘any person who by contract, agreement, or otherwise arranged for disposal or treatment’ of hazardous substances ‘from which there is a release, or a threatened release which causes the incurrence of response costs.’ The plain meaning of this language dictates that the government need only prove: (1) there was a release or threatened release, which (2) caused incurrence of response costs, and (3) that the defendant generated hazardous waste at the clean-up site.” (quotations omitted)). The relevant inquiry in this case is therefore whether each defendant discharged PCBs to the site — not whether it discharged PCBs to the site in sufficient quantity to have justified KRSG’s incurrence of response costs.

1. Menasha

Defendant Menasha owns and operates a recycled paper mill located in Otsego, Michigan, and adjacent to the Kalamazoo River within the site. As the district court explained, between 1957 and 1971, a type of carbonless copy paper — ■ typically referred to as “NCR” paper — -was manufactured; this paper incorporated Ar- *659 odor 1242, which contains PCBs, as a solvent. See KRSG, 3 F.Supp.2d at 802. Although Menasha never used NCR paper in its feedstock, it did use secondary fibers known as double-lined kraft (“DLK”) and old corrugated containers (“OCC”), both of which could have contained small levels of PCBs. J.A. at 575 (Schell Aff.) (“As a result of the intensive recycling efforts which started in the mid-1970s and continue on a large scale basis even today, these PCBs which originated in the NCR paper have slowly infiltrated a large portion of the world’s paper stock, albeit in a steadily decreasing concentration. Thus, in theory, any mill that uses even small quantities of post-consumer paper as furnish could have inadvertently introduced minute quantities of PCBs into the production line.”).

KRSG presented evidence that on two occasions the finished product coming off the rolls at the Otsego mill tested positive for PCBs. There was also testimony that during the pulping process, wastewater that may contain these recycled secondary fibers is lost to the mill’s wastewater treatment system and eventually discharged to the Kalamazoo River. J.A. at 803 (Roys Dep.). To support its theory that the PCBs used in Menasha’s pulping process were eventually discharged into the Kalamazoo River, KRSG presented evidence that Menasha’s effluent tested positive for low levels of PCBs on four separate occasions.

In response, Menasha argues that the wastewater discharge test results submitted by KRSG were unreliable. Specifically, Menasha’s expert testified that the test results from the 1970s were “highly unreliable because the reported analytical detection limits were unachievable using the technology available at that time.” J.A. at 572 (Schell Aff.). Additionally, Menasha argues that because it used contaminated water from the Kalamazoo River in its wastewater treatment and cooling water systems, each of the effluent samples identified as containing PCBs could simply have reflected that fact rather than proving that PCBs were discharged as a result of Menasha’s processes. J.A. at 571 (Schell Aff.).

Applying the correct standard, it is clear that summary judgment in favor of Mena-sha was inappropriate. KRSG has presented direct evidence that Menasha discharged PCBs into the Kalamazoo River. Although Menasha contests the reliability of KRSG’s test results and proffers an alternative hypothesis for the presence of PCBs in its effluent, there is at least a genuine issue of material fact as to whether Menasha contributed to the contamination at the site.

2. Eaton

At all relevant times, Eaton owned three automotive manufacturing facilities near the Kalamazoo River. Although PCBs were admittedly present in the electrical equipment at each of the three plants, there was no evidence of any leaks from the electrical equipment that could have resulted in PCBs being discharged into the Kalamazoo River. Instead, KRSG’s theory of liability is' that there were PCBs in Eaton’s process oils (quench, hydraulic, and cutting oils), which were discharged into the Kalamazoo River. Eaton does not contest that process oils were likely discharged into the Kalamazoo River; it argues instead that its process oils did not contain PCBs.

a. Kalamazoo Facility

Eaton’s Kalamazoo plant, which manufactured transmissions until it was sold in 1985, is located about one-half mile from the Kalamazoo River. Wastewater from the Kalamazoo Plant was discharged into the Zantman Drain, which emptied into the Kalamazoo River. Eaton does not contest that limited amounts of process oils escaped in wastewater.

The wood floors from the Kalamazoo plant were tested in a PCB survey, and these tests revealed PCB contamination. This contamination was primarily concen *660 trated around the electrical equipment, although PCBs were also found in other areas. Eaton’s expert concluded, based on the fact that the contamination was largely localized, that the contamination resulted from limited leakage from the electrical equipment. J.A. at 1921 (Wharton Letter).

KRSG, however, presented evidence that the PCB contamination found in the floor samples was caused, at least in part, by PCB-contaminated quench oils. Stuart Lightfoot, Eaton’s former director of environmental engineering, was questioned about the cause of the contamination at the Kalamazoo plant:

Q: Were you ever able to ascertain the cause of the PCB contamination at the Kalamazoo facility?

A: We think we did.

Q: And what did you think that cause was?

A: A dripping spicket on an internal wet transformer, and a heat treat oil quench operation.

Q: So you think you determined that that [sic] oil quench operation at the Kalamazoo facility was in part the cause of the PCB contamination at that facility?

A: Yes. We think that is true.

J.A. at 2690 (Lightfoot Dep.). Although no PCB testing was ever conducted on the quench oils used at the Kalamazoo facility, Lightfoot explained that he suspected that the facility used quench oils containing PCBs because of their fire-retardant qualities. Additionally, KRSG presented the deposition testimony of Thomas Newell, a former MDNR engineer, who expressed a belief that PCBs may occasionally be in the oils used at the plant at trace contaminant levels. J.A. at 2783 (Newell Dep.). Newell based this opinion on his general knowledge that many of the oils used in the automotive industry were recycled and were contaminated with PCBs.

The district court discounted Lightfoot’s testimony because Lightfoot could not be sure that the quench oils used contained PCBs, and it found that Newell’s general knowledge about the automobile parts manufacturing industry was not probative of what occurred at Eaton. However, the evidence presented by KRSG, although circumstantial, was sufficient to enable a reasonable juror to conclude that Eaton used PCBs in its quench oils at the Kalamazoo facility. The district court therefore erroneously granted summary judgment to Eaton.

b. Marshall Facility

Eaton’s still operable Marshall facility machines, grinds, heat-treats, and assembles components for the automobile industry. The facility is located one-quarter mile from the Kalamazoo River and approximately 30 miles upstream of the site.

With regard to this facility, KRSG presented test results indicating the presence of PCBs in the effluent from the Marshall facility. In 1980, MDNR performed a wastewater survey at the Marshall facility, and samples indicated the presence of PCBs. Newell indicated that the PCBs that were found were “most likely associated with the oils that are used at the plant.” J.A. at 3388 (Newell Letter). Although the results of this one test were never repeated, it is direct evidence that the Marshall facility discharged PCBs into the Kalamazoo River. 7 Summary judgment in Eaton’s favor was therefore improper.

Because we conclude that a material question of fact exists as to the liability of Menasha, Eaton-Kalamazoo, and Eaton-Marshall, we will reverse the district *661 court’s grant of summary judgment and remand for a determination of liability under the appropriate standard. With regard to Eaton’s Battle Creek facility, the district court ruled in favor of Eaton after a bench trial. On remand, the district court should similarly re-evaluate the liability of Eaton-Battle Creek under the proper standard.

C. The District Court’s Evidentiary Ruling

KRSG also argues that the district court erred in refusing to admit a certain document regarding Eaton’s Battle Creek facility under the ancient documents exception to the hearsay rule. “We review factual components of the district court’s evidentiary determinations under an ‘abuse of discretion’ standard, and review its legal conclusions de novo.” Reynolds v. Green, 184 F.3d 589, 596 (6th Cir.1999).

Rule 803(16) of the Federal Rules of Evidence provides an exception to the hearsay rule for “[statements in a document in existence twenty years or more the authenticity of which is established.” Fed.R.Evid., 803(16). Rule 901 provides that “[t]he requirement of authentication or identification as a condition precedent to admissibility is satisfied by evidence sufficient to support a finding that the matter in question is what its proponent claims.” Fed.R.Evid. 901(a). Under the ancient documents provision of Rule 901, a document is authentic if it “(A) is in such condition as to create no suspicion concerning its authenticity, (B) was in a place where it, if authentic, would likely be, and (C) has been in existence 20 years or more at the time it is offered.” Fed.R.Evid. 901(b)(8).

Although we have not often had occasion to address the proper application of the ancient documents exception to the hearsay rule, the Third Circuit has explained:

While the ancient documents provision has not been a subject of frequent discussion in reported opinions, those cases which do address the provision establish that the point of a Rule 901(b)(8) inquiry is to determine whether the documents in question are, in fact, what they appear to be. “Although the rule requires that the document be free of suspicion, that suspicion does not go to the content of the document but rather to whether the document is what it purports to be.... ” Questions as to the documents’ content and completeness bear upon the weight to be accorded the evidence and do not affect the threshold question of authenticity. The determination that a set of documents are, indeed, prim a fa-cie authentic in no way precludes counsel from challenging the content of the documents or from arguing that missing documents subject the contents to a different interpretation.

Threadgill v. Armstrong World Indus., Inc., 928 F.2d 1366, 1375-76 (3d Cir.1991) (quoting United States v. Kairys, 782 F.2d 1374, 1379 (7th Cir.), cert. denied, 476 U.S. 1153, 106 S.Ct. 2258, 90 L.Ed.2d 703 (1986)) (citations omitted).

The excluded document is purportedly an EPA report indicating that Eaton’s Battle Creek facility purchased from Monsanto large quantities of a substance known as Pydraul A-200, which contained PCBs. Specifically, the document indicates that Eaton Battle Creek purchased 1,940 pounds of Pydraul in 1970 and 645 pounds in 1971. J.A. at 4964. At the bench trial KRSG called MDNR employee Tom Rohrer to the stand to identify the document. Rohrer testified that the document came from the files of the Lansing office of MDNR, and that it had been in those files for over twenty years. J.A. at 4445-46 (Rohrer Test.). Rohrer personally reviewed the document in the fall of 1977 through the spring of 1978 in connection with a study MDNR was conducting. Rohrer testified that the document was not prepared by MDNR, but that “[he] be-lievefd] it was prepared by the United States Environmental Protection Agency.” J.A. at 4447 (Rohrer Test.).

*662 KRSG then offered the exhibit as an exception to the hearsay rule as an ancient document, for the purpose of showing that Monsanto made certain sales to Eaton Battle Creek. The court refused to admit the document, explaining that its reliability was seriously in question.

Considering all the circumstances, we find no error in the district court’s ruling. Most important, Rohrer was unable to state with certainty that the document was even prepared by the EPA; he could only state that it was his belief that the document “was a page out of an EPA report if I recall correctly.” J.A. at 4449 (Rohrer Test.). Moreover, Rohrer’s testimony does not establish why the document was in the MDNR files. We therefore affirm the district court’s exclusion of this document from consideration.

III. CONCLUSION

Because the district court applied an incorrect legal standard for determining liability under CERCLA, we REVERSE its judgments. Under the correct standard, the record makes clear that summary judgment was erroneously granted to Menasha and to Eaton with regard to its Kalamazoo and Marshall facilities. Therefore, we REMAND for a determination of the liability of those three facilities using the correct standard, and for a rede-termination of the liability of Eaton’s Battle Creek facility. After liability has been determined, the district court may properly consider the causal link between each defendant’s waste and the resulting environmental harm, along with other relevant equitable factors, in allocating response costs among the liable parties.

Notes

. In that opinion, we affirmed the district court's grant of summary judgment to defendant Benteler Industries.

. The defendants named in the complaint were Rockwell International, Eaton Corporation, Wells Aluminum Corporation, Hercules, Inc., Benteler Industries, Inc., Menasha Corporation, Upjohn Company, and Rock-Tenn Company, Mill Division, Inc.

. In Carter-Jones Lumber Co. v. Dixie Distributing Co., 166 F.3d 840 , (6th Cir.1999), this court reaffirmed this general principle. See id. at 847 (citing Centerior for the proposition that “[i]t is correct that unrelated defendants are to be made severally liable only, rather than jointly and severally liable, in an action for contribution under CERCLA”). The Carter-Jones court reversed the district court’s determination that in a CERCLA contribution action the liability of a corporation and its sole shareholder must be several as a matter of law, but it did so on state law grounds. The court explained: "Section 113(f) of CERCLA, the section governing contribution actions, in no way addresses issues of corporate liability, and it should not therefore be presumed to alter state laws governing the liability of corporations vis a vis their officers and owners.” Id. (footnote omitted).

. See, e.g., Lisa C. Goodheart & Karen A. McGuire, Revisiting the Issue of Causation in CERCLA Contribution Litigation, 82 Mass. L. Rev. 315, 319 (1998); Lemuel M. Srolovic & Pamela R. Esterman, Fold or Fight: The Changing Settlement Calculus in CERCLA Enforcement Actions, 9 Fordham Envtl. L J. 469, 480-81 (1998).

. Although the district court relied on two Fifth Circuit cases in addition to the Acushnet decision to support its causation requirement, those cases are inapposite. In Amoco Oil Co. v. Borden, Inc., 889 F.2d 664 (5th Cir.1989), Amoco sought a declaratory judgment that Borden, from which Amoco had purchased contaminated industrial property, was liable for response cost damages under CERCLA. Borden argued that it should not be held liable because the release that occurred was not sufficiently hazardous to justify any response costs; it thus argued that Amoco failed to prove the requisite causal nexus between the release and response costs. The Fifth Circuit agreed with Borden that use of a "standard of justification” was acceptable for determining whether a release caused the in-currence of response costs; "[T]he question of whether a release has caused the incurrence of response costs should rest upon a factual inquiry into the circumstances of a case and the relevant factual inquiry should focus on whether the particular hazard justified any response actions.” Id. at 670. The Amoco case was concerned with the statutory causation requirement (a causal nexus between the release and the incurrence of response costs), not the type of causation requirement urged by the defendants in the instant case (a causal nexus between a specific defendant's release and the incurrence of response costs). In fact, the Amoco court specifically distinguished the case before it from a case "involving multiple sources of contamination,” in which "a plaintiff need not prove a specific causal link between costs incurred and an individual generator’s waste.” Id. at 670 n. 8. The Fifth Circuit’s decision in Licciardi v. Murphy Oil U.S A., Inc., 111 F.3d 396 (5th Cir.1997) (per curiam), simply applied the principles set forth in Amoco.

. Moreover, it is far from clear that the two analyses — liability and cost allocation — may properly be conflated even in a trial that is not bifurcated. The First Circuit in Acushnet reasoned that “allowing a CERCLA defendant to prevail on issues of fair apportionment, even at the summary judgment stage, is consistent with Congress's intent that joint and several liability not be imposed mechanically in all cases,” and that "the costs and inherent unfairness in saddling a party who has contributed only trace amounts of hazardous waste with joint and several liability for all costs incurred outweigh the public interest in requiring full contribution from de minimis polluters.” Acushnet, 191 F.3d at 78-79. However, we have stated that liability under § 113 is ordinarily not joint and several, but several only. See Centerior, 153 F.3d at 348. There is therefore no possibility that a de minimis defendant will be saddled at the apportionment stage with costs for which it is not responsible.

. Eaton argues that a one-time detection of PCBs in the plant's effluent is insufficient to withstand a motion for summary judgment because there is no evidence that such a result was typical of Eaton-Marshall's effluent. However, one discharge is sufficient to support liability; there is no requirement that the generator typically discharge waste to the site.