dissenting.
The risk that a dairy product will contain salmonella is one of the constants of the business. The risk will vary with the care the participants in the process take. The contract in this ease required the sellers (collectively Valley Lea) to take scrupulous care when processing the milk. In order to protect itself from a number of risks — the risk that Valley Lea would fail to take the necessary precautions, the risk that some contamination would enter the milk in the drying process, and the risk that milk would become contaminated en route— General Foods tested the dry milk before using it. The instructions in this case permitted the jury to infer from General Foods’ knowledge of the general risk of contamination, and from its careful testing, that it “incurred” the risk of contamination and so cannot recover. I believe that the law of Indiana does not permit a jury such latitude. The district court should have given General Foods’ proposed instruction No. 50 rather than its own instruction No. 27.
This is a contract case. General Foods seeks damages for breach of warranty. It wants compensation for consequential damages caused by a concealed defect in a product it purchased. We should start with the rules of contract law contained in Indiana’s version of the Uniform Commercial Code. The UCC provides that a buyer’s acceptance of goods does not alter warranties that apply to the goods, when the defect in the goods is latent. Ind.Code § 26-1-2-316. (I omit the “Ind.Code § 26-1” from other citations to the UCC.) The buyer may recover consequential damages when the seller knows the use to which the goods will be put and the buyer could not reasonably proceed by obtaining some other supplier’s products (by “cover”). UCC § 2-715(2)(a). A buyer’s inspection does not prevent the buyer from recovering damages caused by latent defects. Richards v. Goerg Boat & Motors, Inc., 179 Ind.App. 102, 384 N.E.2d 1084 (1979). Older cases hold the same. Rayl v. General Motors Corp., 121 Ind.App. 608, 101 N.E.2d 433 (1951) (en banc); Teter v. Schultz, 110 Ind.App. 541, 39 N.E.2d 802 (1942); Bank of Kansas City v. Grind-staff 45 Ind. 158 (1873). Under UCC § 2-607(2) the acceptance of goods does not prevent the buyer from recovering damages for the goods’ non-conformity to the contract. See also UCC § 2-608(l)(b) and comment 1, which allows the buyer to revoke acceptance belatedly and hold the seller liable. Hudson v. Dave McIntire Chevrolet, Inc., 180 Ind.App. 646, 390 N.E.2d 179 (1979).
The UCC does not refer to “incurred risk.” Neither do cases decided under the *1102UCC. The doctrine is an offshoot of the Restatement (Second) of Torts § 402A, comment n, which states that a user of a defectively designed good may not recover if he is aware of the defect and nonetheless proceeds unreasonably. The doctrine was not designed for contract cases. I recognize that some Indiana cases have applied it in contract cases without any explanation, and that this binds a federal court hearing a diversity case. See Gregory v. White Truck & Equipment Co., 163 Ind. App. 240, 323 N.E.2d 280 (1975). Nonetheless there is no reason why a federal court should extend the doctrine to foreclose liability that would otherwise be supported by principles of contract law.
So far as I can tell from the confusing and internally contradictory cases announcing and applying the doctrine of incurred risk, no court of Indiana has relieved a seller of liability for a latent defect. The cases say instead that “incurred risk” is a defense to liability when a person knows a particular risk and acts anyway — a doctrine almost antithetical to avoiding liability on the basis of a latent defect. Kroger Co. v. Haun, 177 Ind.App. 403, 379 N.E.2d 1004, 1008 (Ind.App.1978), on which the majority principally relies, states that the doctrine depends on “the perception and voluntariness of a risk and is blind as to the reasonableness of risk acceptance.”
All of the cases that have discussed the doctrine involve risks that are apparent. In Moore v. Moriarty, 415 N.E.2d 779 (Ind.App.1981), for example, one party was pulling a disabled truck with a tractor, while the other steered the truck. The person in the cab of the truck failed to apply the brakes at the right time, injuring the driver of the tractor. The court thought it inappropriate as a matter of law to apply the doctrine of “incurred risk.” One driver could not know that the other would fail to apply the brakes — although he could appreciate the risk that a tow may lead to a mishap. See also, e.g., Power v. Brodie, 460 N.E.2d 1241, 1243 (Ind.App.1984), explaining that “incurred risk” requires “much more than the general awareness of a potential for mishap ... [and entails] acceptance of a specific risk of which the plaintiffs had actual knowledge.”
When the defect in a product is latent, the state has declined to apply the doctrine. See Moore v. Federal Pacific Electric Co., 402 N.E.2d 1291 (Ind.App.1980), in which the plaintiff was electrocuted when he put a screwdriver in switchgear that he knew was energized. He knew of the risk of electrocution; he did not know, however, that a defect in the switchgear made the risk higher than he thought. The court held the doctrine of “incurred risk” applicable “only if Moore had actual knowledge of the danger or defect caused by the Defendants” (Id. at 1293) — in other words, of the defect in the product’s design, not of the general risk of electrocution when you put a screwdriver in a live switch carrying 450 voits.
The sort of risk General Foods understood and accepted here is the sort of risk any user of dairy products understands and accepts — that the product may be contaminated with salmonella. General Foods tried to protect itself from this risk. But it did not accurately understand the degree of risk it faced. Under Indiana law awareness that any product may be defective, and the adoption of a plan of testing to reduce the costs of such defects, does not permit a jury to find that the buyer “incurred” the risk. There is no basis for allowing the jury to find “incurred risk” here.
The majority holds that any problem in the first paragraph of the instruction on incurred risk was overcome by the second paragraph, which says that the defense applies if General Foods “actually knew of the danger and proceeded unreasonably” in using the dry milk. The initial question is what “the danger” means — the danger that any dry milk will be contaminated, or some particular level of danger? A subsidiary question is what it means to proceed “unreasonably” in using dry milk.
The contract between General Foods and its suppliers attempted to allocate duties of quality control. Valley Lea undertook to *1103use the' best methods of sanitation and production; General Foods undertook to test the milk before using it. Each party had a duty of care. Because each could contribute to the loss, this made sense. The seller’s care reduced the amount of contamination, and the buyer’s care reduced the chance that accidentally-contaminated milk would cause loss. If the buyer took no care (for example, skipping the testing) or if the buyer used milk that it thought likely to be contaminated, a jury could find the conduct unreasonable in the sense of tort law. If a person neglects to take precautions that cost less than the injury (if one comes to pass), discounted by the improbability that there will be an injury, that person has behaved unreasonably. United States Fidelity & Guaranty Co. v. Jadranska Slobodna Plovidba, 683 F.2d 1022, 1025-26 (7th Cir.1982); United States v. Carroll Towing Co., 159 F.2d 169, 173 (2d Cir.1947) (L. Hand, J.).
General Foods tested the milk so thoroughly that there was less than a one-in-twenty chance that as much as 5% of the powder would contain undetected salmonella contamination. It is difficult to see how using the milk in an ordinary commercial way could be unreasonable even under the tort standard. The loss in this case came to about $1.2 million. If it is appropriate to interpret the results of the testing as putting an upper limit of 5% on the probability of loss, the expected loss from using the milk was less than $60,000 ($1,200,000 x 0.05 = $60,000). The alternative was to leave the plant idle while seeking another source of milk, which could have imposed commercial losses substantially exceeding that sum. (The parties agree that roller dried powdered milk was not readily available, making “cover” difficult.) Perhaps even on these numbers a jury could have found unreasonable conduct if General Foods had mixed Valley Lea’s milk with that of another supplier for an unusually large production run, thereby increasing the potential loss, but apparently General Foods did not do this.
As it turns out, the results of the testing were seriously misleading. The testing was based on the assumption that the probability of contamination in any one lot was independent of the probability of contamination in another. General Foods assumed, in other words, that contamination would be accidental rather than systematic. To find salmonella in one lot does not give much information about another, if the contamination was introduced by chance after manufacture. General Foods had every reason to believe that the risks were independent. Its contract called for Valley Lea to use the best sanitary methods and to clean all equipment thoroughly after each lot. This would ensure independence. So far as General Foods knew, there was no history of salmonella contamination at Valley Lea.
General Foods had been lied to. Valley Lea had found contamination before; so had federal inspectors. Valley Lea’s officers not only withheld this information from General Foods but also lied when General Foods put the question point blank after finding contamination in lot 329. A memorandum in evidence contains a pledge by officers of Valley Lea to hide the information from General Foods.
Valley Lea also did not routinely clean the equipment after production of each lot, so that contamination might recur. To make matters worse, Valley Lea’s equipment had been set up so that a leaky valve allowed unpasteurized milk to enter the finished product. (There is some suggestion that unpasteurized milk was used deliberately in order to increase the capacity of the plant.) This dramatically increased the risk of contamination, and it linked the risk of one shipment with the risk of another. Valley Lea did not inform General Foods about the failure to clean up or the use of unpasteurized milk.
The knowledge that whatever went wrong with lot 329 probably had gone wrong with the other lots in the shipment would have cast new light on the interpretation of the tests. The samples were not independent. General Foods was taking a much bigger risk than it knew. It says, *1104and Valley Lea does not deny, that it would not have used an ounce of Valley Lea’s milk had it known how Valley Lea was operating its plant; it concedes that it would have been unreasonable to use milk after finding the problem in lot 329 if it had known that the risks of the lots were dependent rather than independent. Yet the instruction did not inform the jury of the significance of dependence versus independence to a judgment that General Foods “incurred” a known risk.
There is, perhaps, another sense in which General Foods “incurred” a risk. It specified that the milk was to be roller dried rather than spray dried. All agree that roller dried milk is exposed to the air longer, and so there is a greater risk of contamination at the processing stage. General Foods took this risk because it believes that roller dried milk gives its chocolate a better taste. It might be a sensible interpretation of the contract — of trade practice modifying the written terms — to say that General Foods accepted all of the increased risk that came from selecting the roller drying process. Cf. Western Industries, Inc. v. Newcor Canada Ltd., 739 F.2d 1198 (7th Cir.1984) (discussing modification of warranties through trade custom). General Foods also willingly incurred any risks associated with its own manufacturing processes, which might argument the loss from the occasional inevitable introduction of salmonella. Here, for example, General Foods shipped the chocolate in liquid state before obtaining the results of its final tests for contamination, and a jury might sensibly have charged to General Foods the costs created by this.
But there is no evidence from which a jury could have found that the problem in this case arose from the selection of the roller rather than the spray drying process. The evidence points to either reckless or deliberate conduct at Valley Lea that permitted unpasteurized milk to enter the product. General Foods’ specification of roller dried milk did not create this risk.
The instructions here were thoroughly unsatisfactory. The jury properly could have been told that General Foods incurred the risk of roller dried milk, that it incurred the costs of unusual production runs or immediate shipment. It did not, however, incur the risks of loss from unpasteurized milk or uncleaned equipment. It did not know that it was taking these risks, and under Indiana law knowledge of the specific character of the risk is an essential element of the defense. The instructions should have separated the risk of roller dried milk from the risk of other milk, the background risk common to all milk from the risk of unpasteurized milk. They should have informed the jury that it was commercially reasonable to use the milk once the tests came out as they did — unless General Foods knew that the risks were dependent rather than independent.
This is a genuinely hard case, however, because General Foods did not tender instructions that made these distinctions explicitly. In a civil case we cannot fault a judge for failing to invent distinctions that the parties did not incorporate in their proposals.
When each party tenders an incomplete or misleading instruction, the judge must either write a correct instruction or choose the one that comes closer to being right. Chase v. Consolidated Foods Corp., 744 F.2d 566 (7th Cir.1984). General Foods’ proposed instruction No. 50 was significantly closer to being right. It would have told the jury that the defense depends on knowledge of a “specific” risk, and that the general risk of salmonella contamination would not be enough. It also required knowledge that the powder was “probably contaminated” — too high a requirement in any test of commercial reasonableness, although arguably justified by the state’s unwillingness (so far) to apply the doctrine of incurred risk to latent defects. This proposal was therefore not free from flaw, but it was closer than the nebulous No. 27 that the jury heard.
No. 27 may too easily be read to penalize diligence. If General Foods had simply dumped the powder into its vats without any testing, it would have recovered on its *1105warranty. The more carefully General Foods tests, the more conscious it is of risks, the less able it is to recover for undetected poison in the food. As a general principle, that must be wrong.