dissenting.
I respectfully dissent. A new trial in this case is not justified. The record before us contains no evidence to support a judgment against Torrington on plaintiffs’ voluntary-undertaking theory or against Textron for product liability. I would render judgment that the plaintiffs take nothing.
*853The plaintiffs did not plead a voluntary-undertaking claim against Torrington. They may not even have thought of this theory of recovery until after the trial was over and a verdict had been returned. But assuming that the claim was somehow tried by consent and not simply by accident, I agree with the Court that the judgment against Torrington must be reversed because the jury was not told what the essential elements of the claim are or asked to decide whether the plaintiffs had proved all of those elements. I disagree that the case should be remanded for a new trial in the interest of justice. I have two reasons.
First: justice does not require that the plaintiffs be given a second trial to do what they should have known to do in the first trial, which was to ask the jury whether they proved each element of their voluntary-undertaking claim. We have sometimes remanded a case for a new trial when parties were reasonably confused about the substantive law,1 but this is not such a case. According to the Court, the elements of a voluntary-undertaking claim are clearly stated in section 323 of the Restatement (Second) of Torts, published in 1965, and in this Court’s 1976 opinion in Colonial Savings Association v. Taylor.2 Findings of negligence and causation — the only findings requested by the plaintiffs in this case — are necessary but not sufficient to establish liability on a voluntary-undertaking claim. As the Court states: “We have never held that a person may be liable on an undertaking theory without establishing rebanee or increased risk of harm, and we decline to do so now.”3 The plaintiffs in this case established neither reliance nor increased risk and requested no jury findings on these issues. Why then should they be given a new trial?
Because, the Court answers, the plaintiffs could not have known before today’s decision that they needed findings on all elements of their cause of action in order to recover; they might reasonably have believed that findings on negbgence and causation were enough. Specifically, the Court says:
We have never considered how an undertaking claim should be submitted to the jury, nor had the courts of appeals directly addressed the issue at the time this ease was tried.
Because neither this Court nor any other appellate court had written about the proper submission of an undertaking claim at the time this case was tried, it is appropriate to remand this case in the interest of justice.4
The confusion in the law that justifies a new trial, the Court says, is not over what the elements of a voluntary-undertaking claim are, but over whether they must all be submitted to the jury.
This is a truly remarkable justification for ordering a new trial. I would have thought that the idea that a plaintiff must prove each element of his cause of action was so basic that it was not necessary for this Court or any court to reaffirm the requirement on a case-by-case basis.5 *854Moreover, we firmly rejected the argument in Keetch v. Kroger Co. that a finding of negligence and causation would support liability for a premises defect without findings on the other essential elements of the liability theory.6 In principle, Keetch is indistinguishable from the present case. The rationale used by the Court to decide both cases is exactly the same.7 The Court in this case cites Keetch as its principal authority. How then does the Court conclude that the plaintiffs could reasonably have believed after Keetch that they could establish liability for a voluntary undertaking without obtaining favorable findings on all elements of the claim? The answer is that the Court treats Keetch as merely an ad hoc determination of how to submit a premises liability claim to a jury that says nothing about how to submit any other claim with multiple elements. To justify a new trial in this case, the Court must rob Keetch of principle.
The weakness in the Court’s position is also demonstrated by the evidence it cites-one case and one article — in support of its assertion that the proper submission of a voluntary-undertaking claim is mired in confusion.8 The case, Park v. Larison,9 decided just four months ago, long after the trial in the present case, holds that the trial court did not err in refusing the plaintiffs’ request for a jury instruction on the elements of a voluntary-undertaking claim, apparently because no such claim was involved in the case. Nothing in the court’s opinion suggests that a plaintiff can establish voluntary-undertaking liability on findings of negligence and causation alone. The article the Court cites, Judge and Jury in the Texas Supreme Court,10 was authored by Dean Powers, Tarrington’s counsel in this case. According to the Court, the article notes “a wide split in authority about whether fact questions that form the predicate for a duty should be decided by the court or the jury.”11 That is true, but only in the context of determining whether a duty exists in particular circumstances. The article never suggests that if a duty exists, essential elements of liability can be decided by the court rather than a jury. It is impossible to imagine that an attorney could reasonably conclude from either the Park case or Dean Powers’ article that a voluntary-undertaking claim could properly be submitted to the jury without mention of all of its essential elements.
There is nothing in the record, or even in the arguments on appeal, to indicate that the plaintiffs thought about Colonial *855Savings, section 323 of the Restatement, Keetch, and Dean Powers’ article, and decided that the only findings needed to support a voluntary-undertaking claim against Torrington were negligence and causation. The Court orders a new trial not because of any confusion in the law, but because it wants the plaintiffs to have one. The result is in the interest of fíat, not in the interest of justice.
Second: the error in submitting the voluntary-undertaking claim to the jury was harmless and cannot justify a new trial unless there was some evidence of each element of the claim, and for three elements there was none.12
The Court concludes that there is some evidence to support the plaintiffs’ contention “that Torrington assumed a broader undertaking to investigate and identify all potentially defective bearings”.13 The Court states that this contention is “supported to some extent by the fact that Torrington ultimately gave Textron a list of serial numbers for both — 03 and — 05 bearings, some of which were manufactured at the New Britain plant.”14 The Court cites no other evidence that could possibly support the plaintiffs’ contention, and that is because none exists. Torring-ton supplied the list of serial numbers to Bell in response to Bell’s request after an investigation into the Rood crash involving a helicopter with a — 03 bearing. Torring-ton argues that any duty it undertook to identify faulty bearings was limited to the — 03 bearing. Inferring from Torring-ton’s response to an inquiry from Bell that Torrington had voluntarily undertaken to identify all its defective bearings is a stretch. But while the response listed — 03 and — 05 bearings, it listed no bearings manufactured by Torrington’s predecessor, and Bell’s parent, Textron. Nothing about the list of serial numbers Torrington gave Bell remotely suggests that Torrington undertook to identify defective bearings that it did not manufacture. All of the evidence shows that Torrington had no reason to undertake an investigation of Tex-tron bearings. Torrington did not own the plant when those bearings were manufactured. The investigation had not indicated that such bearings were defective or contaminated. The shelf life of the bearings was three to five years, and Textron had ceased manufacturing bearings six years before the investigation, so there was no reason to suppose that Textron bearings were still in use. Moreover, the evidence suggests no reason why Torrington would voluntarily undertake to determine for Bell whether Bell’s parent had manufactured defective bearings. If Bell had wanted that information — and there was no reason why it should, given the shelf life of the bearings — it would have asked its parent, Textron, who manufactured the bearings. Not only is there no evidence that Torring-ton ever undertook to investigate Tex-tron’s bearings, all of the evidence indicates that it would never have done so.
To submit a voluntary-undertaking theory to the jury, there must also have been evidence either that Torrington’s negligent performance of its undertaking increased the risk of harm, or that the Navy relied on Torrington’s undertaking. The Court does not even discuss this element of the theory. Assuming that Torrington undertook to inform the Navy whether Textron’s bearings were defective — even though as I have said, there is absolutely no evidence of such an undertaking — the issue becomes whether Torrington’s undertaking made *856the Navy worse off than if Torrington had done nothing. In other words, if Torring-ton had refused to cooperate with the investigation, would the Navy have taken other action that would have reduced the risk of harm from any defective bearings? The plaintiffs have cited no such evidence. When asked at oral argument whether any such evidence existed, counsel answered that it was simply “common sense” that the Navy would have done more. When the Navy did not follow its own safety procedures to reduce the risk of harm from bearings that were identified as possibly defective, and did not follow its shelf life policy to ensure that bearings were not employed after they were no longer dependable, there is no reason to think that it would have undertaken its own investigation to try to determine whether bearings manufactured six years earlier were defective.
Finally, there must have been evidence that Torrington’s failure to use reasonable care in undertaking to determine whether six-year-old Textron bearings were defective caused the helicopter crash, and there was none. The Navy assigned the bearings a three-year shelf life and required that they be removed at the end of that time, as the plaintiffs’ witnesses uniformly conceded at trial. Had the Navy followed its own shelf life policy, the bearing would never have been installed in the helicopter.
The Court’s assertion that evidence on the shelf life of the bearings was conflicting is simply wrong. The Court reads the Navy bearing manual to “suggest[ ] that the nondeteriorative rating of the hangar assembly could be applied to a bearing installed in the assembly.”15 This is not a reasonable reading of the manual. It makes no sense to say that a bearing should not be in service more than three years after it was manufactured, but if it is installed in an assembly it will last forever. Service deteriorates a bearing. Service certainly does not make a bearing nondet-eriorative, and not surprisingly, the Navy manual does not suggest otherwise. All the manual says is that if a bearing is properly installed and maintained, it — “the item” — should last for as long as its shelf life — here, three years. The Court also states that one department of the Navy assigned the — 05 bearing a “0” shelf life, indicating that the bearing was “nondeteri-orative”.16 This conclusion was refuted at trial by plaintiffs themselves. Regarding the Navy materials to which the Court refers, plaintiffs offered the following testimony at trial:
Q What does this say right here, beginning of paragraph 100?
A “Per Ref D Naval Supply Systems Command NavSup is designated the Navy’s managing agent for the shelf-life program. As such, NavSup is responsible for the control of shelf life items to include assigning shelf-life codes.”
Q Read the next sentence.
A “NavSup assigns shelf-life codes based upon advice or decision of military service technical/engineering support activities furnished to the integrated material manager. The shelf-life code assigned by NavSup to the tail rotor drive shaft hanger bearing is zero” — or it could be “no” — “nondeteriorative.”
Q “Non”- — I’m not going to be able to pronounce it either well-“nondeterio-rative.” So, the Navy has reviewed this and has done its own analysis that this bearing — and we know that they have their own shelf life manager; and we know now, based on what you told us, *857that they have a three-year shelf-life policy for this bearing.
A Unless otherwise modified by the depot bearing engineer.
Q Do you know if that has ever been modified by the depot bearing engineer prior — prior to this accident?
A That, I do not know.
The Court adds, as this testimony indicates, that bearing shelf lives could be modified “by the local cognizant depot bearing engineer,”17 but there is no evidence that any modification was ever made affecting the bearing in question.
Not one witness at trial, including the plaintiffs’ witnesses, was confused about the Navy shelf life policy for — 05 bearings. Even if there had been confusion about the Navy policy, responsibility for that confusion would have belonged to the Navy, not Torrington. Torrington never suggested to the Navy that the bearings had a shelf life longer than five years.
The Court simply ignores the inescapable fact that if the Navy had followed its own policy in 1987, nothing Torrington did or did not do in 1990 could have caused any harm. In General Motors Corp. v. Saenz, we held that if an accident would not have occurred had the manufacturer’s warning been heeded, the manufacturer’s failure to give a different warning could not be said to have caused the accident.18 Similarly here, because the Navy’s adherence to its own shelf-life policy would have prevented the accident, nothing Torrington failed to do can be said to have caused the helicopter crash.
The same argument applies to Textron. Assuming that its bearing was defectively designed, manufactured, and marketed, the defects could not have caused the harm in this case if the Navy had followed its own shelf-life policy. For this reason, there was no evidence to support the plaintiffs’ claim against Textron, and no reason now to order a new trial.
Accordingly, I agree with the Court that the judgment for the plaintiffs must be reversed, but I would not remand the case for a second trial.
. See, e.g., Spencer v. Eagle Star Ins. Co. of Amer., 876 S.W.2d 154, 157 (Tex.1994); Boyles v. Kerr, 855 S.W.2d 593, 603 (Tex.1993); Westgate, Ltd. v. State, 843 S.W.2d 448, 455 (Tex.1992).
. 544 S.W.2d 116, 119-120 (Tex.1976).
. Ante at 838 n. 7.
. Ante at 838 (citations omitted).
. Keetch v. Kroger Co., 845 S.W.2d 262, 267 (Tex.1992) ("It should hardly need be said that broad form submission does not entail omitting elements of proof from the charge. Broad form involves inclusion of multiple elements within a single question, usually by adding accompanying instructions, when it is *854feasible to do so. It alters only the form of the charge, not the substance.”) (Hecht, J., concurring).
. Id. at 267.
. The Court says: "Here, the broad-form negligence question allowed the juiy to hold Torrington liable regardless of whether Tor-rington knew or should have known that its services were necessary to protect others. Cf. Spencer v. Eagle Star Ins. Co. of Am., 876 S.W.2d 154, 157 (Tex.1994). The question also allowed an affirmative answer regardless of whether anyone relied upon Torrington’s undertaking, or whether Torrington’s performance of its undertaking increased the plaintiff's risk of harm. Because the question allowed the jury to find Torrington liable even if the plaintiffs did not establish the necessary factual predicates for a negligent undertaking duty, it was erroneous. These essential elements of an undertaking claim should be included in the instructions accompanying a broad-form negligence question. Cf. Keetch v. Kroger Co., 845 S.W.2d 262, 267 (Tex. 1992).” Ante at 840.
. Ante at 841 n. 10.
. 28 S.W.3d 106 (Tex.App.—Texarkana 2000, no pet. hist.).
. 75 TexL.Rev. 1699, 1716 (1997).
. Ante at 841.
. See Wilson v. Texas Parks & Wildlife Dept., 8 S.W.3d 634, 635 (Tex.1999).
. Ante at 839.
. Ante at 839.
. Ante at 846.
. Ante at 846.
. Ante at 846.
. 873 S.W.2d 353, 361 (Tex.1993).