NORRIS, J., delivered the opinion of the court, in which KENNEDY, MILBURN, NELSON, BOGGS, SILER, and BATCHELDER, JJ., joined and in which MERRITT, J., joined as to Part III.C.3. MERRITT, J. (pp. 583-86), delivered a separate opinion concurring in part and dissenting in part. RYAN, J. (pp. 586-95), delivered a separate dissenting opinion in which MARTIN, C.J., DAUGHTREY, and MOORE, JJ., joined and in which MERRITT, J., joined as to Part I.
OPINION
ALAN E. NORRIS, Circuit Judge.This appeal highlights the difficulty that often attends the apportionment of liability for the clean-up costs of sites that have been subjected to long-term environmental degradation. In the present case, brought pursuant to the Comprehensive Environmental Response, Compensation, and Liability Act (“CERCLA”), 42 U.S.C. §§ 9601-9675 (1988 & Supp. V 1993), the environmental damage occurred over a period of decades and during the watch of several owners.
A central concern on appeal is the criteria required under CERCLA before a parent corporation can be held financially liable for pollution that occurred on a site owned by a subsidiary. Because we adopt a stricter standard than did the district court for imposition of such liability, we reverse certain of its determinations and remand for further proceedings.
I. PROCEEDINGS BELOW
In May and June 1991, the district court conducted a fifteen-day bench trial to determine which parties were responsible for clean-up costs related to pollution of a site located in Dalton Township, Michigan. In addition to the five testimony of twenty-nine witnesses, the court received more than 2,300 exhibits and reviewed dozens of deposition transcripts. Given the complexity of the proceedings below, the factual findings contained in the district court’s published opinion are extensive. CPC Int’l, Inc. v. Aerojet-General Corp., 777 F.Supp. 549, 555-70 (W.D.Mich.1991). We summarize them here by way of background.
Beginning in 1957, a series of owners used the Dalton Township site to manufacture chemicals. The initial owner, the Ott Chemical Company (“Ott I”), controlled the site from 1957 until 1965. During this time, the groundwater flowing underneath the' site became contaminated, a development confirmed by tests conducted in 1964.
*576Pollution of soil, surface water, and groundwater continued after the Ott Chemical Company (“Ott II”), a wholly owned subsidiary of CPC International, Inc. (“CPC”), took over ownership of the site in 1965. The use of unlined lagoons as a means of chemical waste disposal was the principal cause of the contamination. According to the district court, this practice spanned the period from 1959 until at least 1968.
Seepage from these lagoons did not, however, constitute the sole source of pollution that occurred during the ownership of Ott I and Ott II. Further contamination emanated from chemical spills from train cars, from chemical drums, from overflows of chemicals contained in a cement-lined equalization basin, and from other sources. Groundwater pollution did not go completely untreated during this time; from 1965 until 1974, purge wells were operated intermittently in an attempt to alleviate the problem.
In 1972, the Story Chemical Company (“Story”) acquired the site from Ott II and continued to operate it until 1977, when bankruptcy ended operations. At that point, the trustee in bankruptcy assumed title to the site and attempted to find a buyer.
Active governmental response to the pollution problems at the site began in 1977, after Story’s bankruptcy, when the Michigan Department of Natural Resources (“MDNR”) visited the site to assess the situation. In view of the severity of the environmental problems and the lack of resources to pay for a cleanup, the MDNR became active in an effort to attract a purchaser who would participate financially in clean-up efforts. This search led to the signing of a document on October 13, 1977, by the Cordova Chemical Company (“Cordova/California”), a wholly owned subsidiary of Aerojet-General Corporation (“Aerojet”), and the MDNR. The district court described the agreement and its aftermath:
It addressed the problem of environmental contamination at the property and set forth obligations with respect to cleanup activities____
... MDNR agreed to remedy the waste container and sludge problems, and Cordova/California agreed to eliminate the phosgene gas and give MDNR $600,000 to defray the costs of the agency’s cleanup of the waste containers, sludge and residential wells.
With respect to Cordova/California’s $600,000 payment and the company’s responsibility or liability for the contamination at the site it was acquiring, the [agreement] stated:
Cordova Chemical Company shall not have any responsibility or liability in connection with any other corrective actions which the Department of Natural Resources or any other governmental agency may hereafter deem necessary____
However, the agreement did not provide for a total cleanup of the site’s severe environmental problems____
In particular, MDNR and Cordova/California did not reach an agreement regarding a remedy for the groundwater contamination problem. Instead, the fate of the groundwater problems was not resolved, with MDNR left to tackle the problem as part of its overall regulatory responsibility for the site.
... Cordova/California and MDNR fulfilled their cleanup obligations under the [agreement].
CPC Int’l v. Aerojetr-General, 777 F.Supp. at 564-67.
Having executed this document, Cordova/California purchased the site the following day from the Story bankruptcy trustee. Cordova Chemical Company of Michigan (“Cordova/Michigan”), a wholly owned subsidiary of Cordova/California, acquired ownership of the site in 1978. Cordova/Michigan retains ownership, -although manufacturing operations at the site ceased in 1986.
The district court made the following observations regarding conditions at the site during the ownership of the Cordova companies:
*577Dining their period of operations, [the companies] neither buried waste nor dumped it onto the ground. No chemical waste was disposed into the unlined lagoons that had been used during the Ott I and Ott II eras. Before beginning chemical manufacturing, Cordova/Miehigan repaired the equalization basin and chemical sewer system. When operating, Cordova/Miehigan discharged chemical waste through off-site disposal or to a sewer that flowed to the Muskegon County treatment facility.
Id. at 556. In short, although the preexisting groundwater contamination problem was not remedied during their ownership, the trial court concluded that neither Cordova/California nor Cordova/Miehigan exacerbated the condition.1
The federal Environmental Protection Agency became involved in cleanup of the site in 1981. Since then, the EPA has formulated a long-term response to the environmental damage that has occurred at the site; the cost of this effort will run into the millions of dollars.
II. CERCLA LIABILITY
Section 107(a) of CERCLA lists the parties who are potentially liable for the cleanup coste of a polluted site.2 42 U.S.C. § 9607(a). For the purposes of this action, those parties include the present owner and operator of a facility from which there is a release of a hazardous substance, any prior owner or operator of a facility whose involvement coincided with disposal of a hazardous substance, and any person who arranged for the disposal or transport of hazardous waste from a facility. 42 U.S.C. § 9607(a)(l)-(3). The parties stipulated that the site is a “facility” as defined by CERCLA, that it contains “hazardous substances,” that “releases” of hazardous substances have occurred and threaten to continue, and that CPC, the MDNR, Aerojet, Cordova/California, and Cordova/Miehigan are “persons” as defined by the statute. Id. at 556.
Because courts that have been asked to render liability decisions in CERCLA actions frequently invoke the remedial purpose of the act, e.g., United States v. Kayser-Roth Corp., Inc., 910 F.2d 24, 26 (1st Cir.1990), cert. denied, 498 U.S. 1084, 111 S.Ct. 957, 112 L.Ed.2d 1045 (1991), we will review that subject before considering the liability of those parties now before us.
Congress enacted CERCLA as a “remedial statute designed to protect and pre-, serve public health and the environment.” Kayser-Roth, 910 F.2d at 26; accord Schiavone v. Pearce, 79 F.3d 248, 253-54 (2d Cir. 1996); Lansford-Coaldale Joint Water Auth. v. Tonolli Corp., 4 F.3d 1209, 1221 (3d Cir. 1993); Anspec Co., Inc. v. Johnson Controls, Inc., 922 F.2d 1240, 1241-42 (6th Cir.1991) (reviewing this circuit’s approach to CERCLA liability). Accordingly, courts generally will not interpret § 9607(a) in a way that apparently frustrates the statute’s goals in the absence of specific congressional intent to the contrary. Anspec, 922 F.2d at 1247 (citing New York v. Shore Realty Corp., 759 F.2d 1032, 1045 (2d Cir.1985)).
*578It must be recognized, however, that it is difficult to divine the specific, as opposed to the general, goals of Congress with respect to CERCLA liability since the statute represents an eleventh hour compromise. See generally Shore Realty Corp., 759 F.2d at 1089-42 (discussing legislative history). As the district court recognized, “some of CERCLA’s provisions are vague and its legislative history sparse.” CPC Int’l v. Aerojet-General, 777 F.Supp. at 571; accord Anspec, 922 F.2d at 1247 (characterizing the legislative history as “scant”); Lansford-Coaldale Joint Water Auth., 4 F.3d at 1221 (“[Congressional intent may be particularly difficult to discern with precision in CERCLA, a statute notorious for its lack of clarity and poor draftsmanship.”).
Courts would not be warranted, therefore, in pointing to the “remedial legislation” litany, see generally Norman J. Singer, 3 Southerland Statutory Construction § 60.01 (5th ed.1992) (the rule that remedial statutes should be liberally construed is “firmly established”); Dennis v. Higgins, 498 U.S. 439, 443, 111 S.Ct. 865, 868, 112 L.Ed.2d 969 (1991) (noting that 42 U.S.C. § 1983, as a remedial statute, should be liberally construed), as a reason for filling in the blanks left by this sketchy legislative history to impose liability under nearly every conceivable scenario. Thus, while the liability provisions concerning facility operators should be construed so that financial responsibility for clean-up operations falls upon those entities that contributed to the environmental problem, the widest net possible ought not be cast in order to snare those who are either innocently or tangentially tied to the facility at issue. In fact, this court has pointed out that, “Congress intended that those responsible for disposal of chemical poisons bear the cost and responsibility for remedying the harmful conditions they created.” Anspec, 922 F.2d at 1247 (emphasis added).
In turning to the specific facts now before us, we adhere to the tenet that liability attaches only to those parties who are culpable in the sense that they, by some realistic measure, helped to create the harmful conditions.
In its effort to discern the sweep of CERCLA liability, the district court concluded that:
CERCLA broadens the potential for liability of parent corporations without discarding entirely the traditional concept of limited liability that is central to corporate law____
Accordingly, it seems that CERCLA’s “owned or operated” language forges a new, middle ground. It is a ground that at once accommodates the general principle of limited liability and the broader principle of liability attaching for operative activity. To permit these principles to coexist under CERCLA, the liability of a parent corporation cannot attach simply because a parent has had involvement with its subsidiary in a manner merely consistent with their investment relationship. Rather, a parent must have actually operated the business of its subsidiary.
In this court’s view, then, a parent corporation is directly liable under section 107(a)(2) as an operator only when it has exerted power or influence over its subsidiary by actively participating in and exercising control over the subsidiary’s business during a period of disposal of hazardous waste. A parent’s actual participation in and control over a subsidiary’s functions and decision-making creates “operator” liability under CERCLA; a parent’s mere oversight of a subsidiary’s business in a manner appropriate and consistent with the investment relationship between a parent and its wholly owned subsidiary does not.
CPC Int’l v. Aerojet-General, 777 F.Supp. at 573 (emphasis added).
On the basis of this “new, middle ground,” the district court found both CPC and Aerojet liable as operators for the disposal of hazardous substances that occurred while their subsidiaries operated the site.
III. DISCUSSION
A CPC
The district court reasoned that liability potentially could attach to CPC as a parent corporation in two ways: direct Lability un*579der CERCLA’s “operator” language or by common law veil-piercing. Clearly, since the facility was titled in the subsidiary’s name, CPC could be found liable as an owner only through veil piercing. The court determined that CPC was liable as an operator of the site for environmental damage that occurred during the ownership of Ott II; this liability was grounded in section 107(a)(2) of CERCLA, which renders “any person [liable] who at the time of disposal of any hazardous substance owned or operated any facility at which such hazardous substances were disposed of.” 42 U.S.C. § 9607(a)(2). To reach this conclusion, the court necessarily had to hold CPC, as a parent corporation, accountable for the environmental conduct of its wholly owned subsidiary corporation, Ott II. And, because the court held CPC directly liable as an operator, it did not reach the question of whether CPC was liable as an owner pursuant to the traditional common law doctrine of veil-piercing.
It is not at all clear from the district court’s opinion whether the basis for finding parental liability as an operator under its “new, middle ground” is the actual operation of the subsidiary’s business or, on the other hand, the exertion of power or influence through active participation in the subsidiary’s business. Although they are used interchangeably in the district court’s opinion, the two concepts are not interchangeable. If anything, the facts recited by the district court support liability under the latter standard but not under the former. This confusion underscores the inevitable difficulty that arises when courts attempt to erect new concepts of corporate liability within the framework of CERCLA in the absence of direction from Congress. We are not persuaded that, in enacting CERCLA, Congress contemplated the abandonment of traditional concepts of limited liability associated with the corporate form in favor of an undefined “new, middle ground.”
Actually, another scenario occurs to us under which one could argue that a parent corporation should be deemed to have directly operated a facility owned by its subsidiary. At least conceivably, a parent might independently operate the facility in the stead of its subsidiary; or, as a sort of joint venturer, actually operate the facility alongside its subsidiary. However, this is not a theory of operator liability relied upon by the district court, or alluded to in its opinion.
CERCLA defines the “owner or operator” of an onshore facility as “any person owning or operating such facility.” 42 U.S.C. § 9601(20)(A)(ii). When the facility has been conveyed to a unit of state or local government, the definition differs. It then includes “any person who owned, operated or otherwise controlled activities at such facility immediately [before the transfer to the governmental authority].” 42 U.S.C. § 9601(20)(A)(iii). It thus appears that the drafters of the statute distinguished an operator from a person who “otherwise controlled” a facility. When the owner of a facility contracts out the daily running of the operation to a third party, that party presumably attains operator status (and its attendant liability). However, when a parent corporation actively participates in the affairs of its subsidiary consistent with the restrictions imposed by traditional corporations law, nothing in the definition just cited or in the rest of the statute indicates that the parent has assumed the role of operator.
Despite the definition of “owner or operator,” several circuits, like the district court below, have determined that parent corporations can attain operator status by exerting significant control over the operations of their subsidiaries. See, e.g., Kayser-Roth, 910 F.2d at 26-27; Schiavone, 79 F.3d at 255; Lansford-Coaldale Joint Water Auth., 4 F.3d at 1221; Nurad, Inc. v. William E. Hooper & Sons Co., 966 F.2d 837, 842 (4th Cir.), cert. denied, 506 U.S. 940, 113 S.Ct. 377, 121 L.Ed.2d 288 (1992); CPC Int'l v. Aerojet-General, 777 F.Supp. at 572-73.
While some may wish to extend the reach of CERCLA to maximize the impact of its remedies, nothing in the statute or its legislative history warrants the invocation by courts of vague, expansive concepts, such as the district court’s “new, middle ground,” which threaten the efficacy of time-honored limited liability protections afforded by the corporate form. As the Court of Appeals for the Fifth Circuit has noted in this context, “[i]f Con*580gress wanted to extend liability to parent corporations it could have done so, and it remains free to do so.” Joslyn Mfg. Co. v. T.L. James & Co., Inc., 893 F.2d 80, 83 (5th Cir.1990).
The district court’s approach presents a number of problems. First, it replaces the relatively bright line provided by the traditional doctrine of piercing the corporate veil with a nebulous “control” test. When, precisely, is a parent acting in a manner consistent with its investment relationship as opposed to a manner that triggers operator liability? The indicia enumerated by the district court, such as participation in the subsidiary’s board of directors and involvement in specific policy decisions, offer little guidance. Certainly, these activities are not grounds traditionally relied upon as warranting the disregard of separate corporate existences.
Second, the threat of unlimited liability will likely deter private sector participation in the cleanup of existing sites. The ease before us illustrates this point. There is no dispute that the MDNR actively sought a private sector partner to take over and assist in the remediation of the site. Aerojet indicated an interest on the condition that it could cap its potential liability for environmental cleanup, which it sought to accomplish through the negotiation of the agreement with the MDNR and the use of subsidiaries.3 To scuttle such sensible and legitimate precautions in favor of an unpredictable “control” test would actually contravene the public interest by discouraging businesses from being involved in such projects.
Accordingly, we reject the district court’s “new, middle ground” as the basis for fixing operator liability and hold that where a parent, corporation is sought to be held liable as an operator pursuant to 42 U.S.C. § 9607(a)(2) based upon the extent of its control of its subsidiary which owns the facility, the parent will be liable only when the requirements necessary to pierce the corporate veil are met. In other words, under the circumstances of this case, whether the parent will be liable as an operator depends upon whether the degree to which it controls its subsidiary and the extent and manner of its involvement with the facility, amount to the abuse of the corporate form that will warrant piercing the corporate veil and disregarding the separate corporate entities of the parent and subsidiary.
Whether the circumstances in this case warrant a piercing of the corporate veil will be determined by state law. See Anspec, 922 F.2d at 1248. Michigan appears to follow the general rule that requires demonstration of patent abuse of the corporate form in order to pierce the corporate veil. There must be such a unity of interest and ownership that the separate personalities of the corporation and its owner cease to exist, and the circumstances must be such that adherence to the fiction of separate corporate existence would sanction a fraud or promote injustice. 1 William M. Fletcher, Fletcher Cyclopedia of the Law of Private Corporations § 41.30 (perm. ed. rev.vol.1990); Stephen H. Schulman et al., Michigan Corporation Law & Practice § 3.9(c) (1991 Supp.); Seasword, v. Hilti, Inc., 449 Mich. 542, 548, 537 N.W.2d 221, 224 (1995) (corporate veil may be pierced where the subsidiary is a “mere instrumentality” of the parent and the separate corporate existence is used to subvert justice or cause result contrary to clearly overriding public policy); see also Bodenhamer Bldg. Corp. v. Architectural Research Corp., 873 F.2d 109, 111-12 (6th Cir.1989) (surveying Michigan corporate veil-piercing decisions). Organization of a corporation for the avowed purpose of avoiding personal responsibility does not in itself constitute fraud or reprehensible conduct justifying a disregard of the corporate form. Gledhill v. Fisher & Co., 272 Mich. 353, 359, 262 N.W. 371, 373 (Mich.1935).4
*581The district court relied upon a number of factors in determining that CPC “actively participated in and exerted significant control over Ott IPs business and decision-making” and was therefore directly liable under 42 U.S.C. § 9607(a)(2) as an operator: 100% ownership of Ott II; participation on Ott IPs board of directors; a cross-pollination of officers who were involved in decision-making and daily operations; active participation by CPC officials in environmental matters; and financial control of Ott II through approval of budgets and capital expenditures. CPC Int’l v. Aerojet-General, 777 F.Supp. at 575. While these factors reveal a parent that took an active interest in the affairs of its subsidiary, they do not indicate such a degree of control that the separate personalities of the two corporations ceased to exist and that CPC utilized the corporate form to perpetrate the kind of fraud or other culpable conduct required before a court can pierce the veil. While CERCLA contemplates allocating financial responsibility to those corporations that cause environmental degradation, it does not authorize assignment of Lability to parent corporations that abide by the proper use of the corporate form.
In summary, then, it seems to us that under the “owned or operated” language of 42 U.S.C. § 9607(a)(2), there are three scenarios under which a parent corporation could be held Lable for the disposal of hazardous substances at a facihty whose owner of record was the parent’s subsidiary corporation. First, as an owner, by piercing the corporate veil. Second, as an operator, where the parent directly operates the faciLty itself, either independently of its subsidiary, or as an actual co-operator alongside the subsidiary.5 Although a parent conceivably could be held Lable under this theory, it is not the one reLed upon by the district court, and, in any event, is not supported by the facts in the record before us.
FinaUy, operator LabiLty may be based upon the conduct of the parent in the course of its affihation with its subsidiary, including the degree of control exerted by the parent over its subsidiary. This is the scenario utilized by the district court, relying upon its “new, middle ground” standard to define the circumstances under which the parent wiL be Lable. As pointed out above, we conclude that this “new, middle ground” is unworkable, and that traditional veil piercing is the only standard under which this scenario for LabiLty can be assessed reLably.
Accordingly, the district court’s finding of operator LabiLty with respect to CPC must be reversed.
B. ' MDNR Liability
The district court rejected LabiLty claims that were advanced against the MDNR on two fronts: as an operator and as an “arranger.” Only the district court’s decision regarding arranger LabiLty is appealed.
CERCLA imposes LabiLty on 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 faciLty ... owned or operated by another party or entity and containing such hazardous substances.
42 U.S.C. § 9607(a)(3).
It is contended that the MDNR incurred arranger LabiLty when it negotiated with Cordova/California for the acquisition of the site and agreed with Cordova/California on a plan to clean up the groundwater contamination.
We agree with the district court’s conclusion that the MDNR escapes LabiLty because its actions were taken in response to the Act’s environmental emergency provision:
No state or local government shaL be Lable under this subchapter for costs or *582damages as a result of actions taken in response to an emergency created by the release or threatened release of a hazardous substance generated by or from a facility owned by another person. This paragraph shall not preclude liability for costs or damages as a result of gross negligence or intentional misconduct by the State or local government. For the purpose of the preceding sentence, reckless, willful, or wanton misconduct shall constitute gross negligence.
42 U.S.C. § 9607(d)(2). Our reading of the record indicates that the MDNR acted in good faith when attempting to address the groundwater contamination of the site. That its efforts proved to be less than entirely successful is unfortunate, but does not subject the agency to liability.6
C. Liability of Aerojet and its Subsidiaries
1. Owner Liability
The district court held both Aerojet and Cordova/Michigan liable as present owners of the site under CERCLA’S section 107(a)(1) which assigns liability to “the owner and operator of a ... facility.” 42 U.S.C. § 9607(a)(1). Cordova/Michigan does not challenge this decision on appeal and we therefore express no opinion with respect to its liability.
As for the parent corporation, Aerojet could be found liable as an owner only through veil piercing, since Cordova/Michigan was the owner of record. The district court appropriately looked to Michigan law to determine whether to pierce the corporate veil and, in its view, the evidence supported a conclusion that Aerojet had “totally dominated Cordova/Michigan, creating a complete identity of interests between the parent and its wholly owned subsidiary.” CPC Int’l v. Aerojet-General, 777 F.Supp. at 578. Accordingly, the court determined that it was appropriate to pierce the corporate veil. Among the grounds cited for its decision, the court highlighted the total ownership by the parent, Aerojet’s active participation in the acquisition of the site, the timing of the incorporation of the subsidiaries, cross-pollination of corporate officers, financial control, and the integration of the businesses. Id. at 577.
These facts, however, fall somewhat short of what is required to pierce the corporate veil under Michigan law. They do not establish that Cordova/Michigan was a mere instrumentality of Aerojet in the sense that the separate corporate personalities of the parent and subsidiary ceased to exist. More tellingly, they do not reveal activity by Aerojet that approaches the level of culpable conduct contemplated by Michigan law as a predicate to disregarding the separate corporate form. Bodenhamer Bldg. Corp., 873 F.2d at 112. While Aerojet obviously sought to limit its liability for existing environmental problems through good faith negotiation with the MDNR and prudent use of the corporate form, there is nothing to suggest that the company acted to subvert justice or with fraudulent intent or otherwise sought to distort the legitimate purposes of the corporate form. Although Aerojet took an active interest in its subsidiaries, the record does not support a conclusion that Cordova/Michigan was not a viable corporate entity. Accordingly, the district court erred when it pierced the corporate veil to assign liability to Aerojet as an owner.
The district court also noted that Cordova/California actually owned the site from October 1977 until November 1978. Although the company began clean-up operations as required by its agreement with the MDNR, the district court found that additional releases of hazardous substances occurred during this period. CPC Int’l v. Aerojet-General, 777 F.Supp. at 579. It thus imposed liability on Cordova/California as a former owner pursuant to section 107(a)(2).
*583This conclusion, however, conflicts with the district court’s summary of activity at the site during the ownership of the Cordova companies. Id. at 556. On remand, therefore, we ask the district court to indicate with greater specificity precisely which portions of the record it relies upon to support a finding that additional releases of hazardous substances occurred during Cordova/California’s brief ownership. Absent such evidence, liability cannot attach to Cordova/California as a former owner of the site.
2. Operator Liability
Finally, we turn to the district court’s alternative imposition of liability on Aerojet pursuant to section 107(a)(2), precisely the same basis for liability that we have already discussed, and rejected, with respect to CPC. Since we decline to adopt the “new, middle ground” control test utilized by the district court, no liability attaches unless the corporate veil can be pierced. And because the record does not support veil piercing, Aerojet cannot be held liable as an operator pursuant to section 107(a)(2).
3. Defenses
Upon remand, the district court should also revisit its treatment of the defense raised by Aerojet, Cordova/California, and Cordova/Miehigan under section 107(b)(3). The pertinent part of the statute follows:
[A] person otherwise liable who can establish by a preponderance of the evidence that the release or threat of release of a hazardous substance and the damages resulting therefrom were caused solely by—
(3) an act or omission of a third party other than ... one whose act or omission occurs in connection with a contractual relationship, existing directly or indirectly, with the defendant ... if the defendant establishes by a preponderance of the evidence that (a) he exercised due care with respect to the hazardous substance concerned, ... and (b) he took precautions against foreseeable acts or omissions of any such third party and the consequences that could foresee-ably result from such acts or omissions ____
42 U.S.C. § 9607(b) (emphasis added).
In parsing the exceptions to the defense, the district court noted that under 42 U.S.C. § 9601(35)(A), the term “contractual relationship” includes deeds transferring title. Thus, the district court concluded that a defense would be unavailable to a defendant who had a direct or indirect contractual relationship with the parties responsible for contaminating the site. CPC Int’l v. Aerojet-General, 777 F.Supp. at 581. Under this view, the defense could not be invoked by any defendant who was a party to a deed with a polluter. The district court, however, ignored the requirement that, in order to render the defense inapplicable, the hazardous substance release must have resulted from the act of a third party “in connection with” the contractual relationship with the defendant. The “in connection with” language of the defense appears to have been designed to preclude a person from escaping liability by contracting for a third party to do his dirty work for him.
As we pointed out above, from what we glean from the district court’s recitation of facts, the release of hazardous substances appears to have been caused solely by the predecessors of these three defendants.
IV. CONCLUSION
For the foregoing reasons, the district court is reversed in part and affirmed in part and this cause is remanded to the district court for further proceedings consistent with this opinion.
. The district court noted that two hazardous chemicals — benzene and 1,2 dichloroethane— were found at the site and used during the Cordova period of ownership. Id. at 556, 579 n. 11. Although Cordova/Miehigan, in an April 5, 1984 letter to the federal EPA, acknowledges generating small quantities of these materials for disposal, we find nothing in the record to support a finding that any additional release of hazardous substances occurred at the site during the Cordova period.
. 42 U.S.C. § 9607(a) states, in part:
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 ... 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 ...
(4) any person who accepts or accepted any hazardous substances for transport to disposal or treatment facilities, ... 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____
. Although it does not affect our resolution of the liability issue, we note that these negotiations occurred before the enactment of CERCLA.
. We do not conclude, as suggested by the dissent, “[t]hat piercing the corporate veil under Michigan law requires showing that the corporate form was used to subvert justice.' ” Rather, it requires that the subsidiary be a “mere instrumentality” of the parent which is used either to subvert justice or to circumvent overriding public policy. In our view, none of these factors fairly *581characterize the relationship between CPC and Ott II.
. Thus, the dissent's characterization of our opinion as holding "that a parent cannot, as a matter of law, be held directly liable under 42 U.S.C. § 9607(a)(2) as an ‘operator’ of a facility owned by its subsidiary corporation” is mistaken.
. We note that, since this case was briefed and argued, the Supreme Court decided Seminole Tribe of Fla. v. Florida, -U.S.-, 116 S.Ct. 1114, 134 L.Ed.2d 252 (1996), which potentially provides the MDNR with an Eleventh Amendment defense. Because this issue has not been raised by the parties through supplemental briefs and the resolution of the case with respect to the MDNR is not affected, we choose not to consider the impact of Seminole Tribe at this point in the proceedings.