The question we decide on this appeal is whether a company's general liability insurance policy, which excludes coverage for property damage caused by pollution, nevertheless provides coverage in this particular case under the policy's exception for "sudden and accidental" polluting events. Throughout the four-year period in which the relevant policies were in effect, the insured's manufacturing operations involved continuing pollution-releasing activity. The alleged "sudden and accidental" events occurred on two days during this period, one in 1973 and the other in 1975. We conclude that these events do not qualify as "sudden and accidental" discharges of pollutants and, accordingly, need not reach the trigger of coverage and notice issues that were decided below.

The general liability insurer-appellant is Lumbermens Mutual Insurance Company (Lumbermens). The manufacturer-appellee is Belleville Industries, Inc. (Belleville).1 In 1973 Belleville acquired an old brick and wood building on the banks of the Acushnet River near its entry into New Bedford Harbor. It carried on the same general process of manufacturing capacitors as had the seller, Aerovox Corporation, for some 26 years. Capacitors are devices which accumulate and hold electric charges. They consist of two oppositely charged surfaces, separated by a dielectric, or insulator. In this particular operation the dielectric was a fluid, "Aroclor," which consisted of a chemical compound, polychlorinated biphenyls, or PCBs. Belleville purchased PCBs from Monsanto Chemical Corporation between 1973 and early 1977, the period when Lumbermens provided general liability coverage for the company. By 1978 the toxicity of PCBs had become so well-recognized that they were outlawed. Ninety-nine percent of the PCBs purchased by Belleville was of a particular kind, Aroclor 1016, which was 99.6% biodegradable. That is, over time most of the pollutant would convert into non-polluting substances, leaving only .4% in a toxic condition.

Late in 1983, the United States and the Commonwealth of Massachusetts brought an action under the Comprehensive Environmental Response, Compensation and Liability Act (CERCLA) and other environmental and civil statutes against Belleville and five other corporations, seeking damages and cleanup costs resulting from pollution of the Acushnet River and New Bedford Harbor. In August 1984, Belleville gave written notice of its asserted liability to Lumbermens and demanded that the insurer defend and indemnify the company. Lumbermens denied coverage and subsequently brought this declaratory judgment action to determine its liability under the policies. Meanwhile, it proceeded to defend Belleville. Ultimately, after the district court issued a partial summary judgment for the governmental plaintiffs on the issue of Belleville's liability, the company joined in a consent decree requiring it to pay $4 million.

These are the salient provisions of the relevant insurance policies:

--Lumbermens' basic undertaking was to pay "all sums which the insured shall become legally obligated to pay as [property] damages because of ... an occurrence....";

--The coverage-invoking event, an "occurrence," is "an accident, ... which results in ... property damage neither expected nor intended....";

--But coverage is excluded for "property damage arising out of the ... release ... of ... pollutants into or upon land, the atmosphere or any water course or body of water;"

--"[B]ut this exclusion does not apply if such ... release ... is sudden and accidental."

The Prior Proceedings. In dealing with the first of several rounds of summary judgment motions in the declaratory judgment action, the district court, though looking on the exclusion provision as "straightforward," refrained from issuing judgment because of its perception that state law precluded any judgment where the underlying claimants, the governmental plaintiffs in the original lawsuits, would not be bound. Subsequently, as we have noted, the sovereigns obtained partial summary judgment on liability in the underlying action. The court, however, continued to question its power to issue judgment in the declaratory judgment action and therefore denied all of Lumbermens' motions for summary judgment. It then certified several questions to the Massachusetts Supreme Judicial Court.

The only question relevant to our discussion sought to determine the meaning of "sudden" in the policy provision creating an exception to the exclusion of coverage for pollution damages. Courts had been divided over whether the term was ambiguous and thus could be interpreted simply as "unexpected," making it potentially applicable to gradual releases of pollutants. The Supreme Judicial Court responded that the term was not, in its view, ambiguous, and "that when used in describing a release of pollutants, 'sudden' in conjunction with 'accidental' has a temporal element." Lumbermens Mutual Casualty Co. v. Belleville Industries, Inc., 407 Mass. 675, 680, 555 N.E.2d 568, 572 (1990). It added, "[s]urely, the abruptness of the commencement of the release or discharge of the pollutant is the crucial element." 407 Mass. at 681, 555 N.E.2d at 572.

Lumbermens then filed new motions for summary judgment, in response to which Belleville proffered a number of alleged "sudden and accidental" events that it claimed mandated coverage under the policies.2 The court rejected most of Belleville's candidates for the "sudden and accidental" exception, ruling that the vast part of the damages was attributable to gradual pollution. Belleville then voluntarily limited its claims to discharges from two allegedly "sudden and accidental" events: a very heavy rainstorm in 1973 and a fire in 1975. On these issues, the parties proceeded with a nine-day, jury-waived trial.

The relevant evidence for present purposes concerns Belleville's manufacturing scenario and process, the magnitude of the rainstorm and its discharges, and the nature of the fire and resulting discharges.

The Manufacturing Process. Aroclor oil, i.e., liquid PCBs, was delivered to storage tanks in the basement of the plant. It was then pumped up to impregnation tanks on the second floor. Aluminum capacitor canisters were then lowered in wire baskets into hot Aroclor oil for a two-day "impregnation" period. The oil then was piped back downstairs to dirty oil storage tanks in the pump room, filtered, and then piped to the original storage tanks for reuse. At the same time, the wire baskets containing the impregnated capacitors were removed from the tank.

Although a drip pan was used to catch oil dripping from the baskets, drips and spills on the pump room floor occurred and were sprinkled with Fuller's earth to absorb them or mopped or squeegeed into sump pits. The contents of the sump pits were pumped into the north trough outside the plant, which led to the Acushnet River. PCBs routinely were dispersed through the north trough, the municipal sewer system, the emission of PCB vapors (which condensed as waxy residues on the plant roof, walls, ground, pavement, and parking lot), and through the leaching of PCB-contaminated waste, such as "reject" capacitors dumped on the ground and on the mud flats of the Acushnet River. During manufacturing, Aroclor in vapor form also was released into the atmosphere by exhaust fans.

The Rainstorm. On December 17, 1973, a violent rainstorm deposited 5.31 inches of water in 24 hours. One person said it had been exceeded only by Hurricane Carol in 1954; an expert's testimony was that a storm of this intensity occurred only once in seven years. Streets were flooded. Inside the plant, water flooded into a shipping area and surrounded an elevator pit, both of which were located on levels below the pump room. Water depth was no greater than one to one-and-a-half inches. Eyewitnesses reported that the pump room was not reached by any waters but that in an adjacent area on the same level, water spurted up through the floor in six-inch geysers; it did not accumulate there but drained down a ramp to a lower level where the shipping-receiving office was located. There the water at its deepest was no more than two inches; employees worked for 90 minutes to two hours cleaning up, by pushing the water into a sump pit with mops and squeegees.

The Fire. On September 24, 1975, a fire broke out in a ventilator shaft leading from the tank room to the roof. Firefighters used powerful hoses, which sent spray up to and through the roof, and caused water to flow down the stairwell and into the impregnation room. They also demolished part of the roof and building to put out the fire. Aroclor, which had accumulated on the walls, floors, and roof of the plant, was released both through the roof--eventually landing on and running off the parking lot--and down the stairs, where it was released via the sump pits.

While there was evidence of the heavy flow of water from the hoses, there was no evidence of the duration of the flow or whether the resultant releases were particularly noteworthy. The product manager observed "a stream of water coming down like a waterfall." The vice president for manufacturing, who was at the rear of the plant at the beginning of the fire, found the fire had been extinguished when he reached the site. He saw the water draining into the sump pits two or three feet below ground level. Another employee reported seeing "a lot of water ... in the aisles" that had been cleared up by the following day.

The District Court Decision. Addressing the rainstorm discharges, the district court first rejected as a "sudden and accidental" release any movement of PCBs from the mud flats to other areas of the harbor. It reasoned that although the storm may have speeded up the process, the process itself was a natural one, the water and wind spreading the PCBs. Being natural, it could not "be considered sudden as that word is used in construing the pollution exclusion under the laws of Massachusetts." Opinion at 19. In any event, it was not accidental, added the court, since Belleville had never taken any steps to reclaim or protect the flats.

As for releases from the interior of the plant, the court rejected any outflow of water through cracks in the floor because a release "by the natural progression of the flow of underground water" is not accidental, id. at 22. It also rejected any release caused by the pumping of water from the sump pits, for pumping is intended, not accidental.

The court did, however, on its own initiative, find one type of release to be both sudden and accidental. Although the plant has been reconstructed since 1973, the court inferred that there must have been a door at the first level on the eastern side to give access to both the parking lot and the tidal flats. It further reasoned that PCB-bearing water "receded under and around that door," id. at 21. This release was sudden, being caused "by operation of flood waters themselves rather than through any natural underground waterflow or natural erosion of sediments within the harbor"; it was "accidental in the sense that the release was both unexpected and occurred without the assistance of any human agency." Id. at 23. The court had earlier, in ruling on Lumbermens' motion for judgment at the end of Belleville's evidence, explained, "And I believe that the accidental event ... meets the standards of the insurance policy, at least until such time as the rushing flood waters recede to the level where manmade pumps, manmade drainage ditches, sump pumps and the like, take over." App. III at 883.

The court rejected as accidental any discharge attributable to the use of the fire hoses. The running off of water from the parking lot or through the storm sewers was not accidental because this was "the normal fashion of rainwater." Opinion at 27. And the release of water from within the plant by the sump pumps was not sudden because the pumps "operated in the fashion they were intended to," id. at 28, and not accidental because use of the sump pumps was intentional.

While there were other rulings made by the court, there is only one other which is relevant to our discussion--its ruling on damages. It placed on Lumbermens the burden to prove the percentage of damage attributable to covered and uncovered risks. It noted the difficulty in such a complex factual situation as this case presented, but observed that the present record made possible a reasonable allocation. It reasoned that virtually all of the pollutants washed out by the 1973 storm were Aroclor 1016, and that by 1983, the threshold year for liability under CERCLA, only .4% of the pollutant remained. It then multiplied the total damages paid by Belleville by .004, and awarded Belleville an indemnification from Lumbermens of $16,000.

We defer, of course, to any factual findings of the district court. Indeed, there is no issue as to underlying facts. The only dispositive issue is whether the underlying facts establish "sudden and accidental" releases within the intendment of the policies. This is a question of law that we must decide de novo.

In setting forth the heroic efforts of the district court to be scrupulously fair and to honor the traditional solicitude given insureds, we also have revealed the Augean labors inherent in microanalysis in a case of this kind. The nature of these efforts, together with a contextual view of the policy language and consideration of relevant case law, persuades us that the "sudden and accidental" exception should not be construed to provide coverage in these circumstances--in other words, when the discharges consisted of long accumulated, unattended, and unsegregated pollutants, and were caused by events not clearly beyond the long-range reasonable expectation of the insured. When, in the case of an insured whose operations involve a likelihood of continuing polluting releases, a court might properly identify a sudden release so beyond the pale of reasonable expectability as to be considered "accidental," we need not decide. We have been told that the particular policy language we must deal with here is no longer in vogue. Happily, courts may be spared further line drawing in this area.

Our analysis begins with an effort to discern the sense of the policy provisions that first exclude damages occasioned by pollution and then carve out an exception for "sudden and accidental" releases. It continues with a survey of some of the problems encountered in a microanalytical approach. And it concludes with a look at relevant case law.

Our reading of the two pollution provisions in the policy suggests that in the "ordinary" case, i.e., a case involving a "clean" operation, such as an office building housing company headquarters, insurers were willing to commit to covering a possible but unlikely event resulting in the release of pollutants. A coverable occurrence would be clearly identifiable as "sudden and accidental" because it would be a marked departure from normal operations. But in the case of a pollution-prone operation, where the emission of pollutants is part and parcel of the daily conduct of business, there is the possibility of infinite variations on the usual theme; i.e., polluting incidents are likely to occur that are on the fringe of normal operations but that the company seeks to characterize as sudden and accidental. As this case illustrates, determining where along the spectrum of polluting events coverage should begin is a perplexing and, ultimately, unsatisfying endeavor. We think it illogical to believe that insurers intended through the "sudden and accidental" exception to buy into a risk and/or litigation package of this nature. See New Castle County v. Hartford Accident and Indemnity Co., 933 F.2d 1162, 1202 (3d Cir.1991) (referring to "the insurers' underlying intent to distance themselves from deliberate polluters--i.e., those who intentionally or knowingly discharge pollutants into the environment").

More important than our bare reading of the language, however, is the actual spectrum of problems, revealed in this case, that stem from microanalysis of a continuous pattern of pollution. The district court's focus on the means by which the PCBs were spread, as a way of identifying "sudden and accidental" pollution, strikes us as a well-meaning but ill-fated effort to distinguish between virtually indistinguishable occurrences. The court concluded that "sudden and accidental" coverage would be unavailable if releases take place through intended or manmade devices and also if releases follow natural underground watercourses. But, as Belleville asks, "[i]s damage resulting from [manmade] automatic sprinklers triggered by a fire not accidental?" Similarly, is a discharge of pollutants necessarily expected simply because the release happens to travel a natural route? We think not.

Moreover, the court seemed, to some extent, to merge the concepts of sudden and accidental. For example, the court found the release of storm water by the sump pumps not to be sudden because the pumps were working as intended, and not accidental because the use of the pumps was intentional. We also have problems with the court's double jump--the first to an inferred door on the first level of the eastern face of the plant; the second to water moving under and around that door.

A more generic problem lies in determining what is sufficiently unexpected. In this case an expert witness had testified that storms of the intensity of that experienced in 1973 occurred no more frequently than once every seven years. Putting aside the thought that events which occur over time with some regularity, like the onset of locusts or gypsy moths, may be said to be expected, we wonder whether a once-in-four-years storm would qualify as unexpected. What about a once-a-year deluge? From a microanalytical viewpoint, almost any event can be labelled unexpected, since history probably never repeats itself precisely. But such an approach would eviscerate the exclusion for pollution.

Perhaps the best evidence of the infeasibility of attempting to assess discrete "fringe" events, in the case of a company with a history of contributing over a lengthy period to a gradual accumulation of pollutants, is the catalogue of "sudden and accidental" discharges submitted by Belleville. See note 2 supra. Other similar litigable possibilities would include an employee tripping and spilling Aroclor oil, a drip pan giving way, a pipe breached.... The prospect is limitless.

One final problem is that of determining the percentage of damage caused by uninsured releases and damage caused by insured "sudden and accidental" releases. In this case the court made a calculation by multiplying a quality of the (covered) pollutant (.4%) and the amount of total damages paid by Belleville ($4 million). Only if we knew what percent of the total pollution (post-CERCLA) caused by the nonbiodegradable residue of Aroclor 1016 was attributable to the 1973 rainstorm runoff around the east side door would we have a figure to multiply with the damages total. But apart from calculations, and the inherent problems of proof that allocation would involve, there is a more basic problem. Belleville argues that the policy language leaves no room for allocation. Once Lumbermens is found obligated to cover a "sudden and accidental" release, Belleville argues, the policy requires it to pay "all sums." The complete wording, however, is "all sums which the insured shall become legally obligated to pay as [property] damages because of ... an occurrence...." Whether this language permits allocation we need not decide. But we do observe that if Belleville's interpretation were correct, the result would be that of a very small tail wagging a very large dog; in this case, the flow of storm waters around a door for a few hours at the most would be sufficient to make Lumbermens indemnify Belleville for four years of pollution at a price of $4 million.

This range of practical problems reinforces our view, based on a common sense reading of the policy language, that the "sudden and accidental" exception to the pollution exclusion was not designed to operate in circumstances such as those existing in this case. We draw further support from the strong body of case law rejecting insurance coverage where a company has for a lengthy period of time purposefully and regularly been carrying on operations involving continual pollution.

In Great Lakes Container Corp. v. National Union Fire Ins. Co., 727 F.2d 30 (1st Cir.1984), although the precise issue of "sudden and accidental" releases was not raised, we had no difficulty in holding that an insurer whose insured was described as discharging chemical pollutants on its land "as a concomitant of its regular business activity," id. at 33, had no obligation to defend or indemnify in light of the "type of activity" described. Id. at 34.

Other circuits have used a similar formulation. In Ogden Corp. v. Travelers Indem. Co., 924 F.2d 39, 42 (2d Cir.1991), dealing with a scrap metal processing and storage business with a 33-year history, the court said that it was "doubtful whether the continuous discharge of pollutants resulting from the purposeful operation of a scrapyard can be construed as accidental." That same court made a more emphatic statement a year earlier in EAD Metallurgical, Inc. v. Aetna Casualty & Sur. Co., 905 F.2d 8, 11 (2d Cir.1990): "damage[ ] resulting from purposeful conduct [ ] cannot be considered 'accidental.' " United States Fidelity & Guar. Co. v. Star Fire Coals, Inc., 856 F.2d 31 (6th Cir.1988), involved a coal loading operation that had been in business for many years, loading as many as five or six 60-car trains a week. A great amount of damaging coal dust had been discharged during a period of seven or eight years when crushed coal unintentionally dropped from the conveyor belt. The court, viewing these incidents "as a normal part of the coal processing operation," held that it was "impossible to characterize these discharges of dust as 'sudden' ": "[t]he 'sudden and accidental' exception to th[e] exclusion is inapplicable here where the pollutants at issue were discharged on a regular ongoing basis." Id. at 35.

More specific rulings, in cases where parties attempted to distinguish discrete episodes of pollution from ongoing activity, have been made by a number of federal district courts and state intermediate appellate courts. Fireman's Fund Ins. Co's. v. Ex-Cell-O Corp., 750 F. Supp. 1340 (E.D.Mich.1990), involved a manufacturer of instrument panels that generated both solid and liquid wastes that entered groundwater. The company, much like Belleville, sought coverage for a tank spill in 1977 and a pipe rupture in 1978. The court held that even if these incidents and damage therefrom could be proved, they were "expected." Id. at 1350. The insurer had proved that "intentional practices, including disposal of waste water into the Pokamoonshine Brook tributary; subsequent disposal to the north end of the plant site; and finally disposal into the lagoon system, resulted in groundwater contamination." Id. at 1350-51. In other words, "[t]he evidence support[ed] insurers' theory that policyholders expected property damage to result from their day-to-day manufacturing processes." Id.

A year earlier the same court had faced the same set of issues. In Ray Industries, Inc. v. Liberty Mutual Ins. Co., 728 F. Supp. 1310 (E.D.Mich.1989), a boat manufacturer had, for 13 years, relied on a contractor to deposit 55-gallon drums of waste in a landfill. In the process some barrels were punctured and crushed. In responding to the insured's argument that discrete sudden and accidental releases occurred each time a barrel was smashed, the court held: "because ... discharges took place continually and regularly for approximately thirteen years, they were not sudden and accidental." Id. at 1318.

The absence of suddenness was the basis for the holding in Industrial Indem. Ins. Co. v. Crown Auto Dealerships, Inc., 731 F. Supp. 1517 (M.D.Fla.1990). In this case a recycler of used crankcase oil stored waste oil sludge in unlined storage ponds. Chemicals leached, spills and leaks occurred, and there was occasional runoff of rainwater. The court viewed the pollution as gradual and the normal result of the recycler's 29 years of operations. The alleged incidents were "clearly cumulative," id. at 1521. "Consequently, the leaching and occasional spills of chemicals and runoff from sludge ponds during major rainfalls cannot be classified as abrupt or sudden events." Id. Similarly, in Fischer & Porter Co. v. Liberty Mutual Ins. Co., 656 F. Supp. 132 (E.D.Pa.1986), a tank spillage was not sudden and accidental because it was "part of the regular conduct of the insured's business." Id. at 136. Even if employees dumped the pollutant, "pollution resulted from voluntary acts within the regular, routine business operations." Id. at 139.

A recent state court case is Outboard Marine Corp. v. Liberty Mutual Ins. Co., 212 Ill.App.3d 231, 156 Ill. Dec. 432, 570 N.E.2d 1154 (Ill.App.Ct., 2d Dist.1991). In that case the insured had, for over a decade, discharged PCBs into "North Ditch" leading into Waukegan Harbor. In trying to establish the exception to the pollution exclusion clause, the insured offered evidence that the harbor contamination "could have resulted from one or more events ... such as a major rainstorm, a flood, or a fire." 156 Ill. Dec. at 443, 570 N.E.2d at 1165. The court held that there is "nothing sudden about discharging pollutants over an 11-year period." Id. 156 Ill.Dec. at 444, 570 N.E.2d at 1166.

Belleville cites no cases supporting its theory. It attempts to distinguish most of the cases cited above by asserting that they did not involve an insured that was discharging pollutants, as it was, under a federal permit. See 42 U.S.C. Sec. 9607(j) (federally permitted release exception to CERCLA liability). Belleville's point in emphasizing its permit is hard to decipher. We think that what it intends us to understand is that, because of the permit, its liability in this case cannot stem from "regular business discharges into the harbor" but only from the releases associated with the rainstorm and fire. Thus, the fact that it was a regular polluter should not enter into our analysis.

Whatever the specific basis of Belleville's liability to the governmental plaintiffs, however, it does not affect our application of the "sudden and accidental" clause. It is, rather, the nature of an insured's enterprise and its historical operations that determine the applicability of the policy provision. Belleville discharged pollutants as an ordinary part of its business operations; we simply cannot analyze these provisions as if it were a manufacturer of health foods that rarely, if ever, experienced a pollution-producing event.

We therefore AFFIRM the judgment below except its declaration that Lumbermens is liable under the 1973 policy; we REVERSE its judgment of liability under the 1973 policy.