POSNER, Circuit Judge.
The appeal and cross-appeals in this diversity suit bring before us a difficult and important question of insurance law, one that we must decide under Illinois law but that has nationwide significance because it involves the interpretation of language that appears in the industry-wide standard-form policy known as Comprehensive General Liability Insurance. George H. Tinker, “Comprehensive General Liability Insur-[807]*807anee — Perspective and Overview,” 25 Federation of Ins. Counsel Q. 217, 218-21 (1975); Abex Corp. v. Maryland Casualty Co., 790 F.2d 119, 121 (D.C.Cir.1986). If a manufacturer sells a defective product or component for installation in the real or personal property of the buyer, but the defect does not cause any tangible change in the buyer’s property until years later, can the installation itself nonetheless be considered a “physical injury” to that property? The defective product or component in such a case is like a time bomb placed in an airplane luggage compartment: harmless until it explodes. Or like a silicone breast implant that is harmless until it leaks. Or like a defective pacemaker which is working fine now but will stop working in an hour. Is the person or property in which the defective product is implanted or installed physically injured at the moment of implantation or installation — in a word, incorporation — or not until the latent harm becomes actual?
The product at issue is a plumbing system, called “Qest,” that appellant Eljer’s U.S. Brass subsidiary manufactured and sold to plumbing contractors all over the United States between 1979 and 1986 (some have continued to be sold for installation in mobile homes, but we can ignore that detail). Between a half million and three-quarters of a million Qest systems were installed in houses and apartments, invariably behind walls or below floors or above ceilings, so that the repair or replacement of the unit requires breaking into walls, floors, or ceilings. Within a year after the first units were sold and installed, complaints about leaks began coming in and the first products-liability lawsuit was brought against U.S. Brass in 1982. By 1990 several hundred additional suits had been filed, some of them class actions, involving altogether almost 17,000 of the systems; and it is estimated that ultimately 5 percent of the total number of Qest systems ever sold will have failed in circumstances likely to provoke a tort suit. Eljer reckons U.S. Brass’s potential tort liability in the hundreds of millions of dollars, most of which Eljer hopes to recover from insurers — but only if it wins this case.
The tort claims are of two basic types— leak and nonleak — with the second requiring further subdivision:
(1) Claims that the system leaked, and caused water damage to the homeowner or forced him to repair or replace the system or deprived him of the use of his property (the leak, or the process of repair or replacement, may have made the home temporarily uninhabitable) or reduced the value of his home; or inflicted more than one of these harms.
(2) Claims that although the system has not yet leaked, (a) the value of the property in which it was installed has fallen as a result of the danger that the system may leak in the future, or the homeowner has as a precaution against that danger incurred the expense of replacing the system without waiting for it to leak; (b) the homeowner has been deprived of the use of his property while his Qest system was being replaced; (c) merely the ominous presence of the nonleaking, but perhaps some day destined to leak, Qest has deprived the homeowner of the use of his home (he’s afraid to turn on the water) — as when a building is evacuated in response to a bomb threat.
Liberty Mutual, the principal defendant, issued a series of annual Comprehensive General Liability Insurance policies to Eljer covering the years 1979 through 1988. Travelers, the other defendant, provided excess coverage for Eljer between 1982 and 1986. The policies cover liability for “property damage” accidentally caused by Eljer or its subsidiaries, defined as “(1) physical injury to or destruction of tangible property which occurs during the policy period, including the loss of use thereof at any time resulting therefrom, or (2) loss of use of tangible property which has not been physically injured or destroyed provided such loss of use is caused by an occurrence [defined, so far as relevant here, as an accident, including continuous or repeated exposure to conditions] during the policy period.” This language was changed in the last three policies (covering 1986 through 1988), but, the parties agree, not materially.
[808]*808Eljer brought this declaratory judgment suit to establish that the physical injury to the property of the buyer of a Qest system occurs when the system is installed in the buyer’s house or apartment, not when it begins to leak or is replaced or is recognized to have reduced the value of the buyer’s property. If this is right, essentially all the “property damage” inflicted by the defective systems occurred during the years in which Liberty’s insurance policies were in force, because those were the years in which the systems were sold and installed. Liberty, however, contends that with respect to Eljer’s liability (actually U.S. Brass’s liability, but we can disregard that detail too) for leaking systems, the “property damage” to which the policies refer does not occur until the system leaks. If it hasn’t leaked, but has been replaced in anticipation of its proving defective, or the value of the home has diminished, or the homeowner has lost the use of his property because he’s afraid to use the plumbing, then, provided at least one other Qest system somewhere else has leaked, there is “property damage” as of the time when the homeowner realizes that the Qest is reducing the value of his home and (or) must be replaced. Liberty argues in the alternative that in the nonleaking case the physical damage occurs when the homeowner replaces his Qest system, since the process of removing the old system causes unquestioned damage to the house.
Many of the leaks upon which tort claims against Eljer are based did not occur until after the expiration of Liberty’s last insurance policy at the end of 1988. And many people did not realize till then that their Qest systems were potentially defective and reducing the value of their homes and ought to be replaced and maybe the home not used until it was replaced because the water should be shut off immediately as a precaution. So if the insurance companies’ arguments prevail, Eljer will not be able to shift a substantial part of the burden of the tort claims against it to Liberty — or to Travelers, whose coverage expired even sooner.
The district judge agreed with the insurance companies that property damage within the meaning of the Comprehensive General Liability Insurance form contract does not occur upon installation of the defective system, except in the case of claims for loss of use where there is no leak (categories (2)(b) and (2)(c) in our typology). In the case of claims that arise from leaks, property damage occurs, in the district judge’s view, when the leak occurs, while in the case of claims for the cost of repair or replacement incurred in anticipation of a leak it occurs at the time of repair or replacement. Eljer challenges these rulings. Liberty and Travelers cross-appeal, arguing mainly that the judge should not have ruled on loss of use claims when there is no leak, because the issue was not presented to him. The insurance companies are correct. Eljer failed to develop evidence or argument concerning this rather esoteric category of claims. It is not even clear that any claims of type (2)(c) have been or will be lodged against it — for how many people will vacate their home because they’re afraid the pipes will leak unless the water is turned off? — in which event there might not even be an actual controversy between Eljer and its insurers over reimbursement for such claims. There are (2)(b) claims — loss of use of property incident to the replacement of a possibly defective Qest system. But because of the absence of evidence or argument concerning loss of use claims, the insurance companies, which may have had grounds for showing that such claims are within a policy exclusion, did not have a fair chance to defend themselves. Burdett v. Miller, 957 F.2d 1375, 1380 (7th Cir.1992). Eljer is not entitled to a declaratory judgment concerning such claims.
The central issue in the case — when if ever the incorporation of one product into another can be said to cause physical injury — pivots on a conflict between the connotations of the term “physical injury” and the objective of insurance. The central meaning of the term as it is used in everyday English — the image it would conjure up in the mind of a person unschooled in the subtleties of insurance law — is of a [809]*809harmful change in appearance, shape, composition, or some other physical dimension of the “injured” person or thing. If water leaks from a pipe and discolors, a carpet or rots a beam, that is physical injury, perhaps beginning with the very earliest sign of rot — the initial contamination (an important question in asbestos cases, as we shall see). The ticking time bomb, in contrast, does not injure the structure in which it is placed, in the sense of altering the structure in a harmful, or for that matter in any, way — until it explodes. But these nice, physicalist, “realistic” (in the philosophical sense) distinctions have little to do with the objectives of parties to insurance contracts. The purpose of insurance is to spread risks and by spreading cancel them. Most people (including most corporate executives) are risk averse, and will therefore pay a premium to avoid a small probability of a large loss. Once a risk becomes a certainty — once the large loss occurs — insurance has no function. The last point at which a Qest plumbing system has an insurable risk of being defective and causing harm is when it is installed. When it starts to leak is too late; the risk has turned into a certainty and cannot be spread by being insured. It may also be too late when so many other Qest systems have failed that the unknowable risk has become a statistical certainty (or nearly so). For the individual homeowner there is still uncertainty as to whether and when his Qest will leak, so he may still be able to buy insurance against such an eventuality, though the premium will be very high because the risk is very great. For Eljer, however, which is liable for all the leaks traceable to the defectiveness of the system (as distinct from mistakes in installation, or other causes for which Eljer is not responsible), the pooling of the risks of the individual homeowners has produced a near certainty of a loss in excess of $100 million. How much in excess no one yet knows, and this leaves some room for insurance. But we are told that Eljer has had no primary liability coverage since the last Liberty policy expired at the end of 1988 and no excess coverage since the last Travelers policy expired at the end of 1986.
It would be an overstatement to say that the insurance policies which Eljer- bought from the defendants had no value unless property damage is deemed to occur in the policy year in which the system is installed rather than in the policy year — more likely post-policy year — in which it fails or is so likely to fail that the prudent homeowner replaces it (perhaps vacating the home until it is replaced) or the market writes down the value of the home to reflect the risk of the system’s failing and causing water damage and loss of use. Some systems malfunctioned in the year of installation; there has therefore been some payout under the insurance policies. And, because of limited liability, corporations have less incentive to buy insurance than private individuals do, David Mayers & Clifford W. Smith, Jr., “On the Corporate Demand for Insurance,” 55 J. Bus. 281 (1982); maybe limited coverage for modest premiums was all Eljer wanted or bought. But the defendants make nothing of the special character of the corporate insurance market — nor is it clear just how special it is, given risk aversion by executives and other employees, and the costs of bankruptcy (id. at 283-85) — instead treating this case as if Eljer had been an individual who had been seeking liability insurance. In these circumstances we should not be quick to assume that the standard general liability policy issued by the American insurance industry provides what to an outsider at any rate appears to be largely though not completely illusory coverage in an important class of cases. Sears, Roebuck & Co. v. Reliance Ins. Co., 654 F.2d 494, 499 (7th Cir.1981); Keene Corp. v. Ins. Co. of North America, 667 F.2d 1034, 1046-47 (D.C.Cir.1981); Hancock Laboratories, Inc. v. Admiral Ins. Co., 777 F.2d 520, 525 (9th Cir.1986).
Nor are literal interpretations, which assume that words or phrases are always used in their ordinary-language sense, regardless of context, the only plausible interpretations of contractual language, especially when the contract is between sophisticated business entities. We should at least ask what function “physical injury” [810]*810might have been intended to perform in a comprehensive general liability policy beyond just drawing a commonsensical, lay person’s, ordinary-language distinction between physical injury and physical noninju-ry.
Apparently the term was intended to distinguish between physical and nonphysical injuries rather than to distinguish between injuries and noninjuries. The 1966 version of the General Comprehensive Liability Insurance policy form had defined “property damage” as “injury to or destruction of tangible property.” Most courts, in apparent harmony with the drafters’ intentions, Willard J. Obrist, The New Comprehensive General Liability Insurance Policy: A Coverage Analysis 7, 17 (Defense Research Institute, Inc. 1966), gave the term “injury” in the new definition a broad interpretation. Elco Industries, Inc. v. Liberty Mutual Ins. Co., 90 Ill.App.3d 1106, 1109-10, 46 Ill.Dec. 319, 323, 414 N.E.2d 41, 45 (1980). Some, however, excluded injuries in which there was no physical contact between the injurer and the property injured by him. (Sola Basic Industries, Inc. v. United States Fidelity & Guaranty Co., 90 Wis.2d 641, 280 N.W.2d 211 (1979), reviews the case law.) For example, if a manufacturer of construction cranes sold a defective crane which collapsed in front of a restaurant, blocking access to it and thereby impairing the restaurateur’s income, and the restaurateur sued the manufacturer and recovered a judgment, the manufacturer’s liability insurer might not have to pay, because the blocking of access might not be considered an injury to tangible property. Tinker, supra, at 232. To allay doubts on this score by sharpening the aim of the 1966 definition, the insurance industry’s committee charged with updating the Comprehensive General Liability Insurance policy form redid the definition in 1973, producing the two-part definition that we quoted earlier from Liberty’s policies. The second part, which is new, explicitly covers the injury inflicted by our hypothetical crane manufacturer — a “loss of use of tangible property which has not been physically injured or destroyed.” The first part of the new definition is the old definition with “physical” prefixed to “injury” to distinguish the two parts. Both cover injury, but the second part covers injury which is not “physical” because there is no physical touching of the tort victim’s property. Society of Chartered Property and Casualty Underwriters, 1973 CGL Changes 9 (1971); Fred L. Bardenwerper & Donald J. Hirsch, General Liability Insurance—1973 Revisions 11, 32 (Defense Research Institute, Inc. 1974); Guide to Liability Insurance 4 (Rough Notes Co. 1973). There was no intent to curtail liability in a case of physical touching, as where a defective water system is installed in a house.
We can now see more clearly that two senses of “physical injury” are competing for our support. One, which the insurers want us to adopt, is an injury that causes a harmful physical alteration in the thing injured. The other, which is what the draftsmen of the Comprehensive General Liability Insurance policy apparently intended and what rational parties to such a policy would intend in order to make the policy’s coverage real and not illusory, is a loss that results from physical contact, physical linkage, as when a potentially dangerous product is incorporated into another and, because it is incorporated and not merely contained (as a piece of furniture is contained in a house but can be removed without damage to the house), must be removed, at some cost, in order to prevent the danger from materializing. There is an analogy to fixtures in the law of real and personal property — improvements to property that cannot be removed without damaging it. See, e.g., UCC § 9-313. In effect we are being asked to choose between “physical injury” and “physical injury.”
Human beings give structure to their experience — cut the world of perception up into usable slices — through language, in accordance with the needs and interests of the moment. For many purposes the distinction between touching and altering is important; if two automobiles collide, it makes a big difference whether they merely touch each other or dent each other. For other purposes, other distinctions are [811]*811more important, such as that between harms from touching, whether or not some harmful physical change ensues, and harms from other interferences with valued interests. Battery is a tort of touching, and, provided the touching is unpermit-ted and offensive, the fact that the victim was not bruised or wounded goes only to the issue of damages. Assault is not a tort of touching — it becomes battery if the assailant touches his victim, as distinct from making a merely threatening gesture. Physical contact, not physical injury in the layman’s sense, distinguishes the two torts. W. Page Keeton et al., Prosser and Keeton on the Law of Torts §§ 9-10, at pp. 39-46 (5th ed. 1984).
Within the class of unintentional torts, too, the distinction between touching and not touching is often made. Under traditional tort law, if you hit a bridge and put it out of commission, you are liable to the owner of the bridge but not to merchants who lose business because customers cannot reach them with the bridge down. Rickards v. Sun Oil Co., 23 N.J.Misc. 89, 41 A.2d 267 (1945). You have not touched the merchants or their property. The doctrine of “economic loss,” illustrated by the Rickards case (and see Petitions of Kinsman Transit Co., 388 F.2d 821, 825 n. 8 (2d Cir.1968); Barber Lines A/S v. M/V Donan Maru, 764 F.2d 50 (1st Cir.1985); Robert L. Rabin, “Tort Recovery for Negligently Inflicted Economic Loss: A Reassessment,” 37 Stan.L.Rev. 1513 (1985)), makes touching a prerequisite to tort liability for damage to property and thus rules out loss-of-use claims by persons with whose property the tortfeasor had made no physical contact.
The evolution of the bipartite structure of the Comprehensive General Liability Policy may have been shaped by this distinction between physical and economic loss (a poor choice of words — all the losses for which tort victims sue are economic). At first the distinction was so ingrained in tort law that there was little demand for liability insurance on the part of injurers who had not physically touched their victims, because such an injurer would have no liability against which to insure. The bar to this sort of liability weakened — though gradually, haltingly, and only in some jurisdictions — through an increasing recognition of exceptions. Barber Lines A/S v. M/V Donau Maru, supra, 764 F.2d at 55-56; Union Oil Co. v. Oppen, 501 F.2d 558 (9th Cir.1974); J’Aire Corp. v. Gregory, 24 Cal.3d 799, 157 Cal.Rptr. 407, 598 P.2d 60 (1979); People Express Airlines, Inc. v. Consolidated Rail Corp., 100 N.J. 246, 495 A.2d 107 (1985); Keeton et al., supra, § 129, at pp. 1001-02. The demand for insurance against liability for loss of use unaccompanied by physical injury in the traditional sense rose concomitantly. It is not surprising that a new definition of property damage — a definition that encompassed two forms of injury, one physical and one not — emerged, and that coverage was extended to the second.
We cannot be confident, given a sparse drafting history, that the evolution of the definition of property damage is, as we have just conjectured, a response to the changing vicissitudes of the economic-loss doctrine. Although the reference to the crane case suggests that it is, for accidents that cause merely delay are standard examples of the operation of the doctrine, Petitions of Kinsman Transit Co., supra, 388 F.2d at 825 n. 8, the timing is off. The cases we have cited for the gradual erosion of the doctrine were decided after the definitional changes, though they may have been anticipated before. It is not even clear that the suits against U.S. Brass for loss of the use of the property in which the Qest systems were installed are affected by that doctrine. The concern behind it is with unanticipated plaintiffs and disproportionate liabilities, and the rubrics under which recovery is denied are duty and proximate cause. A homeowner who replaces a plumbing system that is likely to fail could be thought to be trying to mitigate the consequences of the supplier’s having supplied a defective product, and thus to be curtailing rather than expanding liability. If so, however, denial of insurance coverage would be arbitrary; it would exploit the adventitious circumstance that a “phys[812]*812ical injury” in the purely conventional sense of the word had not yet occurred.
All this may seem an aside, since these appeals (we have held) do not involve loss of use, but only physical injury. The significance of the loss of use provision, however, lies in what it tells us about the meaning of the physical injury provision. Whatever the precise function of the loss of use provision — whatever its precise relation to the economic loss doctrine — it was not added, and injury redescribed as physical injury, in order to curtail the preexisting insurance coverage for injury to tangible property. That coverage was and remains broad. Not promiscuously so. We do not think that every time a component part fails, the resulting injury can be backdated to the date of installation or incorporation. The expected failure rate must be sufficiently high to mark the product as defective — sufficiently high, as is alleged to be the case regarding the Qest plumbing system, to induce a rational owner to replace it before it fails, so likely is it to fail. That condition was satisfied. We believe that the coverage provided by Comprehensive General Liability Insurance encompassed (loss of use to one side) the tort claims against U.S. Brass.
So far we have been discussing the case as if it had to be decided on the basis of first principles, and that is nearly true because the cases are unclear. The parties agree that New York law governs some of the policy years and Illinois law the rest, and they further agree that the relevant law of the two states is the same. The New York law on the question is particularly sparse. We know that the New York courts accepted the incorporation concept of property damage under the 1966 definition in the Comprehensive General Liability Insurance policy. Sturges Mfg. Co. v. Utica Mutual Ins. Co., 37 N.Y.2d 69, 371 N.Y.S.2d 444, 332 N.E.2d 319 (1975); Marine Midland Services Corp. v. Samuel Kosoff & Sons, Inc., 60 App.Div.2d 767, 400 N.Y.S.2d 959 (1977). But we have no cases interpreting the 1973 definition, the one at issue here. The closest are cases on negligent installation. Greenlee v. Sherman, 142 App.Div.2d 472, 536 N.Y.S.2d 877 (1989); Holmes Protection of New York, Inc. v. National Union Fire Ins. Co., 152 App.Div.2d 496, 543 N.Y.S.2d 459 (1989), and Young v. Ins. Co. of North America, 870 F.2d 610 (11th Cir.1989) (per curiam) (applying New York law). Young is representative. The builder of a house installed a gas barbecue grill in the course of construction. The installation was negligent, resulting in a fire years later that destroyed the house, and the court held that physical damage within the meaning of the insurance policy occurred at the time of the fire, not of the installation. The analogy to our case is plain. But neither in Young nor in the other two cases did the court consider the incorporation theory of Sturg-es and Kosoff and whether it had survived the 1973 redefinition of property damage.
We get a little more help from the Illinois cases, although the parties spend too much time discussing United States Fidelity & Guaranty Co. v. Wilkin Insulation Co., 193 Ill.App.3d 1087, 140 Ill.Dec. 907, 550 N.E.2d 1032 (1989), aff'd, 144 Ill.2d 64, 161 Ill.Dec. 280, 578 N.E.2d 926 (1991), where the issue was whether the installation of insulation containing asbestos in schools and other public buildings was property damage. The Illinois Appellate Court held that it was. Interpretation of the opinion is complicated by the fact that the case involved both policies that contained the 1966 definition of property damage and policies that contained the 1973 definition. As to the first, the court unambiguously held that the installation of the defective insulation (defective because it contained asbestos which might leak) was “injury to ... tangible property.” It then discussed whether it was also “physical injury” within the meaning of the first part of the 1973 definition, and here the opinion turns murky. The court begins by listing the cases on which the insurers rely. It distinguishes them at great length, implying that it rejects the insurers' argument. But later the court veers off, remarking, “Even assuming, arguendo, that the buildings had not sustained physical injury,” the defendants may be liable under the second part of the 1973 definition. 193 Ill.App.3d [813]*813at 1097, 140 Ill.Dec. at 913, 550 N.E.2d at 1038. In the course of distinguishing the insurers’ cases the court notes the continuous release of asbestos fibers; and that is unlike the ticking time bomb case (or this case), where the defective product is completely dormant before the explosion (or leak). But the weight of this factor in the court’s decision is unclear; the statement of facts does not mention the emission of asbestos fibers.
The Supremé Court of Illinois affirmed, but on the ground “that asbestos fiber contamination constitutes physical injury to tangible property.”' 144 Ill.2d at 75, 161 Ill.Dec. at 285, 578 N.E.2d at 931 (emphasis added). The insurers in our case treat this as a tacit repudiation of the Illinois Appellate Court’s incorporation theory of physical injury. And they point out that the Qest plumbing system does not “contaminate” the home in which it is installed until it starts to leak. But as the theory of thé Illinois Appellate Court’s decision is unclear,- so is it unclear what if anything the supreme court thought it was repudiating; and of course the adoption of one ground need not imply the repudiation of an alternative ground. The fact of contamination made Wilkin an easier case than ours; it does not follow that the supreme court would decide the more difficult case another way.
The closest case to ours is Marathon Plastics, Inc. v. International Ins. Co., 161 Ill.App.3d 452, 112 Ill.Dec. 816, 514 N.E.2d 479 (1987). Marathon, the insured, sold pipes for installation in an underground water system. The pipes were defective, which caused the system to leak. The leaks caused no physical damage, but of course they impaired the operation of the water system, and Marathon was sued. The court held that Marathon was covered by its liability policy, because its tort liability was for inflicting property damage within the meaning of the 1973 definition. The court relied on earlier Illinois cases, notably Elco Industries, Inc. v. Liberty Mutual Ins. Co., supra, and was explicit that the absence of physical injury in the ordinary sense to the water system was immaterial; it was enough that the leaks reduced the system’s value. 161 Ill.App.3d at 463, 112 Ill.Dec. at 822, 514 N.E.2d at 485.
We can find .no contrary authority in Illinois. The cases on which the insurers rely do not involve incorporation. The insurers place particular weight on Bituminous Casualty Corp. v. Gust K. Newberg Construction Co., 218 Ill.App.3d 956, 161 Ill.Dec. 357, 578 N.E.2d 1003 (1991), because it was decided after the Supreme Court of Illinois decided Wilkin. It is true that the insured lost in Bituminous Casualty, but we can find nothing in the opinion to suggest that the court thought the supreme court had repudiated the incorporation analysis of Elco and Marathon. The underlying tort suit was for the improper design, manufacture, and installation of the heating, ventilating, and air-conditioning system in the State of Illinois Building in Chicago. It was not an incorporation case because there was no suggestion that the system had reduced the value of the building or could not be removed without damage to the building — actually no one wanted it removed, only fixed. There is a danger, which the court was alert to resist, of stretching the concept of incorporation to the point where every defective product is deemed to injure some piece of tangible property, for almost every product is used in conjunction with something else. If the insured had won, a defective light bulb could be described as causing physical injury to the room in which it was flickering. The issue in Bituminous Casualty was not whether the incorporation doctrine is part of Illinois law; that was taken for granted, notwithstanding the supreme court’s decision in Wilkin. The issue was the scope of the doctrine, and is not presented by this case. But we see that the word “physical” in the definition of property damage serves an important limiting purpose even if it is not interpreted to require a change in the property with which the tortfeasor comes in contact.
That covers the Illinois cases, and since the parties agree that Illinois law is authoritative, there is no reason to look further — and not much to find there. It is [814]*814true that several cases, illustrated by New Hampshire Ins. Co. v. Vieira, 930 F.2d 696 (9th Cir.1991), and American Home Assurance Co. v. Libbey-Owens-Ford Co., 786 F.2d 22, 25-26 (1st Cir.1986), read the 1973 definition of property damage restrictively. But they do so without examining the considerations — the drafting history of the provision and the purposes of insurance— that argue against such a reading. So we doubt that the Illinois cases are mere sports likely to be swept away by the state supreme court. Of course we cannot be certain that that court, should it have a case like this, would not sweep aside the whole line of incorporation decisions of the appellate court. The U.S. Supreme Court frequently sweeps aside a line of considered decisions in the federal courts of appeals. E.g., McNally v. United States, 483 U.S. 350, 107 S.Ct. 2875, 97 L.Ed.2d 292 (1987); see id. at 376, 107 S.Ct. at 2890 (dissenting opinion). Even when a state’s intermediate appellate decisions are uniform, or at least reasonably so, as they are here, we are not bound by them. Indiana Harbor Belt R.R. v. American Cyanamid Co., 916 F.2d 1174, 1176 (7th Cir.1990) (dictum). But we do need, in such a case, a reason for predicting that the state’s supreme court will reject the intermediate decisions, West v. American Tel. & Tel. Co., 311 U.S. 223, 237-38, 61 S.Ct. 179, 183-84, 85 L.Ed. 139 (1940); Diesel Service Co. v. AMRAC Int’l Corp., 961 F.2d 635, 639 n. 3 (7th Cir.1992); Tippecanoe Beverages, Inc. v. S.A. El Aguila Brewing Co., 833 F.2d 633, 638-39 (7th Cir.1987)—as we did in Williams, McCarthy, Kinley, Rudy & Picha v. Northwestern National Ins. Group, 750 F.2d 619, 624-25 (7th Cir.1984), but do not here. On the contrary, the drafting history of the property-damage clause, and the probable understanding of the parties to liability insurance contracts, persuade us that the incorporation of a defective product into another product inflicts physical injury in the relevant sense on the latter at the moment of incorporation — here, the moment when the defective Qest systems were installed in homes.
The judgment of the district court is reversed, and the case remanded for the entry of a judgment for the plaintiff that will be consistent with this opinion.
Reversed and Remanded.