Robert M. Hopkins, Jr., was badly hurt when his 1970 Chevrolet pick-up truck failed to keep to the road on a sharp turn and capsized. He contends in this products liability suit that the accident occurred because the driver lost control of the speed of the truck when the secondary butterfly valves of a defective carburetor were locked in an open position. The jury found that the carburetor was defectively designed and that this defect was a producing cause of the accident. Among its defenses General Motors Corporation contended that any malfunction of the carburetor was due to changes in the mechanism effected by Hopkins in the course of its removal and reins-tallation upon the motor of the truck. The jury found that there had been “misuse” of the carburetor by Hopkins and that this was a producing cause of the accident. The trial court disregarded the findings relative to misuse and rendered judgment for Hopkins. The Court of Civil Appeals affirmed, deciding that there was evidence to support all of the jury’s findings, but holding that misuse which is only a concurring cause of plaintiff’s damages is no defense for the manufacturer of a defective product. 535 S.W.2d 880, (Tex.Civ.App.). We affirm the judgment for Hopkins.
DEFECTIVE DESIGN
The quadrajet carburetor installed on the new Chevrolet truck when Hopkins bought *347it in 1970 has two primary valves and barrels through which air and gasoline are fed into the intake manifold and the cylinders of the engine according to direct control from the accelerator pedal. Maximum acceleration and fuel flow is obtained by opening the two secondary valves and barrels, which is accomplished when the driver presses the pedal some two-thirds of the distance down toward the floorboard. If the secondaries are opened while the engine is being started cold, however, the mixture and flow of fuel and air may not be suitable. A lock-out system is designed to prevent this. By that design until the engine is warm a lock-out lever holds a lock-out pin down and thereby prevents the secondary throttle shaft from turning. The evidence offered by Hopkins, which evidence supports the jury findings of design defect and of that defect as a cause of the accident, tended to prove that the lock-out pin hung on top of the lock-out lever — jamming the secondary valves one-third open and feeding a large quantity of fuel into the engine beyond any means of control by the driver (unless he turned off the ignition). GM has argued that the lock-out pin could not possibly hang on top of the lock-out lever when the engine was warm and in operation. Assuming the truth of the evidence in support of the jury findings, as we must do, there is an abundance of proof here that this could and did happen. The lock-out lever can be moved by the bouncing of the vehicle on a rough road. If the accelerator is pumped down to rotate the lock-out pin upward and if the pin comes down when the lever is rotated underneath, the pin can hang — especially if there is a gritty substance on the adjoining surfaces. The circumstances of the road, the operation of the truck, and the action of its engine at the time of the accident all fit these conditions according to the testimony of Hopkins and the driver of the truck. The pin was found in a hung position over the lever upon examination of the carburetor following the accident. Furthermore, a similar malfunction had occurred, according to two witnesses, a few weeks before the accident. Although Hopkins and a mechanic saw the misposition of the lock-out pin and lever, they did not then realize its significance. This prior experience with the runaway carburetor occurred before the time when Hopkins operated his truck with a different carburetor; there is no evidence of his alteration or “misuse” of the GM carburetor prior to that occasion.
The evidence also tends to prove that GM knew of this problem as early as February of 1968. A picture taken in 1968 of this model carburetor with the lock-out pin positioned on top of the lock-out lever was obtained from GM files and introduced into evidence. Correspondence among GM engineers was also introduced which reflects the effort to make a change in the design to prevent this misposition and which states that these engineers considered the change to be an urgent safety matter. Plaintiff’s experts concluded that the hung lock-out pin caused the accident. We agree with the Court of Civil Appeals that the evidence supports the jury findings of design defect 1 and of this defect being a producing cause of the accident.
*348ALTERATIONS BY HOPKINS
Hopkins bought the Chevrolet truck new on August 24, 1970. The first malfunction of the carburetor mentioned above occurred in the Spring of 1971. A few weeks after that occurrence Hopkins removed the GM carburetor and put a Holley carburetor on his truck. He kept the Holley on the truck for about a week and then removed it and reinstalled the original GM quadrajet carburetor. The accident occurred in June of 1971. GM contends that if the carburetor did malfunction, it was because of Hopkins’ alterations in the process of his removal and reinstallation of the GM carburetor. Eleven changes are named in the Court of Civil Appeals opinion, 535 S.W.2d 880, and include the use of improper bolts and hoses and clamps, etc. The jury found that the manner in which Hopkins reinstalled the carburetor constituted a misuse 2 of the vehicle and that such misuse was a producing cause of the accident. These findings were disregarded by the trial court because, as stated in its judgment, they were “without any support in the evidence” and were “immaterial to the establishment of any defense.”
The Court of Civil Appeals found evidence to support the causal connection of only one change made by Hopkins in and around the carburetor. That Court pointed to the “testimony that when the choke rod was disconnected the choke could be ‘blown open’ under such circumstances that the lock-out pin could be positioned on top of the lock-out lever with the secondaries open” and concluded that the jury was entitled to find “that the disconnected choke rod was a contributing factor in the malfunction of the carburetor system.” 535 S.W.2d 887.
GM attempted to explain the accident in several ways. First, it said that the truck was simply driven too fast to negotiate the sharp turn. Neighbors in the area of the mishap testified to the noise made by the truck on the road before it reached the curve. Then GM offered some evidence which suggested that the Holley carburetor was still installed on the truck at the time of the accident. Finally and at great length, GM experts testified about the changes made by Hopkins in the reinstallation of the carburetor and how these changes might have been a factor in causing the accident. Almost all of this testimony tells of the possibilities of what could have happened. There is not the slightest suggestion that these changes could have played any role in the loss of control of the vehicle and its overturning apart from the uncontrolled racing of the engine described by Hopkins and the driver. The only explanation to be found in the evidence of this racing of the engine is the malfunction of the carburetor. No explanation is given which would permit common experience or knowledge apart from expert testimony to determine the reason for the malfunction of the carburetor. The jury and court must therefore depend upon the explanations and opinions of the experts. Under these circumstances testimony of a possibility without any opinion of the reasonable probability would not support a finding of causal connection. Webb v. Western Casualty & Surety Co., 517 S.W.2d 529 (Tex.1974).
All of the testimony about the possibilities (the blowing open of the choke because the rod was not connected or the binding of the valves because of protruding gasket or misalignment due to uneven pressure of the bolts after reinstallation, etc.) was admissible on the primary issue of the causal connection between an original defect of the product and this accident. For *349that matter, the evidence was admissible on the issue of causal connection between Hopkins’ alterations and the accident. Even if it may not be construed as an opinion that the two were probably related, the testimony is surely relevant. See Otis Elevator Co. v. Wood, 436 S.W.2d 324, 331 (Tex.1968). The testimony in this record about possible causes, while relevant, is no more than speculation and conjecture as to what occurred at the time of the accident and cannot alone support a finding that there was a causal connection between the “misuse” and the accident. Insurance Company of North America v. Myers, 411 S.W.2d 710 (Tex.1966). If the manufacturer or supplier of a dangerously defective product is to relieve himself of all or part of the liability for damages caused by the defect, then he must prove the cause-in-fact connection of any misuse or alteration by the same standard as the user faces in connecting the product defect and his damages.
At one point in the presentation of GM’s defensive evidence, the witness went further. The witness Arthur Van Stetlandt stated without qualification his opinion that if the lock-out pin did hang above the lockout lever, it was because the driver made an attempt to feed the engine and the misposition of the thermocoil cover (backwards and upwards) interfered with the operation of the lever. This is legally sufficient evidence of the cause-in-fact connection between the alteration and the accident. We have here then a case where an original product defect and the misuse of the product by the plaintiff-consumer concur to cause the damaging event.
“MISUSE” AS A DEFENSE WHEN A CONCURRING CAUSE
It is often said that misuse is a defense in products liability suits. Generally speaking this is true, because there are a variety of points upon which the unintended or reasonably unforeseeable use or alteration of a product may be relevant to the liability of the supplier of the product. The misuse may bear upon the issue of whether the product was defective when it left the hands of the supplier or the misuse may bear on the issue of what caused the harm.
We cannot charge the manufacturer of a knife when it is used as a toothpick and the user complains because the sharp edge cuts. A harness hook is not necessarily defective simply because it breaks while being used to hold up a 1700 pound weight. Dosier v. Wilcox & Crittendon Co., 45 Cal.App.3d 74, 119 Cal.Rptr. 135 (1975). There are a number of cases where the manufacturer installed or supplied a safety guard or shield but injury occurs after the purchaser of the machine removes or easts aside the safety device. The foreseeability of that deviation in the manufacturer’s intended use of the product is relevant to the basic question of whether the product was unreasonably dangerous when and as it was marketed. McGrath v. Wallace Murray Corp., 496 F.2d 299 (10th Cir.1974). It is because of cases in these areas that the draftsmen of the Restatement wrote the following in Comment (h) following § 402A:
(h) A product is not in a defective condition when it is safe for normal handling and consumption. If the injury results from abnormal handling, as where a bottled beverage is knocked against a radiator to remove the cap, or from abnormal preparation for use, as where too much salt is added to food, or from abnormal consumption, as where a child eats too much candy and is made ill, the seller is not liable. Restatement of Torts 2d § 402A(h) (1965).
Misuse in this application has been defined by the Oregon Supreme Court as “a use or handling so unusual that the average consumer could not reasonably expect the product to be designed and manufactured to withstand it—a use which the seller, therefore, need not anticipate and provide for.” Findlay v. Copeland Lumber Co., 265 Or. 300, 509 P.2d 28, 31 (1973).
There is a related question of the proof of the condition of a product as of the time when the prodúct left the hands of the defendant supplier. If the plaintiff has no evidence of a specific defect in the design or manufacture of the product, he may offer *350evidence of its malfunction as circumstantial proof of the product’s defect. See Williams v. Ford Motor Co., 411 S.W.2d 443 (Mo.App.1966); Greco v. Bucciconi Engineering Co., 407 F.2d 87 (3rd Cir.1969). The age and use of that product during the time intervening between the purchase and malfunction will tend to support or defeat the circumstantial weight of the malfunction as proof of original defect. In a Pennsylvania case a load of steel pipe fell on the plaintiff when the brake locking mechanism of a crane failed. The record contained no proof of the crane’s defect except the malfunction itself, and the plaintiff contended that it was not necessary for him to present evidence of a specific defect to explain why the brake failed. His difficulty in this contention was that the machine had been in use for 20 years, that it was originally a shovel and had been converted into a drag-line and then into a crane, and that many changes and adjustments had been made in the brake bands and pedals and lock. The Court there speaks of identifying the cause of the failure, but it could just as well speak of the question of proof of original defect of the machine. Under the record both questions could not be answered except by speculation. The case was remanded, however, because the trial court erroneously excluded expert testimony as to a design defect of the brake mechanism. Kuisis v. Baldwin-Lima-Hamilton Corp., 457 Pa. 321, 319 A.2d 914 (1974).
In many cases the question is solely one of causation. Whether an original defect is proved or whether that too depends upon the effect of subsequent use as in the Pennsylvania crane case above, causation must be established. Where plaintiff had added a fifth wheel to his tractor, the dis-positive question was whether that change had caused the accident. Dennis v. Ford Motor Co., 471 F.2d 733 (3rd Cir.1973). Where a pipe-bending shoe broke after structural changes in the shoe and while being used on a press different from the type for which it was designed, the trier of fact was required to resolve the causation problem. Mazzi v. Greenlee Tool Co., 320 F.2d 821 (2nd Cir.1963). Where a gas heater which exploded had been converted for propane gas, the question was the cause of the defect and explosion. Hales v. Green Colonial, Inc., 490 F.2d 1015 (8th Cir.1974). When a crane collapsed while lifting without the use of outriggers, the plaintiff admitted that he knew the manufacturer’s directions were to use the outriggers in lifting the heavy load, but he insisted that the outriggers were only for stability and would play no part in the manner in which the crane collapsed. The trial court had directed a verdict for the manufacturer for the reason that the crane was not used in the way it was intended to be used. The case was remanded for trial on the causation question. Kudelka V. American Hoist & Derrick Co., 541 F.2d 651 (7th Cir.1976).
The foregoing are illustrative of most of the cases where unforeseeable use or misuse are relevant to the liability of product suppliers. This is not to suggest that all statements of judges and writers can be so easily organized and summarized. To some courts misuse and contributory negligence are very similar. One opinion seems to hold that the use of strong drink by the driver of an automobile was a complete defense to a products liability suit — that being misuse though contributory negligence would be no defense. Kirkland v. General Motors Corp., 521 P.2d 1353 (Okl.1974); cf. Fields v. Volkswagen of America, Inc., 555 P.2d 48 (Okl.1976). Again, whether foreseeable or not, cars are not intended to be used in collision with other objects; it has been held that this exonerates the manufacturer despite damages due to a danger of his product which played no part in causing the collision. Evans v. General Motors Corp., 359 F.2d 822 (7th Cir.1966). A larger list and discussion of the cases will be found in the following: 1 Frumer and Friedman, Products Liability § 15 (1971); Noel, Defective Products: Abnormal Use, Contributory Negligence and Assumption of Risk, 25 Vanderbilt L.Rev. 93 (1972); Anno. Products Liability: Alteration of Product After It Leaves Hands of Manufacturer or Seller as Affecting Liability for Product—Caused Harm, 41 A.L.R.3d 1251 (1972); Anno. *351Products Liability: Strict Liability in Tort, 13 A.L.R.3d 1057 (1967).
This brings us to our case: where an unreasonably dangerous defect of the product and its unforeseeable misuse are concurring causes of the damaging event. Does the injured user recover all or none or "a portion of his damages? We do not find the answer in precedents in Texas or elsewhere. Nor does the Restatement give us any guidance; section 402A comment h, quoted above, applies only where the product “is safe for normal handling and consumption.” That is not the case where the facts are that the product was defective — because of an unreasonably dangerous design, for example — and the defect is a producing cause of the injury.
We reject misuse as a defense where the product is dangerous for its foreseeable use and that danger is a producing cause of the injury of a bystander or a user who has not himself made some unforeseeable use of the product. On the other hand, product liability was never intended to take the place of insurance, and we see no justification for making the supplier reimburse the plaintiff for that portion of his own damages as he caused by a use of the product which the supplier would not have foreseen — and which use the plaintiff should have foreseen would create or increase the attendant danger. Reduction of the plaintiff’s recovery should be ordered where the misuse is a concurring proximate cause of the damaging event.
The defect of the supplier’s product need be only the producing cause3 of the harm for the supplier to be held liable. C. A. Hoover & Son v. O. M. Franklin Serum Co., 444 S.W.2d 596 (Tex.1969). His liability is not rested upon what he knew or should have known when he manufactured or sold the product; it rests on his placing into the stream of commerce a product which is demonstrated at trial to have been dangerous. The damaging event may not have been reasonably foreseeable at the time of manufacture or sale because the dangerous factor of the product might not then have been even reasonably knowable. The supplier would thus be free of culpability, but a price of his doing business is to protect people from danger from his products — or to pay. That is a heavy burden on the supplier, but the exposure is not infinite. The supplier is not liable so long as the product is safe for its foreseeable uses. He is not ordinarily required to bear the loss of a user who knows and appreciates the danger, has the opportunity to avoid the harm, but proceeds to use the product and thereby assumes the risk. Henderson v. Ford Motor Co., 519 S.W.2d 87 (Tex.1974). We decide now that the supplier should not be required to pay for all of the damages suffered by a user who contributed to the cause of his harm by unforeseeable handling of the product.
As indicated in the discussion above, contentions by the supplier relative to unforeseeable misuse of the product by the consumer usually bear upon the primary issues of defect and/or eause-in-fact of the damaging event. Where the supplier is unsuccessful in the contentions on those issues, he may interpose an affirmative defense and attempt to prove that the consumer plaintiff altered or misused the product in an unforeseen manner, that the misuse was a proximate cause of the damaging event, and that the misuse contributed to the cause of the event by a certain percent or fraction of the total contribution as between the product defect and the misuse.
It is essential that the supplier prove, as an element of this defense, that the consumer plaintiff should have reasonably anticipated as consequences of the misuse that the malfunction or injury, or some similar malfunction or injury, would occur. *352The harm must be reasonably foreseeable 4 to the user if he is to be penalized. Suppose, for example, a young automobile buff such as R. M. Hopkins, Jr. makes an alteration in his carburetor which, though unforeseeable to the manufacturer, could be reasonably expected by one with Hopkins’ knowledge to risk some difficulties in the operation of his vehicle: for example, slow starts or flooding or sluggish acceleration. But suppose that this alteration does far more: the alteration makes it more likely that a defect of design or manufacture, which constitutes a danger in the normal and expected operation of the carburetor, will be activated and cause the vehicle to speed out of control. If the malfunction and damaging event are not reasonably foreseeable to the user, his misuse should not limit his recovery.
To recapitulate, if the product is found to have been unreasonably dangerous when the defendant placed it in the stream of commerce, and if that defect is found to have been a producing cause of the damaging event, and if the plaintiff has misused the product in the sense as defined by the trial court in its charge in the present case, and if that misuse is a proximate cause of the damaging event, the trier of fact must then determine the respective percentages (totalling 100%) by which these two concurring causes contributed to bring about the event. This comparison and division of causes is not to be confused with the statutory scheme of modified comparative negligence which bars all recovery to the plaintiff if his negligence is greater than the negligence of the parties against whom recovery is sought. Art. 2212a, Vernon’s Ann.Civ.St. (P.P. 1976-1977). The defense in a products liability case, where both defect and misuse contribute to cause the damaging event, will limit the plaintiff’s recovery to that portion of his damages equal to the percentage of the cause contributed by the product defect.
In the present case the defendant GM did not plead misuse of the product as a proximate cause, nor did GM object to the inquiry in special issue No. 4 of “producing cause,” nor has GM made any complaint in the course of the appeal on this ground. Further, GM sought no findings of the percent of the cause of the accident to be attributed to the misuse. Even if we were to say that from the defendant’s standpoint the case was tried on the wrong theory (or that GM might have pursued a different theory or contention with the present writing in hand), we cannot order a reversal and remand for a new trial unless we also find error in the trial court’s judgment. Halepeska v. Callihan Interests, Inc., 371 S.W.2d 368 (Tex.1963).
Finding no error in the judgments of the District Court and Court of Civil Appeals, they are affirmed.
YARBROUGH, J., not sitting.