POSNER, Circuit Judge.

Is a medical mishap an "accident" under an insurance policy that provides benefits for accidental injuries that cause the death of the insured? The question arises in this case under ERISA (Employee Retirement Income Security Act, 29 U.S.C. § 1001 et seq.) because the group accident policy at issue in this case was part of an employee welfare plan. But the answer ("no," said the district judge) does not depend on any specific provisions of ERISA, and we must therefore exercise our power to formulate federal common law principles suitable for the governance of pension and employee benefit trusts. Fox Valley Vicinity Construction Workers Pension Fund v. Brown, 897 F.2d 275, 281 (7th Cir. 1990) (en banc); Wickman v. Northwestern National Ins. Co., 908 F.2d 1077, 1084 (1st Cir. 1990).

The insured, a 36-year-old woman suffering from Crohn's Disease (a chronic inflammatory disease of the intestine), was admitted to a hospital with a tentative diagnosis of intestinal obstruction. A catheter was inserted in a vein beneath her clavicle for the purpose of administering nourishment intravenously. This is a standard treatment for Crohn's. Harrison's Principles of Internal Medicine 431-33 (Jean D. Wilson et al. eds., 12th ed. 1991). Several days later she died suddenly. The catheter had become detached from its original position, had entered the heart through a vein, and had punctured the heart. The preliminary death certificate listed the puncture as the cause of death, but the final certificate listed marked nutritional deficiency resulting from Crohn's as the cause of death.

The summary plan document which ERISA requires be given the participant in an ERISA plan (see 29 U.S.C. § 1021(a), 1022(a)(1), and 1024(b)(1)) describes the insurance policy in this case as "offer[ing] full 24 hour protection against accidents anywhere in the world, whether you are on business, pleasure, vacation, at home, or on or off the job," with the usual exclusions, such as suicide, a military mishap, or a plane crash unless the insured is an airline passenger. The document refers the participant to the insurance policy itself for details, and the policy expressly excludes not only "sickness or disease" but also "medical or surgical treatment of a sickness or disease." This still leaves a question whether a mishap in the course of treatment should be classified as part of the treatment itself or as an accident, but only a tiny question, in light of numerous cases illustrated by Whetsell v. Mutual Life Ins. Co., 669 F.2d 955 (4th Cir. 1982); Castorena v. Colonial Life Accident Ins. Co., 107 N.M. 460, 760 P.2d 152 (1988); Reid v. Aetna Life Ins. Co., 440 F. Supp. 1182 (S.D.Ill. 1977), aff'd without opinion, 588 F.2d 835 (7th Cir. 1978), and Krane v. Aetna Life Ins. Co., 698 F. Supp. 220 (D.Colo. 1988); contra, Mayfield v. Metropolitan Life Ins. Co., 585 S.W.2d 163, 168-69 (Mo.App. 1979).

The plaintiff argues that her decedent never received the policy. So what? Nothing in ERISA requires that the insurance policy summarized in the summary plan document be given the insured. The insured is protected by the fact that, in the event of a discrepancy between the coverage promised in the summary plan document and that actually provided in the policy, he is entitled to claim the former. Hansen v. Continental Ins. Co., 940 F.2d 971, 982 (5th Cir. 1991); Edwards v. State Farm Mutual Automobile Ins. Co., 851 F.2d 134, 136 (6th Cir. 1988); McKnight v. Southern Life Health Ins. Co., 758 F.2d 1566 (11th Cir. 1985). The statute requires that the summary plan document be "sufficiently accurate and comprehensive to reasonably apprise" the plan participants of their rights under the plan. 29 U.S.C. § 1022(a)(1). This requirement entitles the participant to rely on the summary plan document, and if he does the plan is estopped to deny coverage.

But only if there is a contradiction between the summary plan document and the policy. There is not in a case such as this where the policy clarifies rather than contradicts the summary. And even if we are not to consider the language of the policy itself but only that of the summary plan document, the plaintiff must lose because a policy of accident insurance does not reach iatrogenic injuries, that is, injuries resulting from medical treatment. Any time one undergoes a medical procedure there is a risk that the procedure will inflict an injury, illustrating the adage that "the cure is worse than the disease." The surgeon might nick an artery; might in fusing two vertebrae to correct a disk problem cause paraplegia; might in removing a tumor from the patient's neck sever a nerve, so that the patient could never hold his head upright again. A simple injection will, in a tiny fraction of cases, induce paralysis. An injection of penicillin could kill a person allergic to the drug. A blood transfusion can infect a patient with hepatitis or AIDS. All these injuries are accidental in the sense of unintended and infrequent. But they are not "accidents" as the term is used in insurance policies for accidental injuries. The term is used to carve out physical injuries not caused by illness from those that are so caused, and while injuries caused not by the illness itself but by the treatment of the illness could be put in either bin, the normal understanding is that they belong with illness, not with accident.

That at least is our view, bearing in mind our responsibility to make the best rule of federal common law we can to guide the bench and bar of this circuit through a welter of conflicting precedent. We acknowledge a glacial drift, described in John Dwight Ingram Lynne R. Ostfeld, "The Distinction Between Accidental Means and Accidental Results in Accidental Death Insurance," 12 Fla.St.U.L.Rev. 1 (1984), toward a liberal construal of "accident" designed to maximize the beneficiary's prospects of recovering insurance proceeds. In the old days courts distinguished between accidental means and accidental result. Death is almost always accidental in the sense of unintended by the deceased, so if an accidental result sufficed, coverage would be assured regardless of the cause of death. Suppose a person who doesn't know he's allergic to penicillin contracts pneumonia and is given an injection of penicillin — the standard treatment for pneumonia — and dies, a foreseeable, in the sense of not improbable, consequence for one who has such an allergy. Or suppose, as sometimes happens, that a person undergoing surgery never recovers consciousness after being put under the general anesthetic. In either case death is unexpected, an accident, but there is nothing accidental in the means by which the person died and therefore (under the traditional view) he can't recover under an accident policy. Order of United Commercial Travelers v. Shane, 64 F.2d 55, 59 (8th Cir. 1933); Caldwell v. Travelers' Ins. Co., 305 Mo. 619, 267 S.W. 907, 921 (1924); Beneficial Standard Life Ins. Co. v. Forsyth, 447 So.2d 459, 462 (Fla.App. 1984); Wilson v. Travelers Ins. Co., 29 A.D.2d 312, 287 N.Y.S.2d 781, 784. It would be different if he twisted his knee playing tennis and the injury caused blood clots that embolized to his lungs and killed him. Insurance Co. of North America v. Thompson, 381 F.2d 677, 681 (9th Cir. 1967); Key Life Ins. Co. v. Gulledge, 245 Ark. 74, 431 S.W.2d 245 (1968); Brown v. State Mutual Life Ins. Co., 377 So.2d 355, 363 (La.App. 1979); National Life Ins. Co. v. Patrick, 28 Ohio App. 267, 162 N.E. 680 (1927); Chase v. Business Men's Assurance Co., 51 F.2d 34, 36 (10th Cir. 1931) (dictum). Then the means of death — the injury to the knee — would be an accident, and the death would be covered.

Yet as Ingram and Ostfeld explain, many courts that adhere to the means-end distinction, and more that replace it by a standard of foreseeability, would find coverage even in our hypothetical penicillin case. E.g., INA Life Ins. Co. v. Brundin, 533 P.2d 236, 243 n. 24 (Alaska 1975); Korfin v. Continental Casualty Co., 5 N.J. 154, 74 A.2d 312 (1950); International Travelers' Ass'n v. Francis, 119 Tex. 1, 23 S.W.2d 282 (1930); Aetna Life Ins. Co. v. Brand, 265 Fed. 6 (2d Cir. 1920). Many of the cases, though, involve workers' compensation or the equivalent — that is, insurance designed to protect workers against a wide range of job-site hazards, and in that setting a work-related illness such as typhoid fever from drinking contaminated water at the job site presents an attractive case for coverage, as in Aetna Life Ins. Co. v. Portland Gas Coke Co., 229 Fed. 552 (9th Cir. 1916), and H.P. Hood Sons, Inc. v. Maryland Casualty Co., 206 Mass. 223, 92 N.E. 329 (1910).

Maybe the tests are no good. That is what Justice Cardozo thought. Landress v. Phoenix Mutual Life Ins. Co., 291 U.S. 491, 495-96, 54 S.Ct. 461, 462, 78 L.Ed. 934 (1934) (dissenting opinion). He thought courts could do no better than to leave the question to "common understanding as revealed in common speech." Connelly v. Hunt Furniture Co., 240 N.Y. 83, 85, 147 N.E. 366, 367 (1925). See also Wickman v. Northwestern National Ins. Co., supra, 908 F.2d at 1085-86; Commercial Travelers Mutual Accident Ass'n v. Kilgore, 201 So.2d 486 (Fla.App. 1967). We agree. The insured ought to know what he is getting in his insurance policy, so that he can decide whether he would like more coverage at a higher price or less at a lower price. A lay person has a clear if inarticulate understanding of the difference between an accidental death and a death from illness, and that understanding will not be altered or improved by head-spinning judicial efforts at definition, such as the mysterious attempt by a court in this state to define disease as that which "originates from a source that is neither traumatic nor physical." Luttrell v. Industrial Commission, 154 Ill. App.3d 943, 955, 107 Ill.Dec. 620, 629, 507 N.E.2d 533, 542 (1987). A person can tell time without being able to define "time" and he can know how to ride a bicycle or shoot pool without being able to explain the principles of physics that enable him to do these things. He can also classify a death as the consequence of illness or accident without being able to define either term.

The plaintiff's counsel conceded at argument that if her decedent died from Crohn's Disease rather than from the migrating catheter (a question that cannot be answered on the basis of the scanty record compiled to date), she has no claim under the policy. The death would be due to illness rather than to accident. That it was sudden and unexpected would make no difference, International Travelers' Ass'n v. Ross, 292 S.W. 193 (Tex.Comm'n App. 1927) — many illnesses kill swiftly and without warning. The plaintiff's counsel conceded further that an "unavoidable" accident occurring in the course of medical treatment — that is, an accident not due to negligence or other malpractice — would not be covered. That concession is at once inevitable, under the ordinary-language approach that we think should govern these cases, and fatal. It is inevitable because no lay person would suppose that a person who died when the chemotherapy administered for his advanced cancer killed him because of his already weakened state had died because of "an accident." That would be a misuse of the English language. Medical treatment is often risky and when the risk materializes and the patient dies we do not call it dying in or because of an accident; it is death from sickness. The concession is fatal to the plaintiff's claim because the presence of negligence or other fault does not convert a medical or any other mishap into an accident. No one would suppose that if the insured had died in an airplane accident that was not anyone's fault the policy would not cover her. There are blameless ("inevitable") accidents and blameworthy accidents; blameworthiness does not make an untoward event "accidental." We do not know why the catheter drifted to and then through the decedent's heart. It may just have been one of those unfortunate events that happen in a small fraction of cases — for a standard complication of the procedure employed in this case (itself a standard procedure for treating the nutritional consequences of Crohn's Disease) is that the catheter may become detached and "embolize into the heart or pulmonary artery." Harrison's Principles of Internal Medicine, supra, at 431. When you die from the standard complications of standard medical treatments you don't, it seems to us, die in or because of an accident; your death is the result of illness.

It would not make sense to allow recovery in these cases. The causal ambiguity presented by the admittedly abbreviated record in the present case is apt to exist in most cases of death from medical treatment and make these cases unusually difficult to decide, which will just make insurance premiums higher (especially since such ambiguities may provoke squabbles between insurance companies — for example, a health insurer and an accident insurer). Medical treatments that involve a nontrivial risk of killing the patient ordinarily are reserved for patients suffering from grave illnesses, and if the patient dies it will often be profoundly unclear whether he died when he died because of the treatment, or because of the disease, or because of both; and if the first, because of some mistake, fairly describable as accidental, in the treatment. A related point is that the concession by plaintiff's counsel converts the accident insurance policy into a medical malpractice policy, requiring proof of negligence, thereby further complicating proof of coverage. The beauty of accident insurance is that you don't have to prove any wrongdoing. But under the plaintiff's interpretation, you do. This is another reason for doubting her interpretation. And still another is that an accident policy confined to a class of accidents that are caused by a tort gives the insured protection when he least needs it, since he would have a damages remedy in such a case.

The gray area between accident and illness is illustrated not by what happened here but by cases in which ordinary language equivocates between the accident and illness classifications (such as Landress, which involved sunstroke, a calamity intermediate between being struck by lighting — clearly an accident — and getting a skin cancer — clearly an illness), or in which an accident results in an infection (in Connelly an undertaker's helper got a scratch while handling a gangrenous corpse and germs entered his bloodstream through the scratch and killed him), or in which an accident occurs in the course of but unrelated to medical treatment — as when the ceiling of the operating room collapses while the insured is on the operating table, and kills him. The last case is closest (though not close) to this case and the defendants' counsel conceded at argument, quite properly in our view, that the policy would cover it. For the average person would say that the decedent had died in an accident in the hospital. The cause of death would be only adventitiously connected to the medical treatment; it is the sort of accident that might equally occur in one's home. Proof of causation would be straightforward; proof of negligence unnecessary. A medical mishap, whether or not caused by negligence, is different. We agree with the district judge that it is not an accident within the meaning either of the policy or of its brief description in the summary plan document.

AFFIRMED.