In Re Silicone Implant Insurance Coverage Litigation

667 N.W.2d 405 (2003)

In re SILICONE IMPLANT INSURANCE COVERAGE LITIGATION.

Nos. C5-01-1546, C3-01-1738, C3-01-1741, C5-01-1742, C4-01-1747, C6-01-1748, C4-01-1778, C9-01-1811, C0-01-1812, C1-01-1821, C4-01-1828, C6-01-1829, C8-01-1850, C2-01-1861, C4-01-1862, C7-01-1869, C6-01-1894, CX-01-1896, C1-01-1897, C9-01-1906; C3-01-1917, C3-01-1920, C7-01-1922, C2-01-1925, C3-01-1965.

Supreme Court of Minnesota.

August 21, 2003. Rehearing Denied September 29, 2003.

Thomas Tinkham, Marianne D. Short, Michael Jurgens, Dorsey & Whitney, LLP, Gary J. Haugen, David F. Herr, R. Lawrence Purdy, Maslon Edelman Borman & Brand LLP, Minneapolis, MN, Dale I. Larson, Douglas L. Skor, Keith A. Dotsetj, Larson King, LLP, John J. Ursu, Thomas A. Boardman, Carol A. Peterson, St. Paul, MN, for Appellant.

Thomas S. Fraser, Richard D. Snyder, Fredrikson & Byron, P.A., Minneapolis, MN, for Respondent Insurers.

Wayne D. Struble, Bowman and Brooke LLP, Minneapolis, MN, Hugh F. Young, Jr., The Product Liability Advisory Council, Inc., Reston, VA, Robert J. Brennan, Lauren E. Sweeney, Porzio, Bromberg & Newman, PC, Morristown, NJ, for Amicus Curiae Plac.

Leonard, Street & Deinard, Frederick W. Morris, Robert P. Thavis, Jeffrey A. Eyers, Minneapolis, MN, Gardere Wynne Sewell LLP, James J. Sentner, Jr., Stephen L. Moll, Houston, TX, for W. Haagman & Company.

McDermott, Will & Emery, Margaret H. Warner, Richard B. Rogers, Washington, DC, Brendel & Zinn, LTD., Sylvia Zinn, Burke Ellingson, St. Paul MN, for Columbia Casualty Company, The Continental Insurance Co., Harbor Insurance Company.

Semmes, Bowen & Semmes, Robert E. Scott, Jr., Paul N. Farquharson, Baltimore, MD, Bassford, Lockhart, Truesdell & Briggs, John M. Anderson, Charles E. Lundberg, Minneapolis, MN, for Federal Insurance Company.

Clausen, Miller, Amy Rich Paulus, Edward M. Kay, Melinda S. Kollross, Chicago, IL, Lind, Jensen, Sullivan & Peterson P.A., Ted E. Sullivan, William L. Davidson, Minneapolis, MN, for Old Republic Insurance Company.

Crowell & Moring LLP, Clifton S. Elgarten, Paul W. Kalish, Sarah C. Lindsey, Washington, DC, Fredrikson & Byron, P.A., Thomas S. Fraser, Richard D. Snyder, Minneapolis, MN, for Century Indemnity Company, Ace Insurance, sa/nv, United States Fire Insurance Company.

Luce, Forward, Hamilton & Scripps LLP, Mitchell L. Lathrop, Bridget K. Moorhead, San Diego, CA, Meahger & Geer, John J. McDonald, Jr., Minneapolis, MN, for Westport Insurance Corporation f/k/a Puritan Insurance Company f/k/a The Manhattan Fire & Marine Insurance Company.

Fedota Childers & May P.C., Jan Michaels, Chicago, IL, Lind, Jensen, Sullivan & Peterson P.A., Thomas D. Jensen, Minneapolis, MN, for The Home Insurance Company, Employers Mutual Casualty Co., Utica Mutual Insurance Company.

Litchfield Cavo, Alan Becker, Chicago, IL, Lind, Jensen, Sullivan & Peterson P.A., Thomas D. Jensen, Minneapolis, MN, for Mt. McKinley Insurance Company f/k/a Gibraltar Casualty Co., Everest Reinsurance Company f/k/a Prudential Reinsurance Company.

Quinlivan & Hughes P.A., Dyan J. Ebert, St. Cloud, MN, Morrison Mahoney & Miller, Michael F. Aylward, John T. Harding, Jr., Boston, MA, for Transamerica Premier Insurance Company n/k/a Tig Insurance Company.

Maslon Edelman Borman & Brand, LLP, Gary J. Haugen, David F. Herr, Dorsey & Whitney, LLP, Thomas Tinkham, Marianne D. Short, Minneapolis, MN, Larson & King, Dale Larson, Keith Dotseth, Douglas L. Skor, 3M Company, Office of the General Counsel, John J. Ursu, Thomas A. Boardman, Carol A. Peterson, St. Paul, MN, for Minnesota Mining & Manufacturing Co.

Hughes Hubbard & Reed, L.L.P. Norman C. Kleinberg, New York, NY, William R. Stein. Robert B. Funkhouser, M. Kathleen O'Connor, Washington, DC, McCollum, Crowley, Vehanen, Moschet & Miller, LTD., Robert L. McCollum, Bloomington, MN, for First State Insurance Company, Twin City Fire Insurance Company.

Morrison, Mahoney & Miller, Arthur Liederman, New York, NY, John T. Harding, Boston, MA, Abrams & Smith, Jerome B. Abrams, Esq., Minneapolis, MN, for European General f/k/a European General Reinsurance Company of Zurich, Swiss Re f/k/a Swiss Reinsurance Company, Zurich.

Bassford, Lockhart, Truesdell & Briggs, John M. Anderson, Charles E. Lundberg, Minneapolis, MN, for Victoria Feuervarsicherungs-Aktien-Gesellschaft n/k/a Victoria Versicherung, A.G., Executive Risk Indemnity.

Johnson & Condon, P.A., Dale O. Thornsjo, Minneapolis, MN, for Chicago Insurance Company, and Interstate Fire & Casualty Co.

Larson & King, Lawrence R. King, St. Paul, MN, for Lakeside Insurance, LTD. and Seaside Insurance, LTD.

Kelly & Berens, George O. Ludcke, Richard A. Kaplan, Minneapolis, MN, for Republic Western Insurance.

Bowman and Brooke LLP, Wayne D. Struble, Minneapolis, MN, for Amicus Curiae Product Liability Advisory Council, Inc.

Livgard & Rabuse, P.L.L.P., Dwight G. Rabuse, Minneapolis, MN, Wiley, Rein & Fielding LLP, Washington, D.C., for Amicus Curiae Complex Insurance Claims Litigation Association.

Birrell & Newmark, LTD., Andrew S. Birrell, R. Travis Snider, Minneapolis, MN, Kirkpatrick & Lockhart LLP, Donald E. Seymour, Paul E. Del Vecchio, Alexander W. Saksen, Pittsburg, PA, for Amicus Curiae Honeywell International Inc.

Rider, Bennet, Egan & Arundel, LLP, Eric J. Magnuson, Michael R. Docherty, Minneapolis, MN, for Amicus Curiae Insurance Federation of Minnesota, National Association of Independent Insurers.

Stephen K. Warch, Assistant Attorney General, Michael J. Tostengard, Assistant Attorney General, St. Paul, MN, for Amicus Curiae Minnesota Dept. of Commerce.

Yaeger, Jungbauer, Barczak & Vucinovich, PLC., Michael L. Weiner, Christopher J. Moreland, Minneapolis, MN, for Amicus Curiae Minnesota Trial Lawyers Association.

Heard, considered, and decided by the court en banc.

OPINION

ANDERSON, PAUL H., Justice.

This appeal stems from a declaratory judgment action brought by several of 3M's high-level, excess-layer, occurrence-based policy insurers. These insurers sought to clarify their coverage obligations in 3M's ongoing silicone gel breast implant mass tort litigation. The insurance policies at issue were in place from 1977 to 1985 and covered claims arising from injuries occurring during that time period. The implant claims for which 3M sought reimbursement were brought in the early 1990s, but were based largely on implantations that occurred during the policy periods, which implants allegedly caused various systemic autoimmune diseases. The Ramsey County District Court determined that the actual-injury trigger, for purposes of determining coverage liability, began at or around the time of implantation when silicone first leaked and came in contact with body tissue stimulating the immune system. The court found that the injury continued after implantation. The court then determined that 3M's losses should be allocated pro rata by time on the risk for the period from implantation through December 31, 1985, the end of the time period during which the policies were in place. The court also concluded that the insurers had breached their implied covenant of good faith and fair dealing and that 3M was entitled to reasonable attorney fees and costs as a result.

The court of appeals affirmed the district court's determinations that the alleged systemic autoimmune diseases constituted a continuing injury and that allocation was appropriate. The court, however, extended the end of the allocation period to the earlier date of the underlying plaintiff's claim or death. The court also reversed the award of attorney fees and costs, concluding that such a remedy is unavailable for this claim. Both sides to this dispute petitioned for and were granted review on the following issues: when and how policy coverage was triggered; whether allocation is appropriate and, if so, when the allocation period should end; whether the insurers are entitled to a judgment reduction; and whether attorney fees and costs are appropriate in this case. We affirm in part and reverse in part.

Between 1977 and 1985, 3M purchased significant amounts of occurrence-based insurance for product liability exposure. 3M purchased primary policies and ascending layers of excess coverage. The petitioner-insurers each provided high-level excess policy coverage, which means that their payment obligations arise only after judgments or settlements have exhausted the substantial primary and lower-level excess policies. Under these occurrence-based policies, coverage is determined by when the alleged bodily injury or property damage took place: all sums related to any such injury or damage that occurred during the policy period are covered by the policy, even if the claim is not asserted until after the end of the policy period.

In 1985, many manufacturers were forced to buy excess coverage in a new form—claims-made policies—which coverage is triggered by the date of the claim instead of the date the injury or damage occurred. These claims-made policies became the new form of excess coverage because product liability insurers no longer offered significant occurrence-based coverage. The claims-made policies were adopted primarily so that insurers could avoid the uncertainty often involved in occurrence-based policies under which insurers may not know the source or totality of their risks at the end of the policy period because claims can be made after expiration of the policy. Under a claims-made policy, insurers do not cover claims submitted after the end of the policy period, even if the injury underlying the claim arose during the policy period. The claims-made policies include a retroactive date that defines the earliest date the injury can have occurred in order for the policy to cover the resulting claim. The most significant difference between occurrence-based and claims-made policies is that occurrence-based policies can be triggered after the expiration of the policy period, while claims-made policies cannot. At the expiration of a claims-made policy, coverage available under the policy disappears.

3M's switch from occurrence-based to claims-made policies was designed to provide seamless coverage: the claims-made policies had retroactive dates that provided coverage immediately upon the expiration of the occurrence-based policies. Here, the district court found that there was no time during the relevant period that 3M was self-insured, uninsured, or, based on its claim history, underinsured. It found that 3M "sought to transfer its product liability risks to the maximum extent reasonably possible."

In 1992, 3M began to be named in thousands of complaints alleging that 3M's silicone gel breast implants caused various symptoms characteristic of a systemic autoimmune disease. The claims arose from injuries allegedly caused by breast implants manufactured by 3M between 1977 and 1984^[1] and implanted between 1977 and 1985. 3M notified its claims-made insurers of the breast implant litigation in April 1992. In July 1993, 3M sent notice of the litigation to its occurrence-based policy insurers for the 1977-1985 period. Though 3M defended the claims on the ground that the underlying tort plaintiffs were unable to prove that the implants caused any of the alleged injuries, it eventually settled a class action suit. The petitioner-insurers have stipulated to the reasonableness of the 1995 settlement. While the class action was settled, 3M has continued to defend against and settle claims with the plaintiffs who opted out of the class settlement.

On September 22, 1994, three of 3M's occurrence-based excess policy insurers commenced this declaratory judgment action against 3M and joined most of 3M's claims-made insurers and other occurrence-based excess insurers to clarify coverage issues raised in the breast implant litigation. The joined occurrence-based excess policy insurers aligned themselves with the other three occurrence-based insurers and became the petitioner-insurers. These petitioner-insurers first sought a declaration that they had no duty to defend, but eventually sought a resolution of the trigger and allocation issues and resolution of the exhaustion requirement. 3M counterclaimed on the declaratory issues, brought claims against the petitioner-insurers alleging breach of contract and breach of the implied covenant of good faith and fair dealing, and asserted statutory and tort claims that were dismissed on motion.

Two of the claims-made insurers that the petitioner-insurers joined in the action, XL Capital Ltd. (XL) and A.C.E. Insurance Co., Ltd. (A.C.E.), moved for their dismissal based on an undertaking 3M had agreed to make in its claims-made policies with these insurers. The undertaking asserted that 3M would agree to a reduction in any judgment against petitioner-insurers if the court determined that XL and A.C.E. shared any common liability with petitioner-insurers. The undertaking thereby protected petitioner-insurers from having to pay losses properly allocated to XL and A.C.E. Agreeing that the undertaking rendered the petitioner-insurers' claims against XL and A.C.E. moot, the district court dismissed XL and A.C.E. from the action in June 1995.

The district court bifurcated the parties' claims into declaratory or nonjury issues and jury issues. In 1996, 1997, and 1999, the court conducted a series of hearings on a number of discrete issues. The first of these nonjury issues tried via bench trial was the question of when coverage was "triggered," and, if triggered, how 3M's losses should be allocated among the insurers. Both sides to the dispute moved for summary judgment on the issue of when actual injury occurred, but the court denied the motions and, in June and July 1996, held a medical trigger bench trial. On July 11, 1996, the court granted 3M partial declaratory judgment on this issue, stating the following:

1. Actual-injury occurs at or about the time silicone-gel breast prostheses are implanted in the body, and insurance coverage is "triggered" at that time.

2. Coverage is triggered continuously for all policies in effect at the time of implant, at the time of manifestation of systemic disease symptoms, and at all times in between those events.

The district court was later called upon to clarify its coverage trigger ruling, and it changed the end date of damages from "the time of manifestation of systemic disease symptoms" to "the earlier of the implant recipient's death, or the date on which the recipient files a lawsuit for damages." It also addressed allocation for the first time and stated that it would apply the "pro rata by time on the risk allocation method," which it later defined as follows: "An individual insurer's share of the damages is determined by multiplying the settlement or judgment amount by a fraction that has as its denominator the entire number of years of the claimant's injury, and as its numerator the number of years within the period when the policy was in effect."

In a later order, on July 14, 1997, the district court sua sponte vacated its coverage trigger and allocation order in light of Domtar, Inc. v. Niagara Fire Ins. Co., 563 N.W.2d 724 (Minn.1997), which case addressed the appropriate allocation method for environmental damage liabilities. The court determined that its earlier ruling that insurance was triggered continuously throughout the period between implantation and suit was inconsistent with the holding in Domtar, because "[w]here * * * a single, discrete occurrence can be identified, the continuous trigger has no applicability." Here, the court stated, the discrete occurrence was the cellular damage caused by the implantation, occurring at or about the time of implant, and "the cellular distortion occurring between implant and encapsulation, constituted the initial, original and precipitating injury." Accordingly, the new finding of the court was that "the implant when inserted into the body is the occurrence that triggers coverage."

Four months later, after receiving motions and memoranda from the parties, the district court again reversed itself and reinstated its earlier "continuous trigger" rulings. It did so because it concluded that its earlier "sua sponte" decision "was incorrect as a matter of fact and law," because the injury at issue is not one injury with continuous leakage, but a consistently recurring injury that takes place each time silicone comes in contact with new cells, creating a new bioreaction. The court went on to state, "It is impossible to tell when any particular injury occurs"; therefore, it deemed allocation to be appropriate.

In January 1998, the judge originally assigned to this case was appointed to another court and was replaced by a colleague from the same judicial district. 3M then moved to have the second judge clarify the allocation rulings, and the judge granted the motion. The second judge vacated the earlier orders as they related to allocation, and invited arguments as to the appropriateness of allocation after 1985. The second judge did not revisit the issue of whether allocation was appropriate in this case, but asserted that he did not believe the law required allocation. Nevertheless, the second judge declined to revisit this issue because the first judge already decided allocation was appropriate and because the second judge could "not say that [the first judge's] conclusion was palpably wrong." Accordingly, the second judge limited his inquiry to determining the appropriate allocation period and established the end date as December 31, 1985, the last day of the last year during which 3M could have purchased occurrence-based insurance. The court decided that allocating losses beyond 1985 for silicone implanted between 1977 and 1985 improperly shifted to 3M losses it had paid its occurrence-based insurers to cover, and would therefore be inequitable.

The district court next conducted a four-month jury trial on 3M's breach of contract claims and the petitioner-insurers' defenses. The first phase of the trial addressed the petitioner-insurers' coverage defenses, which the jury ultimately rejected. The second phase addressed 3M's breach claims. In January 2000, the jury returned its special verdict, which provided the following: 3M's notice to its occurrence-based insurers on July 28, 1993 was not unreasonably late; from July 28, 1993 through June 2, 1994, 3M did not fail to materially and substantially cooperate with the occurrence-based insurers; 3M made no misrepresentations with the intent to deceive and defraud the occurrence-based insurers; and any petitioner-insurers asserting a misrepresentation defense waived any prior misrepresentation defenses by continuing to insure 3M. 3M presented its case on the breach claims, and the court granted the petitioner-insurers a directed verdict on 3M's breach of contract, anticipatory breach, and breach of the implied covenants of good faith and fair dealing claims because 3M "failed to create a jury question for legally recognized consequential contract damages."

In its special verdict, the jury rejected all of the petitioner-insurers' defenses to liability, and the petitioner-insurers then moved to reduce the judgment against them by enforcing the undertaking 3M had made with the previously dismissed claims-made insurers XL and A.C.E. The district court determined that because XL and A.C.E. are not liable for damages arising from implants made before their coverage obligations began, these insurers have no common liability with the petitioner-insurers and allocating some of the judgment to them is inappropriate.

In September 2000, the district court awarded attorney fees and costs to 3M. The court apportioned liability for those costs among all the insurers that had been involved in the litigation, and appointed a referee to determine a reasonable amount. The court awarded attorney fees to 3M based on its findings that the petitioner-insurers breached the covenant of good faith and fair dealing. In a later order, after it was asked to address how the fees should be apportioned among insurers, the court determined that the petitioner-insurers acted individually and could not, therefore, be held jointly and severally liable as 3M requested. The court ordered apportionment of 3M's fees and costs among most of the insurers, including insolvent carriers and those carriers with whom 3M had already settled, and allocated the liability in proportion to policy limits.

Partial final judgment was entered in May 2001 and it included judgments totaling $169,340,679 against most of the remaining petitioner-insurers. The judgment was broken down as follows: $123,030,541 in indemnity costs; $22,456,112 in defense costs; and $23,854,026 in prejudgment interest. This amount did not include the reasonable attorney fees and costs awarded by the district court.

Several petitioner-insurers filed motions to amend the findings or grant judgment notwithstanding the verdict, all of which were denied. 3M also moved for amended findings, or alternatively a new trial on a number of grounds, and its motion was denied.^[2] 3M and the petitioner-insurers each appealed various orders of the district court. The court of appeals made the following rulings relevant to the issues before us: (1) it affirmed the ruling that the insurance policies are triggered by injuries occurring around the time of implantation; (2) it affirmed the ruling that 3M's losses be allocated "pro rata by time on the risk"; (3) it reversed the ruling that the allocation period ended on December 31, 1985 and determined instead that the allocation period should not end until the point at which the claim is filed or the plaintiff dies; and (4) it reversed the award of attorney fees and costs, stating that such a remedy is unavailable because Minnesota's narrow exception to the American rule is limited to "breach of contract by failure to assume the duty to defend." In re Silicone Implant Ins. Coverage Lit., 652 N.W.2d 46 (Minn.App.2002).

3M and the petitioner-insurers petitioned for, and we granted, review of the following issues: (1) when policy coverage was triggered; (2) whether allocation is appropriate in this case; (3) if allocation is appropriate, when the allocation period should end; (4) whether the petitioner-insurers are entitled to a reduction of judgment in recognition of 3M's undertaking with XL and A.C.E.; and (5) whether attorney fees were properly awarded to 3M.

I.

We first address the petitioner-insurers' (insurers) argument that the lower courts erred in concluding that the insurance policies at issue were triggered shortly after implantation of 3M's silicone gel breast implants. The district court held a bench trial to determine when injury arose for purposes of triggering the occurrence-based policies. While the court recognized that "solid conventional science establishes no causal connection whatsoever between silicone gel breast implants and systemic disease," the court assumed legal causation for purposes of this insurance coverage determination. After hearing expert testimony from both sides, the court found that silicone introduced into the body through a breast implant caused a woman's immune system to respond by encapsulating the implant with body tissue to wall the implant off from the rest of the body. The result of this process is chronic inflammation. Prior to the complete encapsulation of the implant, which can take up to 90 days, micro-droplets of silicone leak into the body tissues. A slight migration of the leaked silicone likely occurs during this period as well. The court found that these results are "normal immune responses," and therefore the foregoing events do not constitute the bodily injury necessary to trigger the policies.

Two competing theories exist to explain what takes place after encapsulation. The first hypothesis, accepted by both 3M and the insurers, is that after capsule formation the body is immunologically normalized with respect to the implant and no systemic disease results. The second hypothesis, espoused by the underlying tort plaintiffs (plaintiffs) and by some physicians and scientists, is that an abnormal response occurs, principally an autoimmune response, and that systemic disease symptoms appear. Because the plaintiffs have systemic disease symptoms and legal causation is assumed, the district court adopted the plaintiffs' theory "as being more likely true than not true."

When addressing the assumption that silicone causes disease, experts for both sides agreed that the immune system reacts immediately to the presence of a foreign substance in the body, but the experts offered conflicting opinions regarding the timing of the initial injury. One of 3M's experts hypothesized, "if the implants cause disease, there's only one time at which that injury could start, and that is when the implant goes in." In contrast, an expert for the insurers testified, "if I have to work on the assumption that silicone is a causal factor, then * * * actual injury likely precedes first manifestations [of disease symptoms] by a short period of time."

The district court ultimately found the following:

Leaked silicone is in contact with body tissues from the time of implant until the formation of the protective capsule, a period of several weeks. Silicone is bioreactive during that period and more likely than not that is the period during which cellular abnormality is produced. Thus, bodily injury within the purview of the trigger language occurs at or about the time of implant.

The greater weight of the evidence, in the context of the undisputed fact of systemic disease symptoms and the assumed fact of legal causation and the necessary inference of the occurrence of an abnormality, supports the conclusion that the leaking silicone gel is the cause, the cellular damage is the injury, and the disease symptoms are the effects. Such cellular damage is determinable, constitutes the underlying bodily harm without which there would be no manifestation in the form of disease symptoms, and satisfies the "actual injury" legal standard for trigger.

The district court then employed a "continuous trigger" standard, concluding that "all policies are triggered if they were in effect at the time of implant or at the time of manifestation of symptoms or at any time between those events." After stating that Minnesota follows an "actual-injury" theory to determine when policies have been triggered, the court nonetheless concluded that a "continuous trigger" theory is applicable to this case. Under the "continuous trigger" theory, "the policies in effect at the time of exposure, the time of manifestation, and all the time in between are triggered." N. States Power Co. v. Fidelity and Casualty Co. of N.Y., 523 N.W.2d 657, 662 (Minn.1994) (NSP).

The district court erred in finding that a continuous trigger rule applies. Minnesota follows an "injury-in-fact" or "actual-injury" rule and has explicitly rejected the continuous trigger rule. Id. The court also found that the silicone caused a continuous injury. This was not a finding based on a legal definition of trigger, but was based on the experts' testimony about the nature of autoimmune disease. However, the court appears to have equated a "continuous trigger" with a "continuous injury," which is inaccurate. A trigger is the legal event that activates the insured's policy, while a continuous injury is a factual finding that is based on medical testimony. Thus, we conclude the district court erred as a matter of law in finding that the "continuous trigger" rule applies and therefore erred in concluding that coverage is "triggered continuously for all policies in effect at the time of implant, at the time of manifestation of systemic disease symptoms, and at all times between those events."

Approximately two weeks later, in July 1996, the district court clarified its "trigger ruling," noting that its prior rulings were based on the "propositions that the systemic diseases in question are insidious and progressive and that they are characterized by continuously occurring injuries." The court of appeals held that the district court did not err in finding that "injury occurs on a cellular basis shortly after implant." In re Silicone Implant Ins. Coverage Lit., 652 N.W.2d at 59.

A district court's determination of the timing of an underlying plaintiff's injury is a question of fact. "Findings of fact, * * * shall not be set aside unless clearly erroneous, and due regard shall be given to the opportunity of the trial court to judge the credibility of the witnesses." Minn. R. Civ. P. 52.01; see also Farmers Ins. Group v. Hastings, 366 N.W.2d 293, 294 (Minn.1985) (holding that the district court's determination that a "blow was delivered intentionally," for purposes of a homeowner's insurance policy is reviewable under a clearly erroneous standard). Such factual determinations will, therefore, be upheld unless clearly erroneous and we will reverse the district court's findings regarding the timing of the injury for purposes of triggering the policies only if those findings are clearly erroneous.

We begin our analysis with the policy language. The insurers provided 3M with occurrence-based excess liability coverage. The policies at issue indemnify 3M "for all sums which the insured shall become legally obligated to pay as damages because of injury or damage to which this policy applies." The policies apply to "injury or damage arising out of: bodily injury or property damage caused by an occurrence."^[3]

Both sides agree that the actual-injury trigger rule is the proper method of determining which policies are activated by an occurrence. We adopted the "actual-injury" or "injury-in-fact" trigger in NSP. Under such a rule, "the time of the occurrence is not the time the wrongful act was committed but the time the complaining party was actually damaged." Singsaas v. Diederich, 307 Minn. 153, 156, 238 N.W.2d 878, 880 (1976). Thus, under the actual-injury trigger rule, only those policies in effect when the bodily injury or property damage occurred are triggered. NSP, 523 N.W.2d at 662; see also Jenoff, Inc. v. N.H. Ins. Co., 558 N.W.2d 260, 261 (Minn.1997); Fairview Hosp. and Health Care Serv. v. St. Paul Fire & Marine Ins. Co., 535 N.W.2d 337, 341 (Minn.1995); Singsaas, 307 Minn. at 155, 238 N.W.2d at 879-81.

To trigger a policy, "the insured must show that some damage occurred during the policy period." NSP, 523 N.W.2d at 663. For purposes of the actual-injury trigger theory, an injury can occur even though the injury is not "diagnosable," "compensable," or manifest during the policy period as long as it can be determined, even retroactively, that some injury did occur during the policy period. Am. Home Prods. Corp. v. Liberty Mut. Ins. Co., 748 F.2d 760, 765-66 (2d Cir. 1984).

3M argues that the district court's conclusion that the policies were triggered "at or about the time of implant" was based on substantial evidence in the form of expert testimony and, therefore, is not clearly erroneous. In response, the insurers argue that the court erred as a matter of law because it found that nonexistent cellular injuries triggered the policies in question. The insurers reason that the cellular injuries the court found are "fictional" and therefore those injuries cannot trigger the policies because no injury actually took place during the policy period. They contend that in using a nonexistent cellular injury to calculate when the policies were triggered, the court dispensed with the actual-injury trigger theory and erred as a matter of law. Thus, the insurers urge us to conclude that, here, the only policy coverage triggers are the demonstrable manifestations of the plaintiffs' autoimmune disease.^[4]

The insurers' argument that a "fictional" cellular injury cannot trigger coverage is unpersuasive. We acknowledge that conceptually the plaintiffs' injuries are fictional because medical science has failed to establish a causal connection between silicone gel breast implants and systemic autoimmune disease. Nevertheless, we conclude for purposes of this insurance coverage litigation, which arises from a settlement that is premised on the notion that silicone gel breast implants caused plaintiffs' injuries, that the district court properly assumed the plaintiffs' injuries are authentic and were caused by the silicone gel breast implants. This assumption, on which the district court's finding of cellular injury is based, was necessary to determine when the policies were triggered. The experts agreed that if silicone caused autoimmune disease, then cellular injuries would occur sometime before the symptoms appeared. By arguing that the cellular injury is "fictional" or "nonexistent," the insurers, in effect, are arguing that the plaintiffs were not injured. But, as the insurers concede, they cannot at this post-settlement stage of the proceedings deny coverage for the losses 3M incurred in its settlement on the theory that the silicone did not injure the plaintiffs.

Having dismissed the insurers' argument that "fictional" cellular injury cannot trigger coverage, we now must determine whether the district court's finding that the injuries occurred on or about the time of implant is clearly erroneous. The district court accurately concluded that Minnesota follows an "actual injury" trigger rule. The court then heard expert medical testimony to determine the timing of the plaintiffs' bodily injuries. Experts for 3M testified that injury occurs at or about the time of implant, while the insurers' experts urged the court to find that injury occurs shortly before manifestation of symptoms. When there is conflicting medical testimony, we give deference to the fact finder. See Raze v. Mueller, 587 N.W.2d 645, 648 (Minn.1999) (stating that we give deference to the jury's verdict when there is conflicting medical testimony); Gaspers v. Minneapolis Elec. Steel Castings Co., 290 N.W.2d 743, 745 (Minn. 1980) (stating that when the opinions of medical experts conflict, the function of the trier of fact is to resolve the conflict). Here, the court weighed the conflicting expert medical testimony and after doing so determined that "bodily injury" occurs at the time of implant. Based upon our review of the record, we conclude that the district court's determination of when bodily injury occurred is not clearly erroneous.

We turn next to the insurers' argument that the district court dispensed with the actual-injury trigger rule when it determined that the policies were triggered at or about the time of implantation. After reviewing conflicting medical testimony, the court found that the damage done to the plaintiffs occurred on a cellular level, "at or about the time of implant," years before the plaintiffs began experiencing symptoms of systemic disease. This finding is consistent with the actual-injury trigger rule, which requires that bodily injury occur during the policy period, but does not require that the injury be diagnosable or even evident during the policy period. Am. Home Prods., 748 F.2d at 765-66. In American Home Products, the court stated:

For example, a person may suffer an injury or illness that does not become diagnosable until after some period of gestation; it may be possible after diagnosis to infer that the harm must have begun some time prior to diagnosability because of the stage of the illness at the time it is diagnosed and the fact that the type of illness that is diagnosed does not occur without a gestation period.

Id. at 765. We conclude that the district court did not dispense with the actual-injury trigger rule and that the court's finding that the plaintiffs' injuries occurred at or about the time of implantation is not clearly erroneous. Accordingly, because damage occurred at or about the time of implantation, we conclude that the policies were triggered at or about the time of implantation.

II.

Having concluded that the insurance policies were triggered at or about the time of implantation, the next issue we must decide is whether the district court erred in deciding to allocate 3M's losses from those injuries among the insurers. 3M asserts that the decision to apply allocation is purely legal and therefore reviewable de novo. We have stated, however, that damages are very fact-dependent, so "trial courts must be given the flexibility to apportion them in a manner befitting each case." NSP, 523 N.W.2d at 663. Such language indicates that allocation decisions should be reviewed under an abuse of discretion standard and that is the standard we apply here.

The district court found that from the time of implantation, the damages were continuous and the "actual injury" continued to occur as silicone came in contact with new cells. The court therefore determined that 3M's losses should be allocated pro rata by time on the risk among all triggered policies. The court of appeals affirmed the allocation ruling, concluding that the district court had made specific findings that "the injury causing event was the continuous leakage of silicone that comes into contact with the body's cells." Based on this continuous injury finding, the court of appeals affirmed the need for allocation.

All parties and both lower courts recognized that the most apposite case law for the allocation issue consists of three of our earlier cases addressing environmental damage liability: NSP, 523 N.W.2d 657; SCSC Corp. v. Allied Mut. Ins. Co., 536 N.W.2d 305 (Minn.1995); and Domtar, 563 N.W.2d 724. In these three cases, we discussed the pro rata by time on the risk allocation method and how it applies to continuous injuries arising from environmental contamination.

In NSP, the insured, NSP, was required by the Minnesota Pollution Control Agency (MPCA) to clean up two adjacent property sites that were contaminated by NSP's use of the properties as coal-tar gasification sites. NSP, 523 N.W.2d at 658. The MPCA discovered that as a result of NSP's operations on those sites between 1910 and 1933, the groundwater at both sites was contaminated with coal tars and spent oxide waste. The MPCA required NSP to pay clean-up and monitoring costs for the sites. Id. at 659. NSP then sought coverage for its costs from its comprehensive general liability insurers.

We discussed in NSP the special problems associated with environmental liability insurance cases, where damages are continuous and where "for all practical purposes the bodily injury or property damage suffered during different policy periods is indivisible." Id. at 663 (quoting Kenneth S. Abraham, Environmental Liability Insurance Law: An Analysis of Toxic Tort and Hazardous Waste Insurance Coverage Issues, 120 (1991)). We noted that determining how to allocate damages in such cases "may require a more flexible approach. As with all insurance contract-related issues, courts must consider many factors when deciding this issue, including the policy language, parties' intent or reasonable expectations, canons of construction and public policy." Id. at 661. We went on to state that a pro rata by limits approach to allocation, as advocated by NSP, is inconsistent with the actual-injury trigger theory that we adopted. Id. at 662. In so doing, we stated that the goal of the actual-injury trigger theory is to ensure that insurers are not made liable for injuries occurring outside of their policy periods. We said:

Where the policy periods do not overlap, therefore, the insurers are consecutively, not concurrently liable. A "pro rata by limits" allocation method effectively makes those insurers with higher limits liable for damages incurred outside their policy periods and is therefore inconsistent with the actual injury trigger theory.

Id.

In NSP, we chose the time on the risk allocation method because it has the advantage of being a "more or less per se rule." Id. at 663. "This method assumes that the damages in a contamination case are evenly distributed (or continuous) through each policy period from the first point at which damages occurred to the time of discovery, cleanup or whenever the last triggered policy period ended." Id. Because the contamination in NSP was "regarded as a continuous process in which the property damage is evenly distributed over the period of time from the first contamination to the end of the last triggered policy (or self-insured) period," there was no period during which more or less damage occurred, so allocation according to time on the risk was appropriate. Id. at 664.

One year later, in SCSC, we revisited the issue of allocation in the environmental liability context. SCSC, 536 N.W.2d at 305. In SCSC, a dry cleaning and laundry supply distribution facility purchased, stored, repackaged, and distributed perchloroethylene (PCE), which the MPCA had identified as a volatile organic compound. Id. at 308. SCSC stored the PCE in two above-ground tanks from which the PCE was dispensed through an outgoing fill pipe to the trucks that delivered the chemical to retailers. Id. at 309. The MPCA discovered PCE contamination in the groundwater near the SCSC plant and SCSC was required to pay clean-up costs, for which it sought reimbursement from its general liability insurers. Id. at 309-10. At trial, SCSC alleged that the contamination was not the result of consistent dripping of PCE from the fill pipe during the normal course of operations, as was asserted by the insurers, but was instead the result of a significant spill that occurred in August 1977. The jury found that "property damage arose in August 1977, as the result of an unintended, unexpected, sudden and accidental event, and that the damage was neither divisible nor attributable to an overriding cause." Id. at 310. The district court adopted a "vertical triggering" approach by which the primary policies for 1977 were first in line for coverage and paid out in full, then the excess policies for 1977, then the primary policies for the next year, and so on until the insurers' full liabilities were paid. Id. at 317.

On appeal, we reversed the district court's vertical triggering scheme and also refused to allocate pro rata by time on the risk as we did in NSP. We did so because "NSP was an equitable decision based upon the complexity of proving in which policy periods covered property damage arose," and in SCSC no such complexity existed as a result of the jury having determined the damage arose from a single event in 1977. Id. at 318. We noted that the jury found the damage was not divisible: "the only covered `occurrence' was the 1977 spill. The continual leaching of the chemicals from the soil into the groundwater did result in damages to SCSC because of property damage," but that damage is not covered by insurers that were not on the risk in 1977, the year during which the only covered "occurrence" occurred. Id. at 318. We refused to allocate any damages to insurers that were not on the risk in 1977. All the damages from continued leaching potentially could have been covered if there were enough insurance coverage in the 1977 policies, but SCSC could not look to insurers from later years to help cover that liability.

Finally, in Domtar, the insured sought reimbursement for clean-up costs it incurred in association with its tar refining plant. Domtar, 563 N.W.2d at 728. The plant was operated from 1924 to 1929 and from 1934 to 1948, and it was dismantled in 1954 or 1955. Id. Pollution was first detected in 1979. The property where the plant was located was subsequently declared a Superfund site and the MPCA named Domtar as one of the responsible parties. Id. at 729. Domtar sought declaratory judgment of liability for reimbursement for clean-up costs against its 1956-1970 insurers. It limited its claim to insurers from this time period because earlier policies had been lost and later insurers were dismissed from the action. Id. The record indicated two general causes for Domtar's share of damages: (1) damage was caused by leaks during routine waste-handling and accidental spillage during plant operation; and (2) the bulk of the damage arose from residual sludge discharges from the storage tanks during dismantling of the plant before the property was sold. Id. All pollutants were discharged before Domtar sold the property and before the 1956-1970 insurers sold policies to Domtar. Experts for both sides agreed that the contamination could not be apportioned among causes because leakage to the groundwater had become commingled with and inseparable from other migrating contaminants. Id. at 730. Domtar asserted that the damage continued after the plant was dismantled because "contaminants were migrating deeper into the soil and through the groundwater during the ensuing years, including the present time * * * [and] `property damage at the site was indivisible * * * [and] it continued and expanded' over the years." Id.

In contrast to Domtar's assertions, experts for the insurance companies testified that contamination occurred in the years following the initial spills and leakage, and "the contamination has been ameliorated by biodegradation in the ensuing years," during which time period the 1956-1970 insurers provided coverage. Id. The jury determined that the property damage commenced in 1933 and that "some" property damage took place during each of the insurers' policy periods, rejecting insurers' argument that no "appreciable" damage occurred during their policy periods. Id. The district court determined that liability costs would be allocated evenly from 1933 to the year in which clean-up efforts began and that Domtar would be responsible for the costs outside of the insurers' policy periods, i.e., before 1956 and after 1970. Id. Domtar appealed the allocation ruling and the court of appeals affirmed. Id. at 730-31.

In Domtar, we summarized NSP as establishing that in "continuous and indivisible environmental contamination cases" (1) general liability policies are triggered when property damage occurred during the policy period; (2) insurer liability is consecutive, limited to property damage occurring during the insurer's policy period; and (3) one way to allocate loss among consecutively liable insurers, in the absence of applicable policy language, is pro rata by time on the risk. Id. at 732 (emphasis added). We then discussed the shifting of burdens in cases like these.

[T]he insured bears the burden of proving that a policy has been triggered, but if the insured proves when the contamination began and when it ended or was discovered, then the trial court should presume that property damage was continuous from its initiation until the time of clean up or discovery. The burden of proof then shifts to any party seeking to demonstrate that no appreciable damage occurred during a particular time period. All policies in effect when damages occurred are triggered, and liability is allocated to each policy according to the proportion of time each was on the risk.

Id. Accordingly, in Domtar, we affirmed the use of allocation pro rata by time on the risk, specifically rejecting Domtar's argument that this allocation method unfairly allocated losses to Domtar by allocating losses to periods during which Domtar was uninsured, self-insured, or underinsured. Id. at 732-33. We also emphasized the limits of our holding, however, and attempted to remedy some of the confusion created by our discussion of allocation in NSP and SCSC:

The proper scope of coverage also will depend on the facts of the case. When environmental contamination arises from discrete and identifiable events, then the actual-injury trigger theory allows those policies on the risk at the point of initial contamination to pay for all property damage that follows. [citing SCSC] * * * It is only in those difficult cases in which property damage is both continuous and so intermingled as to be practically indivisible that NSP properly applies. NSP provides a judicially manageable way for trial courts to adjudicate certain pollution-coverage disputes when it is difficult to determine when an "event" or "occurrence" or "damage" giving rise to legal liability has occurred. NSP does not establish hard-and-fast rules; it offers a practical solution in the face of uncertainty.

Id. at 733-34.

Domtar established guidelines for allocating losses from a continuing injury, like the immune diseases at issue here, using an injury-in-fact approach. The first, and most obvious, is that only insurance policies that are appropriately "triggered" are on the risk. Therefore, before an allocation discussion can occur, the district court needs to identify the triggered policies among which to allocate. The second, and most helpful guideline in this case, is that when there is a continuing injury that "arises from discrete and identifiable events, then the actual-injury trigger theory allows those policies on the risk at the point of initial contamination to pay for all property damage that follows." Id. at 733. In other words, the issue of allocation should be raised only if the triggering injury does not "arise[ ] from discrete and identifiable events."

In determining whether the district court erred in choosing to allocate 3M's losses, we follow the analytical progression provided in Domtar. First, we determine whether the plaintiffs' injuries are continuous. If they are not, under the actual-injury trigger theory, the policies on the risk at the time of the injury would pay all losses arising from that injury. Here, the court found that the injuries are continuous, so we move to the next determination: whether the continuous injury arose from some discrete and identifiable event. If it does, the policies on the risk at the time of that event are liable for all sums arising from the event. If not, allocation may be appropriate.

It is at this point in the dispute that the two sides diverge in their allocation analysis. Relying on the analytical framework from Domtar, 3M asserts that the time of implantation of the silicone gel breast implant is the discrete and identifiable event that the district court labeled as the onset of the continuing injury, so allocation among the triggered policies is not appropriate. Instead, 3M asserts that any policy in place at the time of implant is liable up to the limits of the policy for all sums paid in settlement of injuries allegedly arising from that implantation. This is an application of the classic actual-injury trigger rule applied to a continuing injury whose origin can be clearly established. Unlike in Domtar, where there was "agreement that the contamination could not be apportioned among causes," Domtar, 563 N.W.2d at 730, 3M asserts that here the cause is clear and it is akin to SCSC, in which no allocation was applied because the continuous leaching of chemicals was attributable to a discrete and identifiable 1977 chemical spill. SCSC, 536 N.W.2d at 318. This result, 3M argues, will advance the general principle underlying the actual-injury trigger rule, which is to allow policies on the risk at the point of initial contamination to pay for the resulting property damage. See Domtar, 563 N.W.2d at 733; SCSC, 536 N.W.2d at 318.

In contrast, the insurers argue that the continuing injuries at issue here are equally as difficult to define and assign to specific time periods as are the damages involved in environmental contamination cases. The insurers cite the court of appeals' observation that "[t]he district court specifically found that putting an implant in the body was not the injury or the injury-causing event. Instead, the court found that the injury-causing event was the continuous leakage of silicone that comes into contact with the body's cells, causing incremental cellular damage and eventually producing disease." In re Silicone Implant Ins. Coverage Lit., 652 N.W.2d at 60. Unlike in SCSC where there was "a single event of" spillage of a contaminant that for some time afterward leached into and damaged the soil, SCSC, 536 N.W.2d at 318, the insurers argue that with silicone breast implants, "[a]s cells later come into contact with the silicone and provoke an autoimmune response, new cell distortions, and hence new injuries, occur." The insurers assert that the underlying rationale for apportioning loss that we have used in environmental cases applies here. Policies are designed to cover injuries from a certain time period, and the insurers claim the pro rata by time on the risk method achieves this result without forcing insureds to specifically prove how much damage took place during a specific policy period.

We find 3M's arguments to be more consistent with our analysis in Domtar and the district court's findings. In our actual-injury trigger framework, allocation is meant to be the exception and not the rule because "[i]t is only in those difficult cases" that allocation is appropriate. Domtar, 563 N.W.2d at 733. If we can identify a discrete originating event that allows us to avoid allocation, we should do so. Here, the district court labeled the time of implant as the beginning of the continuing injury process. The implantation, therefore, is a readily identifiable discrete event from which all of the plaintiffs' alleged injuries arose. Such implantation is more akin to the single spill that led to continuing soil damage in SCSC than it is to the situation in NSP or Domtar where "contamination could not be apportioned among causes." Id. at 730.

Accordingly, we conclude that this case is not one of the "difficult cases" in which allocation is appropriate and, therefore, we hold that the lower courts erred in allocating the damages among the insurers in this case. Id. at 733. Consistent with our actual-injury trigger theory, we hold that those insurers on the risk at the time of implantation are liable up to the limits of their respective policies for 3M's losses arising from that implantation.

Having decided that allocation is not appropriate in this case, we need not address 3M's claim that the allocation period should not have been extended by the court of appeals. Similarly, absent allocation, the insurers' argument that the judgment against them should be reduced in accordance with 3M's undertaking to XL and A.C.E. is moot because 3M did not have policies with XL and A.C.E. at the times of implantation and XL and A.C.E., therefore, share no common liability with the respondent insurers. Thus, we decline to address this issue as well.

III.

We turn next to 3M's argument that even absent statutory authorization or breach of a contractual duty to defend, it is entitled to coverage-action attorney fees and costs based on the district court's finding that the insurers breached the implied covenant of good faith and fair dealing. The determination of whether an insured is entitled to attorney fees predicated on a finding of a breach of the implied covenant of good faith and fair dealing is a question of law. See Paidar v. Hughes, 615 N.W.2d 276, 279 (Minn.2000). We review questions of law de novo. Frost-Benco Elec. Assoc. v. Minn. Pub. Util. Comm., 358 N.W.2d 639, 642 (Minn.1984).

Under Minnesota's common law, "each party bears [its] own attorney fees in the absence of a statutory or contractual exception." Ly v. Nystrom, 615 N.W.2d 302, 314 (Minn.2000); Garrick v. Northland Ins. Co., 469 N.W.2d 709, 714 (Minn.1991) (noting that for over 100 years it has been the law in Minnesota that attorney fees are only recoverable by a prevailing party when there is statutory authorization or a contractual agreement allowing those fees). However, in the insurance context, we have carved out a narrow exception to the general rule: attorney fees are recoverable when an insurer breaches its duty to defend. See Morrison v. Swenson, 274 Minn. 127, 142 N.W.2d 640 (1966).

In Morrison, the insured commenced a declaratory judgment action against his insurance company to force it to defend him in a personal injury action. 274 Minn. at 132, 142 N.W.2d at 644. We affirmed the district court's determination that the insurance company was required to defend. 274 Minn. at 137, 142 N.W.2d at 647. Turning to the issue of whether the insured could recover the attorney fees he incurred in the declaratory judgment action, we recognized that, absent statutory authorization, attorney fees are generally not recoverable. Id. However, we awarded the insured attorney fees for the following reasons:

[T]his action is in the nature of an action to recover damages for breach of contract. Legal fees incurred in the declaratory judgment action were damages arising directly as the result of the breach. We think that the injured party in an action of this kind ought to be permitted to recover whatever expenses he has been compelled to incur in asserting his rights, as a direct loss incident to the breach of contract.

274 Minn. at 138, 142 N.W.2d at 647.

In Abbey v. Farmers Insurance Exchange, we clarified our holding in Morrison and concluded that, absent statutory authorization, an insured's ability to recover attorney fees is limited to situations where the insurer has breached its contractual duty to defend. 281 Minn. 113, 119, 160 N.W.2d 709, 712 (1968) (holding that where the insured was seeking recovery of disability benefits from the insurer, the Morrison exception did not allow the insured to recover attorney fees incurred in his declaratory judgment action against the insurer). In subsequent cases, we have similarly refused to extend the Morrison exception. See Rent-A-Scooter, Inc. v. Universal Underwriters Ins. Co., 285 Minn. 264, 268-69, 173 N.W.2d 9, 12 (1969) ("[A]bsent statutory authority or specific provision in the insurance contract itself, the insured may not recover attorneys' fees incurred in an action against the insurer to establish coverage under an insurance policy."); Garrick, 469 N.W.2d at 714 (holding that Minn.Stat. § 555.08 (1990) could not be extended to provide attorney fees "beyond the typical Morrison-type exception, i.e., fees incurred as a direct loss incident to the breach of a contractual duty to defend").

Here, although the insurers did not have a duty to defend and there is not a relevant statute authorizing an award of attorney fees, the district court nevertheless granted 3M reasonable attorney fees based on the court's finding that the insurers breached their implied covenant of good faith and fair dealing.^[5] The court found that the insurers' "actions, whether viewed individually, as I do, or collectively as 3M does, demonstrate a course of conduct that constitutes a practical repudiation of their individual insurance contracts and of the insurance program in which they were knowing and enthusiastic participants." The court reasoned as follows:

The Insurers did not deal fairly with 3M. Under such circumstances an award of attorneys' fees is appropriate, is within the reasoning expressed in the duty to defend cases, and is required as a practical matter if commercial general liability insurance is to work in a coherent manner in litigious modern industrial society and economy.

The court of appeals reversed the district court, concluding that the Morrison exception is limited to damages resulting from the insurer's breach of its duty to defend. In re Silicone Implant Ins. Coverage Lit., 652 N.W.2d at 73. Furthermore, the court concluded that even if attorney fees were recoverable, 3M would not be entitled to an award of attorney fees here because the district court found that 3M suffered no legally recognizable damages as a result of any breach of the implied covenant of good faith and fair dealing and, therefore, 3M failed to establish that any damages flowed directly from the alleged breach.

3M asserts that we recognized the potential application of the Morrison exception outside of the duty to defend context in American Standard Insurance Co. v. Le, 551 N.W.2d 923 (Minn.1996). On this basis, 3M argues that it is entitled to an award of attorney fees on the ground that the insurers breached their contractual duty to reimburse 3M for its defense costs and therefore the insurers acted in violation of their implied covenant of good faith and fair dealing.^[6] In response, the insurers assert that Morrison only applies when the insurer has breached its contractual duty to defend.

3M cites the following language in American Standard to support its position that an award of attorney fees is appropriate when an insurance company breaches its duty to reimburse defense costs:

The insured is not entitled to recover attorney fees incurred in maintaining or defending a declaratory action to determine the question of coverage unless the insurer has breached the insurance contract in some respect—usually by wrongfully refusing to defend the insured.

American Standard, 551 N.W.2d at 927. 3M contends that as a result of this language, a breach of the duty to reimburse defense costs, not just a breach of the duty to defend, supports an attorney fee award. We disagree.

In American Standard, the insurer assumed the insured's defense in a civil action subject to a reservation of its right to assert that the insured's claim was not covered under his policy. 551 N.W.2d at 924. The insurer subsequently commenced a declaratory judgment action to resolve the coverage issue. Id. The district court found that the insurer was obligated to defend and to indemnify the insured, and also ordered the insurer to pay the insured the attorney fees he incurred defending the declaratory judgment action. Id. at 924-25. The issue on appeal to this court was whether the district court properly awarded the insured attorney fees. Id. at 925.

We began our analysis in American Standard by recognizing that Morrison is "the limited exception to the general rule" and is limited to damages resulting from a breach of contract by the insurer's failure to defend. American Standard, 551 N.W.2d at 926. We also noted that we have "consistently resisted efforts to expand the Morrison holding to allow collection of attorney fees in actions which do not involve the insurer's breach of contract by failure to assume the duty to defend." Id. We then overruled Econ. Fire & Casualty Co. v. Iverson, 445 N.W.2d 824 (Minn.1989), and Lanoue v. Fireman's Fund American Insurance Cos., 278 N.W.2d 49 (Minn.1979), to the extent that they were inconsistent with Morrison "as a limited exception to the general rule that legal fees are not recoverable absent statutory authority." American Standard, 551 N.W.2d at 927-28. Thus, we concluded that because the insurer had undertaken the defense of the insured, the district court erred in awarding the insured attorney fees incurred in defense of the declaratory action. Id. at 928.

Taken in the context of the entire opinion, the language 3M points to from American Standard does not support its argument that the Morrison exception is applicable outside of the duty to defend context. In American Standard, we reiterated that Morrison is restricted to damages resulting from a breach of the insurer's duty to defend and we overruled two cases with contradictory holdings. Id. at 926. An expansion of the Morrison exception was not necessary to reach our holding that the insured was not entitled to attorney fees. In fact, in reaching our holding we applied the Morrison exception and concluded that attorney fees were not recoverable in a declaratory judgment action to determine whether the insurer was required to indemnify the insured when the insurer did not breach its duty to defend. Therefore, 3M is incorrect that American Standard expanded the Morrison exception.

3M also argues that, even if American Standard does not expand the Morrison exception, the award of attorney fees based on the insurers' breach of their duty to reimburse defense costs is justified here because an award of fees is consistent with the rationale underlying Morrison. 3M reasons that an insured who has bargained for an insurer to pay for defense costs is indistinguishable from an insured who has bargained for an insurer to provide the defense. 3M asserts that in both cases the insured contracts to avoid the burdensome expense of litigation only to have litigation thrust upon it by the insurer in a coverage action. We disagree and conclude that the district court's award of attorney fees to 3M premised on the insurers' failure to reimburse defense costs expands the Morrison exception beyond its rationale.

The facts prompting recognition of the Morrison exception involved the insurer's breach of its duty to defend. Morrison, 274 Minn. at 132, 142 N.W.2d at 644. In Morrison, we reasoned that the expenses the insured incurred in bringing the declaratory judgment action to establish the insurer's duty to defend were "a direct loss incident to the breach of contract," and therefore awarded the insured attorney fees. Id. at 138, 142 N.W.2d at 647. These expenses were "a direct loss incident to the breach of the contract" because the insured had contracted with the insurer to avoid the burden of litigation. The burden of litigation extends beyond the monetary costs of litigation and encompasses hiring attorneys and managing lawsuits. As the insurers argue, if an insurer breaches its duty to defend, the insured must do twice what it contracted to avoid: hire attorneys and manage a lawsuit for both the underlying case and the declaratory judgment proceeding.

In contrast, the agreement to reimburse the insured for defense costs by its high-level, excess insurance providers does not involve the promise to relieve the insured from the burdens of litigation. The insured must still hire an attorney and manage the underlying litigation. An agreement to reimburse the insured's defense costs is simply an agreement for the payment of money. Attorney fees are not recoverable in declaratory judgment actions to establish that the insurer must pay the insured money. See Abbey, 281 Minn. at 119, 160 N.W.2d at 712 (holding that an insured was not entitled to recover attorney fees incurred in his declaratory action against the insured where the insured was seeking to recover disability benefits). Accordingly, we affirm the court of appeals and hold that the district court erred in finding that because the insurers breached the implied covenant of good faith and fair dealing, 3M was entitled to attorney fees.

Affirmed in part and reversed in part.

CASEY, FREDERICK J., Acting J., concurs.^[*]

PAGE, GILBERT, and HANSON, JJ., took no part in the consideration or decision of this case.

NOTES

[1] Because it acquired the original manufacturer, McGhan Medical Corp., in 1977 and sold it in 1984, 3M is considered the manufacturer for that period.

[2] The court did amend a finding that is not relevant to this appeal.

[3] Occurrence is defined as "an accident, event or happening, including injurious exposure to conditions, which results, during the policy period, in bodily injury or property damage neither expected nor intended from the standpoint of the insured."

[4] As the insurers point out, the plaintiffs generally did not begin to experience autoimmune disease symptoms until well after the policies at issue expired.

[5] The district court made this finding despite its prior ruling in which it granted the insurers a directed verdict on 3M's breach of the implied covenant of good faith and fair dealing claim. The court found that although 3M submitted sufficient evidence to create a jury issue as to liability, 3M's claim failed because 3M failed to show that the breach resulted in damages.

[6] While 3M states that it "contracted with the insurers to avoid the costs of defending the underlying tort litigation," this statement is misleading-not all of the insurers had a contractual duty to reimburse 3M for its defense costs. In fact, policies issued by 14 of the insurers do not provide for the payment of 3M's defense costs and 3M's arguments therefore do not apply to these insurers. Accordingly, our discussion is relevant for only those policies that include reimbursement for defense costs.

[*] Appointed pursuant to Minn. Const. art. VI, §§ 2, 10, and Minn.Stat. § 2.724, subds. 1, 2 (2002).