A comprehensive general liability (hereafter sometimes referred to as a CGL) insurance policy is typically structured as a broad grant of coverage followed by a separate section imposing limitations on that coverage. The general rule is that the policyholder has the burden of proving that the loss is within the scope of coverage, while the insurance company has the burden of proving that the loss falls within any express limitation on coverage. In this case, a corporation bought two CGL policies providing, among other things, broad coverage for liability imposed on the policyholder in the event of property damage. Each policy expressly excluded liability arising from pollution (the pollution exclusion) “unless” the release of the pollutants was “sudden and accidental.” The issue here is this: Does the policyholder or the insurer have the burden of proving whether the release of the pollutants was “sudden and accidental”?
The majority places the burden on the policyholder. I disagree. It is a general principle of insurance law that an insurance company relying on a *1201separately stated limitation of coverage must prove that the loss is within the limitation provision. Applying that principle, I would hold that the insurer must demonstrate the inapplicability of any limitation contained in the exclusion clause, however that limitation may be phrased. Thus, here the insurer would have to prove that the release of the pollutants was not sudden and accidental.
I
Aydin Corporation (Aydin) made and repaired electrical transformers at a plant in Palo Alto. Toxic chemicals used in the process and waste materials were stored in underground tanks and above the ground in metal drums. In 1980 Aydin discovered that the land and groundwater at the site were contaminated with polychlorinated biphenyl (PCB), a highly toxic pollutant. Sludge leaking from holes in the underground tanks contained PCB. When the California Department of Health Services learned of the contamination, it issued a remedial action order, directing Aydin to clean up the site.
Aydin was insured with First State Insurance Company (First State). Each of the two CGL policies that Aydin had purchased from First State promised to indemnify Aydin for “all sums which the Insured shall be obligated to pay by reason of the liability imposed upon the Insured by law . . . because of: [^[] . . . ft¡] B. Property Damage, as hereinafter defined . . . flO to which this policy applies, caused by an Occurrence, as hereinafter defined . . . .” The policies define an “Occurrence” as “an accident or event including continuous repeated exposure to conditions, which results, during the policy period, in . . . Property Damage neither expected nor intended from the standpoint of the Insured. . . .”
Under the heading, “Exclusions,” each policy precluded coverage for • “any liability of any Insured arising out of the discharge, dispersal, release or escape of smoke, vapors, soot, fumes, acids, alkalis, toxic chemicals, solids, liquids or gases, waste materials or other irritants, contaminants or pollutants into or upon land . . . unless such discharge, dispersal, release or escape is sudden and accidental.” (Italics added.)
Aydin brought an action for declaratory relief to determine the extent of the coverage under the policies. First State contended there was no coverage because the release of pollutants from Aydin’s underground storage tanks was not “sudden and accidental,” as required by the policy.
The case was tried before an advisory jury. The trial court instructed the jury that First State had to prove that Aydin’s liability arose from the *1202discharge of pollutants, and that the discharge was not sudden and accidental. The advisory jury found in Aydin’s favor, and the trial court adopted its findings. The Court of Appeal reversed, holding that Aydin had the burden of proving that the discharge of PCB was “sudden and accidental” within the meaning of the policy language.
II
To establish coverage under an insurance policy, the insured must show that the occurrence on which the claim is based falls within the scope of basic coverage. (Weil v. Federal Kemper Life Assurance Co. (1994) 7 Cal.4th 125, 148 [27 Cal.Rptr.2d 316, 866 P.2d 774, 32 A.L.R.Sth 789]; Searle v. Allstate Life Ins. Co. (1985) 38 Cal.3d 425, 438 [212 Cal.Rptr. 466, 696 P.2d 1308].) The insurance company then has the burden of proving that the particular claim is excluded by some other provision of the policy. (Garvey v. State Farm Fire & Casualty Co. (1989) 48 Cal.3d 395, 406 [257 Cal.Rptr. 292, 770 P.2d 704]; Searle v. Allstate Life Ins. Co., supra, at pp. 437-438; Clemmer v. Hartford Insurance Co. (1978) 22 Cal.3d 865, 880 [151 Cal.Rptr. 285, 587 P.2d 1098].) The latter rule has existed for some 100 years. (See e.g., Dennis v. Union Mut. Life Ins. Co. (1890) 84 Cal. 570, 572 [24 P. 120]; Western Assur. Co. of Toronto v. J. H. Mohlman Co. (2d Cir. 1897) 83 Fed. 811, 818; Meadows v. Pacific Mut. Life Ins. Co: of California (1895) 129 Mo. 76 [31 S.W. 578, 582-583]; see cases cited in 33 C.J., Insurance, § 832, p. 111, fn. 95.) Although many decisions have relied on this rule, only rarely have courts discussed the reasons underlying it. Perhaps the most helpful explanation can be found in Jewelers Mutual Insurance Company v. Balogh (5th Cir. 1959) 272 F.2d 889. There, the federal reviewing court observed that when an insurer sells a policy purporting to provide broad coverage for all but certain excluded risks, to require the policyholder to affirmatively establish the nonexistence of each of the listed exclusions would be inconsistent with the comprehensive nature of the policy. In the words of the court: “If the insurer’s contention [that the policyholder has the burden of proving policy exclusions] is sound, then as a condition precedent to liability, the assured would have to establish by a preponderance the negative of each of these manifold exceptions. . . . Instead of the policy affording coverage against all causes of damage except those specifically excluded, it would amount only to a named perils cover since the assured to negative the exception would have to establish . . . [an] event not specifically excluded.” (Id. at p. 892.) Another federal decision succinctly noted that if the policyholder had the burden of proving that coverage was not expressly excluded, “the inclusive character of the coverage of the insurance policy would be a delusion, and a snare.” (Chase Rand Corporation v. Central Ins. Co. (S.D.N.Y. 1945) 63 F.Supp. 626, 629.)
*1203Often, as here, an exclusion clause itself contains restrictions limiting the scope of the exclusion. In that situation, does the general rule that an insurance company must establish the applicability of an exclusion also require it to negate any limitations contained in the exclusion clause? A leading insurance treatise has this to say: “In many instances, a policy exception is itself subject to exceptions or limitations. When this is the case, the insurer in proving that the loss comes within the exception must also proceed further to show that the exception or limitation to the exception does not preclude the application of the exception. Consequently, if an exception of liability for injuries intentionally inflicted does not extend to all injuries so inflicted . . . the [insurer] does not sustain its burden of proof unless it negatives the exception in the exemption.” (19 Couch on Insurance (2d ed. 1983) § 79:385, p. 338.) Thus, the general rule is that the insurance company has the burden of proving that limitations or exceptions to an excluded risk are inapplicable. In my view, this rule makes sense, as I discuss below.
Whether to describe the scope of a coverage limitation by phrasing it as an exception is generally a question of semantics, not substance, as this case illustrates. The two CGL policies at issue here excluded coverage for property damage liability resulting from the discharge of pollutants “unless such discharge ... is sudden and accidental.“ The majority concludes that this phrasing makes the “sudden and accidental” language a specific exception to the general pollution exclusion. Yet if the policies had instead excluded from coverage “any gradual or intentional discharge of pollutants,” this limitation would serve to define the exclusion—thus, under well-established insurance law, placing the burden of proof on the insurance company—and could not be characterized as an exception to it. Although there would be no change in meaning, no change in substance, the majority would allow a strictly semantic variation to determine which party had the burden of proof.
Aydin’s brief in this court furnishes another illustration. If a health insurance policy excludes from coverage “pre-existing heart disease,” the term “pre-existing” defines the exclusion and so, under well-established insurance law, an insurer seeking to rely on the exclusion to deny coverage for expenses incurred to treat a heart ailment would have the burden of proving that the insured had already contracted the disease when the policy was purchased. Yet if, as the majority holds today, a policyholder must prove the existence of any exception to an excluded risk, a health insurer could shift this burden of proof to the insured merely by a subtle rephrasing of the policy to exclude coverage for “all heart disease, unless it develops after the inception of the insurance policy.”
To be sure, an insurer generally has a right to decide what coverage to offer, and an insurer may control the scope of coverage, and the obligation it *1204thereby incurs, by the provisions for coverage and exclusions from coverage it states in the policy. To protect the reasonable expectations of those who purchase insurance policies, however, courts have rightly insisted that any exclusions to broadly worded coverage provisions be clearly expressed, and courts have resolved ambiguities as to the meaning of exclusionary provisions in favor of coverage. In addition, because allocation of the burden of proof effectively determines the scope of coverage when evidence of the fact at issue is nonexistent or closely balanced, courts have also insisted that the insurer, and not the policyholder, bear the burden of proving the existence of facts necessary to establish exclusions to broadly worded coverage provisions. Permitting an insurer to use subtle changes in the wording of policy exclusions to manipulate allocations of the burden of proof significantly weakens the protections afforded policyholders by these established principles of insurance law. The rule allocating to the insurer the burden of proving both exclusions and the nonexistence of any qualifications or exceptions to those exclusions is fundamentally more consistent with the spirit of established insurance law.
Also, not to apply these established principles of insurance law could considerably and unnecessarily increase the cost of litigation simply to determine whether subtly worded limitations to coverage contained in an exclusion clause have shifted to the policyholder the burden of proving such limitations.1 Theoretically, policy language can be subdivided endlessly, piling exception upon exception, shifting the burden of proof back and forth, back and forth, with each successive exception. Such shifting increases the cost and complexity of litigation, outweighing any conceivable benefit to be gained from shifting the burden of proof.
For the reasons set forth above, I would hold that once a policyholder has shown that a claim falls within the policy’s basic grant of coverage, the insurer must prove any separate limitation on that coverage, no matter how that limitation may be phrased. Therefore, here First State would have the burden of establishing that the chemical contamination of Aydin’s property did not result from a “sudden and accidental” discharge of pollutants from Aydin’s underground storage tanks.
Ill
In support of its holding, the majority relies on five decisions from other jurisdictions that have placed on the policyholder the burden of proving that *1205the discharge of pollutants was “sudden and accidental.” (E.I. du Pont de Nemours v. Allstate Ins. (Del. 1997) 693 A.2d 1059, 1061; Highlands Ins. Co. v. Aerovox Inc. (1997) 424 Mass. 226 [676 N.E.2d 801, 804-805]; SCSC Corp. v. Allied Mut. Ins. Co. (Minn. 1995) 536 N.W.2d 305, 314; Northville Industries v. Nat. Union Ins. (1997) 89 N.Y.2d 621 [657 N.Y.S.2d 564, 679 N.E.2d 1044, 1048-1049]; Sinclair Oil Corp. v. Republic Ins. Co. (Wyo. 1996) 929 P.2d 535, 543.)2 I find these decisions unpersuasive.
Two of the decisions on which the majority relies (E.I. du Pont de Nemours v. Allstate Ins., supra, 693 A.2d at p. 1061; Sinclair Oil Corp. v. Republic Ins. Co., supra, 929 P.2d at p. 543) offer no reason whatsoever for shifting the burden of proof to the policyholder. Two others (SCSC Corp. v. Allied Mut. Ins. Co., supra, 536 N.W.2d at p. 314; Northville Industries v. Nat. Union Ins., supra, 679 N.E.2d at pp. 1048-1049) dispose of the issue in a single paragraph, without extended analysis. The fifth decision (Highlands Ins. Co. v. Aerovox Inc., supra, 676 N.E.2d at pp. 804-805) merely adopts the reasoning of the federal reviewing court in Aeroquip Corp. v. Aetna Cas. and Sur. Co., Inc. (9th Cir. 1994) 26 F.3d 893, 895 (Aeroquip). Because the majority here too relies on Aeroquip, I now turn to a discussion of that case.
Aeroquip involved the same issue that we consider here: When a comprehensive general liability policy’s clause excluding certain pollution coverage contains an exception for the “sudden and accidental” discharge of pollutants, is it the insurer or the policyholder that must prove the exception to the exclusion? (Aeroquip, supra, 26 F.3d 893.) The case arose under California law, but this court had never considered the issue. In these circumstances, “a federal court must predict how the highest state court would decide the issue . . . .” (In re Kirkland (9th Cir. 1990) 915 F.2d 1236, 1239.) The federal court in Aeroquip concluded that this court would likely impose the burden on the policyholder. (Aeroquip, supra, at p. 895.) It cited three reasons for its conclusion.
First, the Aeroquip court reasoned that imposing on the policyholder the burden of proving the “sudden and accidental” limitation on the pollution *1206exclusion would be consistent with the rule that the insured has the initial burden of proving that an occurrence is within the scope of the basic coverage. This is true, the court said, because “[t]he ‘sudden and accidental’ exception creates coverage where it would otherwise not exist . . .” {Aeroquip, supra, 26 F.3d at p. 895.) The majority in this case echoes that view, stating that the “sudden and accidental” limitation in the exclusion clause “reinstates” coverage (maj. opn., ante, at p. 1192) and therefore should be “construed as a coverage provision” {id. at p. 1193).
Aeroquip and the majority are both wrong in concluding that the “sudden and accidental” limitation creates or reinstates coverage. It does neither. All it does is to place a limit on the scope of the pollution exclusion. To construe this limitation as a coverage clause runs counter to “the basic principle that exclusion clauses subtract from coverage rather than grant it.” {Weedo v. Stone-E-Brick, Inc. (1979) 405 A.2d 788, 795 [81 N.J. 233]; see also Sturla, Inc. v. Fireman’s Fund Ins. Co. (1984) 67 Hawaii 203 [684 P.2d 960, 965]; Continental Cas. Co. v. Pittsburgh Coming Corp. (7th Cir. 1990) 917 F.2d 297, 300.) Here, the events for which Aydin seeks coverage are within the broad grant of coverage of the comprehensive policies. Therefore, Aydin is entitled to that coverage unless the insurer shows that the loss falls within the pollution exclusion.
Second, according to the federal court in Aeroquip, to require the insurer to prove that the discharge of pollutants was not “sudden and accidental” would give policyholders “an incentive to avoid finding out whether pollutants are being gradually discharged, because preservation of ignorance would increase the likelihood of insurance coverage.” {Aeroquip, supra, 26 F.3d at p. 895.) But it is farfetched to assume that policyholders decide upon a course of conduct based upon the apportionment of the burden of proof in the exclusion clauses of their liability insurance policies. Even if they did, the onerous consequences of allowing pollution to continue would far outweigh any potential benefit to policyholders from the speculative possibility that their insurers would ultimately be unable to defeat coverage. Almost inevitably, that pollution will be discovered, leading to governmental cleanup orders. The longer pollution continues unabated, the more difficult and costly it will be to comply with these orders.
Third, Aeroquip asserts that requiring the policyholder to prove the limitation at issue is appropriate because it “places the burden on the party who will generally have access to facts that show the discharge of pollutants was sudden and unexpected.” {Aeroquip, supra, 26 F.3d at p. 895.) I find this reasoning unpersuasive. True, policyholders may have better access to the facts leading to the loss because they have better access to their own insured *1207property than insurers. But if such access were the sole basis for apportioning burdens of proof, then policyholders would always have the burden of negating policy exclusions. That is not the law, however. As the majority acknowledges, it is “well established”' that the insurer, not the policyholder, must prove that a loss within the broad coverage grant falls within an excluded risk. (Maj. opn., ante, at p. 1188.)
The majority in this case adopts the reasoning of Aeroquip, supra, 26 F.3d 893. In addition, the majority points to a fourth reason for its holding, namely, implementation of the parties’ intent. It states: “The fact that different policy language might result in a different allocation of the burden of proof should hardly come as a shock. Rather, it arises from the parties’ general freedom to contract as they deem fit.” (Maj. opn., ante, at pp. 1192-1193.) I have no quarrel with the majority’s proposition that parties have the freedom to contract in any way they like. But, as I pointed out in the preceding paragraph, under general principles of law the burden of proof is on the insurance company to establish that a particular claim is excluded. Here, the parties did not agree to a different allocation of the burden of proof. Therefore, First State rather than Aydin should have the burden to prove that the discharge of pollutants was not “sudden and accidental.” (See AIU Ins. Co. v. Superior Court (1990) 51 Cal.3d 807, 822 [274 Cal.Rptr. 820, 799 P.2d 1253] [“Because the insurer writes the policy, it is held ‘responsible’ for ambiguous policy language, which is therefore construed in favor of coverage.”]; Reserve Insurance Co. v. Pisciotta (1982) 30 Cal.3d 800, 808 [180 Cal.Rptr. 628, 640 P.2d 764] [exclusionary clauses are construed narrowly against the insurer].)3
Conclusion
A comprehensive general liability policy is expected to provide broad, all-encompassing coverage. “ ‘Obviously the very name of the [comprehensive general liability] policy suggests the expectation of maximum coverage. Consequently the comprehensive policy has been one of the most preferred by businesses and governmental entities over the years because that policy *1208has provided the broadest coverage available. All risks not expressly excluded are covered, including those not contemplated by either party.’ ” {Quaker State Minit-Lube, Inc. v. Fireman’s Fund Ins. Co. (¿.Utah 1994) 868 F.Supp. 1278, 1295.)
Before today’s decision, a business entity purchasing a comprehensive general liability policy could reasonably expect that to obtain the policy’s promise of broad protection from liability, it had only to show that it had incurred such liability, and that the insurance company could defeat the coverage only by showing that the claim fell within one of the policy’s limitations or exclusions. The majority’s holding frustrates this expectation, and it gives insurers an unanticipated and unwarranted windfall.
I would reverse the judgment of the Court of Appeal.
Werdegar, J., concurred.
Respondent’s petition for a rehearing was denied October 14, 1998, and the opinion was modified to read as printed above. Mosk, J., Kennard, J., and Werdegar, J., were of the opinion that the petition should be granted.
Thus, although the careful syntactic parsing of the policy language in Justice Mosk’s dissenting opinion may be analytically sound, allocation of the burden of proof should not turn on such subtle grammatical distinctions.
The majority also notes that some federal courts have predicted that other state high courts as well would place the burden on the insured. (Maj. opn., ante, at p. 1190.) Other federal decisions, however, have predicted that certain state high courts would place the burden on the insurer. (EDO Corp. v. Newark Ins. Co. (D.Conn. 1995) 878 F.Supp. 366, 371 [predicting Connecticut would place burden on insurer]; A.Y. McDonald Industries v. INA (N.D. Iowa 1993) 842 F.Supp. 1166, 1171 & fn. 2 [predicting Iowa would place burden on carrier]; U.S. Fidelity & Guar. v. Morrison Grain Co. (D.Kan. 1990) 734 F.Supp. 437, 442-443 [predicting Kansas would place burden on insurer]; see also State of N.Y. v. Blank (2d Cir. 1994) 27 F.3d 783, 788-789 [wrongly predicting New York would place burden on insurer]; New Castle County v. Hartford Acc. and Indent. Co. (3d Cir. 1991) 933 F.2d 1162, 1181-1182 [wrongly predicting Delaware would place burden on insurer].)
The majority’s holding may in some cases conflict with Evidence Code section 520. That statute provides, “The party claiming that a person is guilty of crime or wrongdoing has the burden of proof on that issue.” The intentional, unauthorized discharge of dangerous chemicals onto the land is a wrongful and sometimes criminal act. To the extent that First State here asserts that the “sudden and accidental” limitation on the pollution exclusion is inapplicable because Aydin intentionally rather than accidentally discharged PCB onto its property, First State may, under Evidence Code section 520, have the burden of establishing the truth of this allegation. (But see FMC Corp. v. Plaisted & Companies (1998) 61 Cal.App.4th 1132, 1160 [72 Cal.Rptr.2d 467] [section 520 inapplicable to violation of “essentially regulatory” statutes prohibiting toxic contamination, when much of the contamination occurred before the statutes were enacted].) Because the parties have not addressed this issue, I do not consider it further.