*556Opinion
BAXTER, J.
—Plaintiff’s ankles were badly injured when her General Motors (GM) car collided with another vehicle. She sued GM, asserting that defects in her automobile allowed its left front wheel to break free, collapse rearward, and smash the floorboard into her feet. GM denied any defect and claimed that the force of the collision itself was the sole cause of the injuries. Expert witnesses debated the issues at length. Plaintiff prevailed at trial, and the Court of Appeal affirmed the judgment.
We granted review to resolve three questions. First, may a product’s design be found defective on grounds that the product’s performance fell below the safety expectations of the ordinary consumer (see Barker v. Lull Engineering Co. (1978) 20 Cal.3d 413, 426-432 [143 Cal.Rptr. 225, 573 P.2d 443, 96 A.L.R.3d 1]) if the question of how safely the product should have performed cannot be answered by the common experience of its users? Second, in an action for enhanced collision injuries caused by an uncrashworthy vehicle, where a correct general instruction on legal cause is given, is it error to refuse a defense instruction that any defect cannot be a legal cause of injury if the accident would have produced the same injury even without the defect? Third, if the refusal is error, is it reversible per se? (See, e.g., Self v. General Motors Corp. (1974) 42 Cal.App.3d 1, 10-11 [116 Cal.Rptr. 575].)
We reach the following conclusions: The trial court erred by giving an “ordinary consumer expectations” instruction in this complex case. Moreover, the court should have granted GM’s request for a special instruction explaining its correct theory of legal cause. However, neither error warrants reversal unless it caused actual prejudice, and both errors were harmless on this record. We will therefore affirm the Court of Appeal’s judgment.
Facts
On the early afternoon of January 16,1984, plaintiff was driving her 1982 Camaro in the southbound center lane of Bolsa Chica Road, an arterial street in Westminster. There was a slight drizzle, the roadway was damp, and apparently plaintiff was not wearing her seat belt. A 1972 Datsun, approaching northbound, suddenly skidded into the path of plaintiff’s car. The Datsun’s left rear quarter struck plaintiff’s Camaro in an area near the left *557front wheel. Estimates of the vehicles’ combined closing speeds on impact vary from 30 to 70 miles per hour.1
The collision bent the Camaro’s frame adjacent to the wheel and tore loose the bracket that attached the wheel assembly (specifically, the lower control arm) to the frame. As a result, the wheel collapsed rearward and inward. The wheel hit the underside of the “toe pan”—the slanted floorboard area beneath the pedals—causing the toe pan to crumple, or “deform,” upward into the passenger compartment.
Plaintiff received a fractured rib and relatively minor scalp and knee injuries. Her most severe injuries were fractures of both ankles, and the more serious of these was the compound compression fracture of her left ankle. This injury never healed properly. In order to relieve plaintiff’s pain, an orthopedic surgeon fused the joint. As a permanent result, plaintiff cannot flex her left ankle. She walks with considerable difficulty, and her condition is expected to deteriorate.
After the accident, the Camaro was acquired by a salvage dealer, Noah Hipolito. Soon thereafter, plaintiff’s son, Jeffrey Bishop, and her original attorney, Richard Hawkins, each inspected and photographed the car and its damaged floorboard area. The failed bracket assembly was retrieved. However, Hipolito later discarded the damaged toe pan, repaired the Camaro, and resold it. Thus, except for the bracket assembly, no part of the vehicle was retained as evidence.
Plaintiff sued GM for her ankle injuries, asserting a theory of strict tort liability for a defective product. She claimed the severe trauma to her ankles was not a natural consequence of the accident, but occurred when the collapse of the Camaro’s wheel caused the toe pan to crush violently upward against her feet. Plaintiff attributed the wheel collapse to a manufacturing defect, the substandard quality of the weld attaching the lower control arm bracket to the frame. She also claimed that the placement of the bracket, and the configuration of the frame, were defective designs because they did not limit the wheel’s rearward travel in the event the bracket should fail.
The available physical and circumstantial evidence left room for debate about the exact angle and force of the impact and the extent to which the toe pan had actually deformed. The issues of defect and causation were addressed through numerous experts produced by both sides in such areas as *558biomechanics, metallurgy, orthopedics, design engineering, and crash-test simulation.
Plaintiff submitted the results of crash tests, and also asserted the similarity of another real-world collision involving a 1987 Camaro driven by Dana Carr. According to plaintiff’s experts, these examples indicated that Camaro accidents of similar direction and force do not generally produce wheel bracket assembly failure, extensive toe pan deformation, or severe ankle injuries such as those plaintiff had experienced. These experts opined that without the deformation of the toe pan in plaintiff’s car, her accident could, not have produced enough force to fracture her ankles.
A metallurgist testifying on plaintiff’s behalf examined the failed bracket from her car. He concluded that its weld was particularly weak because of excess “porosity” caused by improper welding techniques. Plaintiff’s experts also emphasized the alternative frame and bracket design used by the Ford Mustang of comparable model years. They asserted that the Mustang’s design, unlike the Camaro’s, provided protection against unlimited rearward travel of the wheel should a bracket assembly give way.
GM’s metallurgist disputed the claims of excessive weakness or porosity in the bracket weld. Expert witnesses for GM also countered the assertions of defective design. GM asserted that the Camaro’s bracket was overdesigned to withstand forces in excess of all expected uses. According to expert testimony adduced by GM, the Mustang’s alternative frame and bracket configuration did not fit the Camaro’s overall design goals and was not distinctly safer for all collision stresses to which the vehicle might be subjected. Indeed, one witness noted, at least one more recent Ford product had adopted the Camaro’s design.
A second major thrust of GM’s defense was that the force of the collision, rather than any product defect, was the sole cause of plaintiff’s ankle injuries. Using the results of accident reconstruction, computer simulations, and actual crash tests, GM sought to prove that the probable collision force concentrated on the left front wheel of plaintiff’s Camaro exceeded the “yield strength” of any feasible weld or design.
By similar means, GM also sought to show that plaintiff’s ankle injuries were not caused by the upward movement of the toe pan, but by the inertial forward and downward motion of plaintiff’s unrestrained body and legs against the toe pan at the instant of impact. From plaintiff’s other injuries, and from photographs showing the general pattern of damage to the Camaro’s interior, GM’s experts inferred that plaintiff was not wearing her seat belt and had locked or braced her legs in reaction to the imminent collision.
*559Hence, they concluded, her rigid ankles had absorbed the Ml force of her inertial forward movement, which was sufficient to cause the fractures. Based on their test results, GM’s witnesses opined that plaintiff’s ankles had probably moved forward, struck the toe pan, and broken before significant deformation of the toe pan occurred.
The court instructed the jury that a manufacturer is liable for “enhanced” injuries caused by a manufacturing or design defect in its product while the product is being used in a foreseeable way. Over GM’s objection, the court gave the standard design defect instruction without modification. (See BAJI No. 9.00.5 (7th ed. 1986).) This instruction advised that a product is defective in design “if it fails to perform as safely as an ordinary consumer would expect when used in an intended or reasonably foreseeable manner or if there is a risk of danger inherent in the design which outweighs the benefit of the design.” (Italics added.)
The jury was also told that in order to establish liability for a design defect under the “ordinary consumer expectations” standard, plaintiff must show (1) the manufacturer’s product failed to perform as safely as an ordinary consumer would expect, (2) the defect existed when the product left the manufacturer’s possession, (3) the defect was a “legal cause” of plaintiff’s “enhanced injury,” and (4) the product was used in a reasonably foreseeable manner.
With respect to all theories of liability, the instructions indicated that “[a] legal cause of injury is a cause which is a substantial factor in bringing about the injury.” (See BAJI No. 3.76 (7th ed. 1986).) However, the trial court refused the following instruction proffered by GM: “If you find that the subject Camaro . . . was improperly designed, but you also find that [plaintiff] would have received enhanced injuries even if the design had been proper, then you must find that the design was not a substantial factor in bringing about her injuries and therefore was not a contributing cause thereto.”
In a series of special findings, the jury determined that the Camaro contained a defect (of unspecified nature) which was a “legal cause” of plaintiff’s “enhanced injury.” The jury further concluded that although plaintiff was guilty of comparative fault, her conduct was not a legal cause of her enhanced injuries. Plaintiff received an award of $1.65 million.
GM appealed. Among other things, it argued that the trial court erred by instructing on ordinary consumer expectations in a complex design-defect case, and by failing to give GM’s special instruction on causation.
*560Following one line of authority, the Court of Appeal concluded that a jury may rely on expert assistance to determine what level of safe performance an ordinary consumer would expect under particular circumstances. Hence, the Court of Appeal ruled, there was no error in use of the ordinary consumer expectations standard for design defect in this case.
The Court of Appeal agreed with GM that its specific instruction on causation should not have been refused. However, the court rejected precedent suggesting that an error of this kind is reversible per se. Here, the Court of Appeal ruled, the error was harmless. After dismissing GM’s remaining appellate claims, the Court of Appeal affirmed the judgment. We granted review.
Discussion
1. Test for design defect.
A manufacturer, distributor, or retailer is liable in tort if a defect in the manufacture or design of its product causes injury while the product is being used in a reasonably foreseeable way. (Cronin v. J.B.E. Olson Corp. (1972) 8 Cal.3d 121, 126-130 [104 Cal.Rptr. 433, 501 P.2d 1153] (Cronin); Greenman v. Yuba Power Products, Inc. (1963) 59 Cal.2d 57, 62 [27 Cal.Rptr. 697, 377 P.2d 897, 13 A.L.R.3d 1049] (Greenman).) Because traffic accidents are foreseeable, vehicle manufacturers must consider collision safety when they design and build their products. Thus, whatever the cause of an accident, a vehicle’s producer is liable for specific collision injuries that would not have occurred but for a manufacturing or design defect in the vehicle. (Cronin, supra, at p. 126.)
In Cronin, supra, a bread van driver was hurt when the hasp retaining the bread trays broke during a collision, causing the trays to shift forward and propel him through the windshield. He sued the van’s producer, alleging that the hasp had failed because of the defective metal used in its manufacture. The court instructed that the driver could recover if he proved a defect, unknown to him, which caused injury while the van was being used as intended or designed. The manufacturer appealed the subsequent damage award. It urged the court should have instructed that liability could not be imposed unless the defect rendered the product “unreasonably dangerous.”
We rejected this contention, holding that the “unreasonably dangerous” test derived from the Restatement (see Rest.2d Torts, § 402A) is inapplicable in California. As we observed, the Restatement defines “unreasonably dangerous” as “dangerous to an extent beyond that which would be contemplated by the ordinary consumer who purchases it, with the ordinary knowledge common to the community as to its characteristics.” (Id., com. i, p. 352, *561italics added.) The original purpose of this formula, we explained, was to make clear that common products such as sugar, butter, and liquor are not defective simply because they pose inherent health risks well known to the general public. However, Cronin indicated, the formula had been applied so as to force injured persons to prove both an actual defect and “unreasonable” danger. (8 Cal.3d at pp. 132-133.)
This “double burden,” Cronin reasoned, ran contrary to the purpose of Greenman, supra, to relieve persons injured by defective products from proof of elements that ring of negligence. Instead, Cronin concluded, an injured plaintiff should recover so long as he proves that the product was defective, and that the defect caused injury in reasonably foreseeable use. (Cronin, supra, 8 Cal.3d at pp. 133-134.)
In Barker v. Lull Engineering Co., supra, 20 Cal.3d 413 (Barker), the operator of a high-lift loader sued its manufacturer for injuries he received when the loader toppled during a lift on sloping ground. The operator alleged various design defects which made the loader unsafe to use on a slope. In a pre-Cronin trial, the court instructed that the operator could recover only if a defect in the loader’s design made the machine “ ‘unreasonably dangerous for its intended use.’ ” (Id., at p. 417.) The operator appealed the defense verdict, citing the “unreasonably dangerous” instruction as prejudicial error.
The manufacturer responded that even if the “unreasonably dangerous” test was inappropriate for manufacturing defects, such as the substandard fastener material in Cronin, it should be retained for design defects. This rule would not produce the undue double burden that concerned us in Cronin, the manufacturer insisted, because unreasonable danger is part of the definition of design defect, not an additional element of strict product liability. Without this limitation, the manufacturer contended, juries would lack guidance when determining if a defect had sprung not from a mistake in supply or assembly, but from a flaw in the product’s specifications.
The Barker court disagreed. It reasoned as follows: Our concerns in Cronin extended beyond double-burden problems. There we also sought to avoid the danger that a jury would deny recovery, as the Restatement had intended, “so long as the product did not fall below the ordinary consumer’s expectations as to [its] safety. . . .” (Barker, supra, 20 Cal.3d at p. 425, fn. omitted.) This danger was particularly acute in design defect cases, where a manufacturer might argue that because the item which caused injury was identical to others of the same product line, it must necessarily have satisfied ordinary consumer expectations. (Id., at p. 426.)
*562Despite these difficulties, Barker explained, it is possible to define a design defect, and the expectations of the ordinary consumer are relevant to that issue. At a minimum, said Barker, a product is defective in design if it does fail to perform as safely as an ordinary consumer would expect. This principle, Barker asserted, acknowledges the relationship between strict tort liability for a defective product and the common law doctrine of warranty, which holds that a product’s presence on the market includes an implied representation “ ‘that it [will] safely do the jobs for which it was built.’ ” (20 Cal.3d at p. 430, quoting Greenman, supra, 59 Cal.2d at p. 64.) “Under this [minimum] standard," Barker observed, “an injured plaintiff will frequently be able to demonstrate the defectiveness of the product by resort to circumstantial evidence, even when the accident itself precludes identification of the specific defect at fault. [Citations.]” (20 Cal.3d at p. 430, italics added.)
However, Barker asserted, the Restatement had erred in proposing that a violation of ordinary consumer expectations was necessary for recovery on this ground. “As Professor Wade has pointed out, ... the expectations of the ordinary consumer cannot be viewed as the exclusive yardstick for evaluating design defectiveness because ‘[i]n many situations ... the consumer would not know what to expect, because he would have no idea how safe the product could be made.’ ” (20 Cal.3d at p. 430, quoting Wade, On the Nature of Strict Tort Liability for Products (1973) 44 Miss. L.J. 825, 829, italics added.)
Thus, Barker concluded, “a product may be found defective in design, even if it satisfies ordinary consumer expectations, if through hindsight the jury determines that the product’s design embodies ‘excessive preventable danger,’ or, in other words, if the jury finds that the risk of danger inherent in the challenged design outweighs the benefits of such design. [Citations.]” (20 Cal.3d at p. 430, fn. omitted.) Barker held that under this latter standard, “a jury may consider, among other relevant factors, the gravity of the danger posed by the challenged design, the likelihood that such danger would occur, the mechanical feasibility of a safer alternative design, the financial cost of an improved design, and the adverse consequences to the product and to the consumer that would result from an alternative design. [Citations.]” (Id., at p. 431.)
Barker also made clear that when the ultimate issue of design defect calls for a careful assessment of feasibility, practicality, risk, and benefit, the case should not be resolved simply on the basis of ordinary consumer expectations. As Barker observed, “past design defect decisions demonstrate that, as a practical matter, in many instances it is simply impossible to eliminate the *563balancing or weighing of competing considerations in determining whether a product is defectively designed or not. . . .” (20 Cal.3d at p. 433.)
An example, Barker noted, was the “crashworthiness” issue presented in Self v. General Motors Corp., supra, 42 Cal.App.3d 1. The debate there was whether the explosion of a vehicle’s fuel tank in an accident was due to a defect in design. This, in turn, entailed concerns about whether placement of the tank in a position less vulnerable to rear end collisions, even if technically feasible, “would have created a greater risk of injury in other, more common situations.” (Barker, supra, 20 Cal.3d at p. 433.) Because this complex weighing of risks, benefits, and practical alternatives is “implicit” in so many design-defect determinations, Barker concluded, “an instruction which appears to preclude such a weighing process under all circumstances may mislead the jury.” (Id., at p. 434.)
Campbell v. General Motors Corp. (1982) 32 Cal.3d 112 [184 Cal.Rptr. 891, 649 P.2d 224] (Campbell) provided additional strong hints about the proper use of the ordinary consumer expectations prong of Barker. Plaintiff Campbell, a bus passenger, was thrown from her seat and injured during a sharp turn. She sued GM, the manufacturer of the bus, alleging that the vehicle was defectively designed because there was no “grab bar” within easy reach of her seat. Campbell presented no expert testimony, but she submitted photographs of the interior of the bus, showing where safety bars and handles were located in relation to the seat she had occupied. At the conclusion of her case in chief, GM moved for nonsuit, arguing that her evidence of design defect and proximate cause was not sufficient. The trial court granted the motion, but we reversed.
We emphasized that in order to establish a design defect under Barker's ordinary consumer expectations test, it was enough for Campbell to show “the objective conditions of the product” so that the jurors could employ “[their] own sense of whether the product meets ordinary expectations as to its safety under the circumstances presented by the evidence. [Fn. omitted.] Since public transportation is a matter of common experience, no expert testimony was required to enable the jury to reach a decision on this part of the Barker inquiry." (Campbell, supra, 32 Cal.3d at p. 126.)
“Indeed, it is difficult to conceive what testimony an ‘expert’ could provide. The thrust of the first Barker test is that the product must meet the safety expectations of the general public as represented by the ordinary consumer, not the industry or a government agency. ‘[0]ne can hardly imagine what credentials a witness must possess before he can be certified as an expert on the issue of ordinary consumer expectations.’ ” (Campbell, *564 supra, 32 Cal.3d at pp. 126-127, quoting Schwartz, Foreword: Understanding Products Liability (1979) 67 Cal.L.Rev. 435, 480, italics added.)
Had we ended our discussion at this point, it would have been clear that a product violates ordinary consumer expectations only when the circumstances arouse such reasonable expectations based on common experience of the product’s users. However, dictum in the next paragraph of Campbell injected ambiguity. We said, “The quantum of proof necessary to establish a prima facie case . . . under the first [i.e., ordinary consumer expectations] prong of Barker cannot be reduced to an easy formula. However, if the product is one within the common experience of ordinary consumers” (italics added), it will generally be enough for the injured plaintiff to show the circumstances of the accident and “the objective features of the product which are relevant to an evaluation of its safety. . . .” (32 Cal.3d at p. 127.) One might infer from this passage that the ordinary consumer expectations prong of Barker is not limited to product performance “within the common experience” of the product’s ordinary consumers.
Several subsequent Court of Appeal cases considered the point. In Bates v. John Deere Co. (1983) 148 Cal.App.3d 40 [195 Cal.Rptr. 637], plaintiff caught his leg in a commercial cotton picker while clearing debris from the moving machinery. He claimed the machine should have included an emergency shutoff switch within reach of the remote position from which its sole operator periodically had to undertake this debris-clearing task. Defense experts suggested that such a feature might induce a false sense of security and make the machine even more dangerous.
The trial court properly found a design defect under the risk-benefit test, but defendant challenged the court’s additional use of the ordinary consumer expectations test. Although it saw no need to decide the issue, the Court of Appeal agreed that “[w]e, too, find it difficult to apply the . . . [consumer expectations] test to these facts, in part because it is difficult to conceive that an ordinary consumer would know what to expect concerning the safety design of a commercial cotton picker. [Citing Barker.]” (148 Cal.App.3d at p. 52.)
In Lunghi v. Clark Equipment Co. (1984) 153 Cal.App.3d 485 [200 Cal.Rptr. 387], the Court of Appeal upheld the trial court’s refusal to instruct on reasonable consumer expectations because unassisted lay jurors “would not know what to expect” about the safety design of a Bobcat model 440 loader, and no experts had testified on the issue. However, the Court of Appeal remarked that on retrial, “appellants are free to present evidence in the form of expert opinions on the reasonable expectations of consumers of the product involved here. . . .” (Id., at p. 496.)
*565In Akers v. Kelly Co. (1985) 173 Cal.App.3d 633 [219 Cal.Rptr. 513] (Akers), there was an accident involving a “dockboard,” a spring-loaded plate which attaches to a loading dock and adjusts to form a bridge between the dock and truck beds of different elevations. Several hours after the prongs of a forklift struck the dockboard, it suddenly flew apart, injuring a nearby worker. Experts debated at length whether the dockboard’s components should have been designed to withstand forklift impacts, and whether a failure in design was a cause of the accident. Over defendant’s objection, the trial court instructed only on the consumer expectations test for design defect.
The Court of Appeal affirmed. It declined to read Campbell as limiting the consumer expectations test to products or accidents of common experience. (Akers, supra, 173 Cal.App.3d at p. 650.) That test, said Akers, “is entirely appropriate in a case such as this one. There are certain kinds of accidents— even where fairly complex machinery is involved—which are so bizarre that the average juror, upon hearing the particulars, might reasonably think: ‘Whatever the user may have expected from that contraption, it certainly wasn’t that.’ Here, a dockboard flew apart and injured [plaintiff]. A reasonable juror with no previous experience of dockboards could conclude that the dockboard in question failed to meet ‘consumer expectations’ as to its safety. . . .” (Id., at p. 651.) This was so, the Court of Appeal concluded, even though expert testimony might be necessary to establish that the manufacturer was responsible for the flaw which caused the product to fail. (Ibid.)
To similar effect is West v. Johnson & Johnson Products, Inc. (1985) 174 Cal.App.3d 831 [220 Cal.Rptr. 437, 59 A.L.R.4th 1] (West). The plaintiff in West became seriously ill in February 1980, during her menstrual period. At this time, there were increasing indications that tampon use sometimes causes toxic shock syndrome (TSS). After reading medical reports, plaintiff’s physicians belatedly concluded that she had suffered TSS caused by tampons which defendant had designed and produced. At trial, experts debated the nature of plaintiff’s illness, and they also disputed whether the tampon design and materials used by defendant encouraged TSS. The trial court instructed only on the consumer expectations prong of Barker.
On appeal, defendant argued that the risk-benefit test alone was proper. However, West agreed with Akers that Campbell does not preclude the consumer expectations test in complex cases involving expert testimony. In a time before general awareness and warnings about TSS, the court reasoned, plaintiff “had every right to expect” that use of this seemingly innocuous product “would not lead to a serious (or perhaps fatal) illness. . . .” Hence, the consumer expectations instruction was appropriate. (West, supra, 174 Cal.App.3d at p. 867.)
*566Finally, in Rosburg v. Minnesota Mining & Mfg. Co. (1986) 181 Cal..App.3d 726 [226 Cal.Rptr. 299], plaintiff claimed she was entitled to judgment under the consumer expectations test because her own testimony that she believed her breast implants would last a lifetime without leaking was the only lay evidence of what consumers expected. However, the Court of Appeal ruled that breast implant performance is beyond common experience, and that expert testimony on what the consumer should expect was therefore relevant and admissible. Here, the court observed, both plaintiff’s surgeon and another defense expert had insisted that failures were expectable and patients were not advised otherwise. Hence, there was substantial evidence to support the finding below that no defect was proven under the consumer expectations test. (Id., at pp. 7S2-733.)2
In Barker, we offered two alternative ways to prove a design defect, each appropriate to its own circumstances. The purposes, behaviors, and dangers of certain products are commonly understood by those who ordinarily use them. By the same token, the ordinary users or consumers of a product may have reasonable, widely accepted minimum expectations about the circumstances under which it should perform safely. Consumers govern their own conduct by these expectations, and products on the market should conform to them.
In some cases, therefore, “ordinary knowledge . . . as to . . . [the product’s] characteristics” (Rest.2d Torts, supra, § 402A, com. L, p. 352) may permit an inference that the product did not perform as safely as it should. If the facts permit such a conclusion, and if the failure resulted from the product’s design, a finding of defect is warranted without any further proof. The manufacturer may not defend a claim that a product’s design failed to perform as safely as its ordinary consumers would expect by presenting expert evidence of the design’s relative risks and benefits.3
However, as we noted in Barker, a complex product, even when it is being used as intended, may often cause injury in a way that does not engage its *567ordinary consumers’ reasonable minimum assumptions about safe performance. For example, the ordinary consumer of an automobile simply has “no idea” how it should perform in all foreseeable situations, or how safe it should be made against all foreseeable hazards. (Barker, supra, 20 Cal.3d at p. 430.)
An injured person is not foreclosed from proving a defect in the product’s design simply because he cannot show that the reasonable minimum safety expectations of its ordinary consumers were violated. Under Barker’s alternative test, a product is still defective if its design embodies “excessive preventable danger” (20 Cal.3d at p. 430), that is, unless “the benefits of the . . . design outweigh the risk of danger inherent in such design” (id., at p. 432). But this determination involves technical issues of feasibility, cost, practicality, risk, and benefit (id., at p. 431) which are “impossible” to avoid (id., at p. 433). In such cases, the jury must consider the manufacturer’s evidence of competing design considerations (id., at pp. 433-434), and the issue of design defect cannot fairly be resolved by standardless reference to the “expectations” of an “ordinary consumer.”
As we have seen, the consumer expectations test is reserved for cases in which the everyday experience of the product’s users permits a conclusion that the product’s design violated minimum safety assumptions, and is thus defective regardless of expert opinion about the merits of the design. It follows that where the minimum safety of a product is within the common knowledge of lay jurors, expert witnesses may not be used to demonstrate what an ordinary consumer would or should expect. Use of expert testimony for that purpose would invade the jury’s function (see Evid. Code, § 801, subd. (a)), and would invite circumvention of the rule that the risks and benefits of a challenged design must be carefully balanced whenever the issue of design defect goes beyond the common experience of the product’s users.4
*568By the same token, the jury may not be left free to find a violation of ordinary consumer expectations whenever it chooses. Unless the facts actually permit an inference that the product’s performance did not meet the minimum safety expectations of its ordinary users, the jury must engage in the balancing of risks and benefits required by the second prong of Barker.
Accordingly, as Barker indicated, instructions are misleading and incorrect if they allow a jury to avoid this risk-benefit analysis in a case where it is required. (20 Cal.3d at p. 434.). Instructions based on the ordinary consumer expectations prong of Barker are not appropriate where, as a matter of law, the evidence would not support a jury verdict on that theory. Whenever that is so, the jury must be instructed solely on the alternative risk-benefit theory of design defect announced in Barker. 5
GM suggests that the consumer expectations test is improper whenever “crashworthiness,” a complex product, or technical questions of causation are at issue. Because the variety of potential product injuries is infinite, the line cannot be drawn as clearly as GM proposes. But the fundamental distinction is not impossible to define. The crucial question in each individual case is whether the circumstances of the product’s failure permit an *569inference that the product’s design performed below the legitimate, commonly accepted minimum safety assumptions of its ordinary consumers.6
GM argues at length that the consumer expectations test is an “unworkable, amorphic, fleeting standard” which should be entirely abolished as a basis for design defect. In GM’s view, the test is deficient and unfair in several respects. First, it defies definition. Second, it focuses not on the objective condition of products, but on the subjective, unstable, and often unreasonable opinions of consumers. Third, it ignores the reality that ordinary consumers know little about how safe the complex products they use can or should be made. Fourth, it invites the jury to isolate the particular consumer, component, accident, and injury before it instead of considering whether the whole product fairly accommodates the competing expectations of all consumers in all situations (see Daly v. General Motors Corp., supra, 20 Cal.3d 725, 746-747). Fifth, it eliminates the careful balancing of risks and benefits which is essential to any design issue.
In its amicus curiae brief, the Product Liability Advisory Council, Inc. (Council) makes similar arguments. The Council proposes that all design defect claims be resolved under a single risk-benefit analysis geared to “reasonable safety.”
We fully understand the dangers of improper use of the consumer expectations test. However, we cannot accept GM’s insinuation that ordinary consumers lack any legitimate expectations about the minimum safety of the products they use. In particular circumstances, a product’s design may perform so unsafely that the defect is apparent to the common reason, experience, and understanding of its ordinary consumers. In such cases, a lay jury is competent to make that determination.
Nor are we persuaded by the Council’s proposal. In essence, it would reinvest product liability claims with the requirement of “unreasonable danger” that we rejected in Cronin and Barker.
When use of the consumer expectations test is limited as Barker intended, the principal concerns raised by GM and the Council are met. Within these limits, the test remains a workable means of determining the existence of *570design defect. We therefore find no compelling reason to overrule the consumer expectations prong of Barker at this late date, and we decline to do so.7
Applying our conclusions to the facts of this case, however, we agree that the instant jury should not have been instructed on ordinary consumer expectations. Plaintiff’s theory of design defect was one of technical and mechanical detail. It sought to examine the precise behavior of several obscure components of her car under the complex circumstances of a particular accident. The collision’s exact speed, angle, and point of impact were disputed. It seems settled, however, that plaintiff’s Camaro received a substantial oblique blow near the left front wheel, and that the adjacent frame members and bracket assembly absorbed considerable inertial force.
An ordinary consumer of automobiles cannot reasonably expect that a car’s frame, suspension, or interior will be designed to remain intact in any and all accidents. Nor would ordinary experience and understanding inform such a consumer how safely an automobile’s design should perform under the esoteric circumstances of the collision at issue here. Indeed, both parties assumed that quite complicated design considerations were at issue, and that expert testimony was necessary to illuminate these matters. Therefore, injection of ordinary consumer expectations into the design defect equation was improper.
We are equally persuaded, however, that the error was harmless, because it is not reasonably probable defendant would have obtained a more favorable result in its absence. (E.g., Pool v. City of Oakland (1986) 42 Cal.3d 1051, 1069 [232 Cal.Rptr. 528, 728 P.2d 1163]; see Cal. Const., art. VI, § 13; Code Civ. Proc., § 475.) In assessing prejudice from an erroneous instruction, we consider, insofar as relevant, “(1) the degree of conflict in the evidence on critical issues [citations]; (2) whether respondent’s argument to the jury may have contributed to the instruction’s misleading effect [citation]; (3) whether the jury requested a rereading of the erroneous instruction [citation] or of related evidence [citation]; (4) the *571closeness of the jury’s verdict [citation]; and (5) the effect of other instructions in remedying the error [citations].” (Pool, supra, 42 Cal.3d at pp. 1069-1070, quoting LeMons v. Regents of University of California (1978) 21 Cal.3d 869, 876 [148 Cal.Rptr. 355, 582 P.2d 946].)
Here there were no instructions which specifically remedied the erroneous placement of the consumer expectations alternative before the jury. Moreover, plaintiff’s counsel briefly reminded the jury that the instructions allowed it to find a design defect under either the consumer expectations or risk-benefit tests. However, the consumer expectations theory was never emphasized at any point. As previously noted, the case was tried on the assumption that the alleged design defect was a matter of technical debate. Virtually all the evidence and argument on design defect focused on expert evaluation of the strengths, shortcomings, risks, and benefits of the challenged design, as compared with a competitor’s approach.
Neither plaintiff’s counsel nor any expert witness on her behalf told the jury that the Camaro’s design violated the safety expectations of the ordinary consumer. Nor did they suggest the jury should find such a violation regardless of its assessment of such competing design considerations as risk, benefit, feasibility, and cost. The jury never made any requests which hinted it was inclined to apply the consumer expectations test without regard to a weighing of risks and benefits.
Under these circumstances, we find it highly unlikely that a reasonable jury took that path. We see no reasonable probability that the jury disregarded the voluminous evidence on the risks and benefits of the Camaro’s design, and instead rested its verdict on its independent assessment of what an ordinary consumer would expect. Accordingly, we conclude, the error in presenting that theory to the jury provides no basis for disturbing the trial judgment.8
*5722. Causation instructions.
GM next claims the trial court committed prejudicial error by refusing to instruct that any design defect was not a “substantial” or “contributing” cause of plaintiff’s “enhanced” injuries if those same injuries would have occurred even with a nondefective design. This proffered instruction conformed to GM’s trial theories that given the angle and force of the collision, the wheel would have collapsed regardless of any defect, and, in any event, that the wheel’s collapse played no part in the ankle injuries plaintiff received.
A party is entitled upon request to correct, nonargumentative instructions on every theory of the case advanced by him which is supported by substantial evidence. The trial court may not force the litigant to rely on abstract generalities, but must instruct in specific terms that relate the party’s theory to the particular case. (E.g., Hasson v. Ford Motor Co. (1977) 19 Cal.3d 530, 543 [138 Cal.Rptr. 705, 564 P.2d 857, 99 A.L.R.3d 158]; Borenkrautw. Whitten (1961) 56 Cal.2d 538, 545-546 [15 Cal.Rptr. 635, 364 P.2d 467]; Phillips v. G. L. Truman Excavation Co. (1961) 55 Cal.2d 801, 806 [13 Cal.Rptr. 401, 362 P.2d 33]; Selfw. General Motors Corp., supra, 42 Cal.App.3d 1, 10 (Self).)
GM’s proposed instruction was correct in form and substance. A manufacturer is liable only when a defect in its product was a legal cause of injury. (Cronin, supra, 8 Cal.3d at pp. 133-134.) A tort is a legal cause of injury only when it is a substantial factor in producing the injury. (Mitchell v. Gonzales (1991) 54 Cal.3d 1041, 1048-1054 [1 Cal.Rptr.2d 913, 819 P.2d 872].) If the external force of a vehicle accident was so severe that it would have caused identical injuries notwithstanding an abstract “defect” in the vehicle’s collision safety, the defect cannot be considered a substantial factor in bringing them about. (E.g., Doupnik v. General Motors Corp. (1990) 225 Cal.App.3d 849, 862-864 [275 Cal.Rptr. 715]; Endicott v. Nissan Motor *573 Corp. (1977) 73 Cal.App.3d 917, 926 [141 Cal.Rptr. 95, 9 A.L.R.4th 481]; Self, supra, 42 Cal.App.3d at p. 10.)9
The general causation instruction given by the trial court correctly advised that plaintiff could not recover for a design defect unless it was a “substantial factor” in producing plaintiff’s “enhanced” injuries. However, this instruction dealt only by “negative implication” (Self supra, 42 Cal.App.3d at p. 10) with GM’s theory that any such defect was not a “substantial factor” in this case because this particular accident would have broken plaintiff’s ankles in any event. As we have seen, GM presented substantial evidence to that effect. GM was therefore entitled to its special instruction, and the trial court’s refusal to give it was error.10
GM argues vigorously that the error is reversible per se. GM claims a California rule that the erroneous denial of instructions explaining a “central theory” of a party’s case is prejudicial as a matter of law.
Substantial authority supports GM’s view. However, GM’s contention is out of step with the usual rules governing instructional error. More significantly, it overlooks the proper application of California’s constitutional requirement that a judgment not be reversed unless error caused actual prejudice in light of the whole record. Hence, we conclude, the error at issue *574must be subjected to an examination whether actual prejudice occurred under the particular circumstances.
A judgment may not be reversed on appeal, even for error involving “misdirection of the jury,” unless “after an examination of the entire cause, including the evidence,” it appears the error caused a “miscarriage of justice.” (Cal. Const., art. VI, § 13.) When the error is one of state law only, it generally does not warrant reversal unless there is a reasonable probability that in the absence of the error, a result more favorable to the appealing party would have been reached. (People v. Watson (1956) 46 Cal.2d 818, 835 [299 P.2d 243].)
Thus, when the jury receives an improper instruction in a civil case, prejudice will generally be found only “ ‘[w]here it seems probable that the jury’s verdict may have been based on the erroneous instruction . . . .’” (LeMons v. Regents of University of California, supra, 21 Cal.3d 869, 875, quoting Robinson v. Cable (1961) 55 Cal.2d 425, 428 [11 Cal.Rptr. 377, 359 P.2d 929].) That assessment, in turn, requires evaluation of several factors, including the evidence, counsel’s arguments, the effect of other instructions, and any indication by the jury itself that it was misled. (Pool v. City of Oakland, supra, 42 Cal.3d 1051, 1069-1070.)
However, a substantial body of California decisions recites that the erroneous denial of correct specific instructions covering a civil litigant’s supportable “theory of the case” is “inherently” prejudicial. Decades old, this principle has been stated, or at least implicitly applied, in a wide variety of situations, ranging from the complete preclusion of a claim or defense (e.g., Hasson v. Ford Motor Co., supra, 19 Cal.3d 530, 548 [contributory negligence]; Phillips v. G. L. Truman Excavation Co., supra, 55 Cal.2d 801, 806 [same]; Bernal v. Richard Wolf Medical Instruments Corp. (1990) 221 Cal.App.3d 1326, 1337-1338 [272 Cal.Rptr. 41] [warranty theories in product liability action]; Paverud v. Niagara Machine & Tool Works (1987) 189 Cal.App.3d 858, 862-864 [234 Cal.Rptr. 585] [superseding cause]; White v. Uniroyal, Inc. (1984) 155 Cal.App.3d 1, 29-33 [202 Cal.Rptr. 141] [peculiar risk doctrine]) to mere lack of specificity in relating correct general principles to the particular facts (e.g., Borenkraut v. Whitten, supra, 56 Cal.2d 538, 544-546 [specific duty of care when priming automobile carburetor]; Williams v. Carl Karcher Enterprises, Inc. (1986) 182 Cal.App.3d 479, 489-490 [227 Cal.Rptr. 465] [affirmative duty to eliminate known dangerous condition in restaurant]; Ng. v. Hudson (1977) 75 Cal.App.3d 250, 261-262 [142 Cal.Rptr. 69] [“proximate cause” as including aggravation of dormant preexisting condition]; Self, supra, 42 Cal.App.3d 1, 10 [defect not “substantial factor” if same injury would have occurred regardless of defect]; see also, *575e.g., Lopez v. Ormonde (1968) 258 Cal.App.2d 176, 180 [65 Cal.Rptr. 513] [refusal of imminent peril instructions; prejudice assumed]; Edelman v. Zeigler (1965) 233 Cal.App.2d 871, 883-884 [44 Cal.Rptr. 114] [refusal of res ipsa loquitur instructions; prejudice assumed]).
The rationale generally given is that an error of this nature prevents jury consideration of the omitted “theory” and thus denies, to that extent, the right to a jury trial. We once declared that “[s]uch an error cannot be cured by the beneficent provisions of article VI, section [13 of the California Constitution]” (Phillips v. G. L. Truman Excavation Co., supra, 55 Cal.2d at p. 808), and this reasoning has been followed with little elaboration in more recent cases. (E.g., White v. Uniroyal, Inc., supra, 155 Cal.App.3d at p. 33; Ng v. Hudson, supra, 75 Cal.App.3d at pp. 261-262; Fish v. Los Angeles Dodgers Baseball Club (1976) 56 Cal.App.3d 620, 641 [128 Cal.Rptr. 807, 91 A.L.R.3d 1].)
The “inherent prejudice” line of authority is not unbroken. A number of decisions, when addressing erroneous denials of specific “theory” instructions, have assessed the actual effect of the error on the judgment. (Agarwal v. Johnson (1979) 25 Cal.3d 932, 951-952 [160 Cal.Rptr. 141, 603 P.2d 58] [court instructed generally on respondeat superior liability, but failed to instruct sua sponte on limits of vicarious liability for punitive damages; any error deemed harmless under Cal. Const., art. VI, § 13]; Hildebrand v. Los Angeles Junction Ry. Co. (1960) 53 Cal.2d 826, 831, 832 [3 Cal.Rptr. 313, 350 P.2d 65] [court instructed generally that “party who asserts the affirmative of an issue” has burden of proof, but refused plaintiff’s specific instruction that defendant has burden of proving contributory negligence; error deemed harmless after “review [of] entire record” under art. VI, former § WA (now § 13)]; Walbrook Ins. Co. v. Liberty Mutual Ins. Co. (1992) 5 Cal.App.4th 1445, 1461-1462 [7 Cal.Rptr.2d 513] [refusal in insurance bad faith action to instruct that amount of underlying personal injury verdict furnishes inference of value of claim; error deemed harmless under “all the circumstances” where other instructions invited jury to consider “strength and weaknesses” of third party’s claim]; Sesler v. Ghumman (1990) 219 Cal.App.3d 218, 226 [268 Cal.Rptr. 70] [refusal to instruct in detail on duty of care when turning left against multiple lanes of traffic; LeMons factors applied to assess prejudice; misleading argument of plaintiff’s counsel emphasized]; Montez v. Ford Motor Co. (1980) 101 Cal.App.3d 315, 322 [161 Cal.Rptr. 578] [refusal to give plaintiff’s proffered instruction defining manufacturing defect deemed harmless under art. VI, § 13]; Wechlo v. Winyard (1973) 33 Cal.App.3d 990, 996 [109 Cal.Rptr. 462] [refusal of last clear chance instruction deemed prejudicial because of evidence that jury focused closely on issues of negligence and contributory negligence]; *576see Continental Airlines, Inc. v. McDonnell Douglas Corp. (1989) 216 Cal.App.3d 388, 404-408 [264 Cal.Rptr. 779] [error to omit elements of fraud by nondisclosure, but judgment may be upheld where missing elements necessarily found in connection with other theories].)
In deciding what standard of reversibility should apply to the erroneous omission of instructions explaining the theory of a claim or defense, we take guidance from a recent decision of this court, People v. Cahill (1993) 5 Cal.4th 478 [20 Cal.Rptr.2d 582, 853 P.2d 1037] (Cahill). There we abrogated California’s traditional rule that admission of an involuntary confession as evidence in a criminal case was reversible per se. In so doing, we examined at length the purpose and application of the California Constitution’s reversible error provision, article VI, section 13.
As we explained, the predecessor of article VI, section 13 (art. VI, § 4 1/2) was added to the California Constitution in 1911. Its purpose was to counteract prior assumptions that a reviewing court could not consider the trial evidence in deciding whether an error had caused prejudice. (Cahill, supra, 5 Cal.4th at pp. 489-490.) People v. O’Bryan (1913) 165 Cal. 55 [130 P. 1042] soon made clear that “[under] the new constitutional provision the appellate courts are empowered to examine ‘the entire cause, including the evidence’ and are required to affirm the judgment ... if error has not resulted ‘in a miscarriage of justice.’ [Citation.]” (Cahill, supra, 5 Cal.4th at p. 490, quoting O’Bryan, supra, at p. 64, italics in original.)
Confusion arose after O’Bryan because California courts developed a “variety of differently worded tests” to determine whether a miscarriage of justice had occurred. (Cahill, supra, 5 Cal.4th at p. 492.) People v. Watson, supra, 46 Cal.2d 818 resolved this confusion by articulating a “generally applicable” standard. For most errors, this test permits reversal “ ‘only when the [reviewing] court after an examination of the entire cause, including the evidence, is of the “opinion” that it is reasonably probable that a result more favorable to the appealing party would have been reached in the absence of the error.’ ” (Cahill, supra, 5 Cal.4th at p. 492, quoting Watson, supra, 46 Cal.2d at p. 836.) Of course, the Watson test has since been applied equally in civil and criminal cases.
Both O’Bryan and Watson recognized that certain limited forms of error would constitute a miscarriage of justice without regard to the state of the evidence. O’Bryan spoke particularly of criminal law errors which, under Anglo-American standards of justice, deny the accused a determination of guilt or innocence “ ‘by an orderly legal procedure in which the substantial rights belonging to defendants shall be respected.’ ” (Cahill, supra, 5 Cal.4th at p. 501, quoting O’Bryan, supra, 165 Cal. at p. 65, italics added by Cahill.)
*577In Cahill, supra, we sought to give meaning to the distinction articulated by O’Bryan. As we observed, “the kinds of errors that, regardless of the evidence, may result in a ‘miscarriage of justice’ because they operate to deny a criminal defendant the constitutionally required ‘orderly legal procedure’ (or, in other words, a fair trial)—for example, the denial of the defendant’s right to a jury trial or to an impartial trial judge [citation]—all involve fundamental ‘structural defects’ in the judicial proceedings . . . rather than the improper admission of a particular item of evidence.” (Cahill, supra, 5 Cal.4th at pp. 501-502.)
Cahill noted that by their nature, “ ‘structural defects in the constitution of the trial mechanism,’ ” such as those automatically reversible in criminal cases under federal constitutional law (see Arizona v. Fulminante (1991) 499 U.S. 279, 309-310 [113 L.Ed.2d 302, 330-332, 111 S.Ct. 1246]), are not susceptible to conventional harmless-error analysis. (Cahill, supra, 5 Cal.4th at p. 493.) Additional examples from California criminal law, as cited by Cahill, include improper denial of the right to separate counsel (see, e.g., People v. Douglas (1964) 61 Cal.2d 430, 436-439 [38 Cal.Rptr. 884, 392 P.2d 964]), conflict of interest on the part of counsel (see, e.g., People v. Mroczko (1983) 35 Cal.3d 86, 104-105 [197 Cal.Rptr. 52, 672 P.2d 835]), ineffectual waiver of right to jury trial (see, e.g., People v. Holmes (1960) 54 Cal.2d 442 [5 Cal.Rptr. 871, 353 P.2d 583]), and discrimination in jury selection (see, e.g., People v. Wheeler (1978) 22 Cal.3d 258, 283 [148 Cal.Rptr. 890, 583 P.2d 748]). (Cahill, supra, 5 Cal.4th at p. 493.)
By contrast, Cahill observed, admission of an involuntary confession is mere “trial error,” that is, “ ‘error which occurred during the presentation of the case to the jury. . . .’” This category of error, said Cahill, “ ‘may . . . be quantitatively assessed in the context of other evidence presented in order to determine whether its admission was [prejudicial or harmless].’ ” (5 Cal.4th at p. 502, quoting Arizona v. Fulminante, supra, 499 U.S. at pp. 307-308 [113 L.Ed.2d at pp. 329-330].) Indeed, Cahill noted, the “admission or rejection of evidence” is among those specific forms of error for which article VI, section 13, “by its terms, directs that. . . prejudicial [effect] . . . be determined ‘after an examination of the entire cause, including the evidence.’ ” (Cahill, supra, 5 Cal.4th at p. 502.)
Earlier California authorities justified an exception for involuntary confessions on grounds that a confession is such persuasive evidence of guilt. The cases reasoned that it would be extremely difficult to determine whether improper admission of this “evidentiary bombshell” was harmless in a particular case. (See, e.g., People v. Jacobson (1965) 63 Cal.2d 319, 330 [46 Cal.Rptr. 515, 405 P.2d 555]; People v. Schader (1965) 62 Cal.2d 716, 731 *578[44 Cal.Rptr. 193, 401 P.2d 665]; People v. Parham (1963) 60 Cal.2d 378, 385 [33 Cal.Rptr. 497, 384 P.2d 1001].)
Cahill explained, however, that “[i]n relying upon this rationale . . . , the California decisions . . . lost sight of the principal purpose and significance of . . . California’s constitutional provision explicitly addressing the matter of reversible error. The recognition that confessions, ‘as a class,’ ‘[a]lmost invariably’ will provide persuasive evidence of a defendant’s guilt [citation] . . . simply means that the improper admission of a confession is much more likely to affect the outcome of a trial than are other categories of evidence, and thus is much more likely to be prejudicial under the traditional harmless-error standard. But . . . that consequence does not, in our view, justify the judicial adoption of a state-law rule that automatically and monolithically treats all improperly admitted confessions as requiring reversal of the defendant’s conviction; the California constitutional reversible-error provision was adopted for the specific purpose of eliminating just such a prophylactic approach to reversible error. [Fn. omitted.]” (5 Cal.4th at p. 503, italics in original.)
Cahill next rejected contentions that a reversible-per-se rule should be retained in order to deter the extraction of involuntary confessions, a particularly “egregious” form of official misconduct, and to guard against the unreliability of such confessions. (5 Cal.4th at pp. 505-507.) Finally, for several reasons, Cahill declined to retain the rule as a matter of stare decisis.
In particular, Cahill noted that “retention of a reversible-per-se rule, solely on the basis of stare decisis, would fail to give proper recognition to the important public policies underlying the reversible error provision set forth in California’s Constitution—policies that remain of vital significance today. . . .” (5 Cal.4th at p. 508.) Among other things, Cahill explained that when a defendant has received a fair and accurate trial despite some error, “reversal of the judgment will result either in a superfluous retrial in which the outcome is a foregone conclusion or, even more unfortunately, in a new trial whose result is altered by the loss of essential witnesses or testimony through the passage of time. In either event, public confidence in the operation of the criminal justice system is diminished.” (Id., at p. 509.)
These principles, properly adapted, apply with equal or even greater force to the issue before us. Of course, we are here concerned with a civil, not a criminal trial. But the constitutional requirement of actual prejudice cannot apply any less stringently to a civil judgment than to a criminal conviction, in which the rights of an accused threatened with deprivation of liberty are at stake.
*579Indeed, as in Cahill, the express terms of article VI, section 13 of the California Constitution weigh against automatic reversal for the kind of error we consider here. The constitutional provision explicitly mentions “misdirection of the jury” as error which warrants reversal only if, “after an examination of the entire cause, including the evidence, the court [concludes] . . . that the error . . . resulted in a miscarriage of justice.” The word “misdirection” logically includes every kind of instructional error. It seems manifest that incorrect, ambiguous, conflicting, or wrongly omitted instructions may equally “misdirect” the jury’s deliberations. Nothing in the language or history of article VI, section 13 suggests that its requirement of actual prejudice, determined by reference to “the entire cause, including the evidence,” applies to some forms of “misdirection,” but not others.
Moreover, erroneous refusal of a proffered civil instruction clearly is not a fundamental denial of the orderly legal procedure due a criminal accused. Nor is it a “ ‘structural [defect] in the . . . trial mechanism’ ” that defies evaluation for harmlessness. Instead, like the improper admission of evidence at issue in Cahill, it is trial error, a mistake that occurred in presentation of the case to the jury. By its nature, error of this kind “ ‘may ... be quantitatively assessed in . . . context... in order to determine whether its [commission] was [prejudicial or harmless].’ ” (Cahill, supra, 5 Cal.4th at p. 502, quoting Arizona v. Fulminante, supra, 499 U.S. 279, 307-308 [113 L.Ed.2d 302, 329-330].)
We are not persuaded otherwise by earlier pronouncements that certain kinds of erroneous instructional omissions in civil cases are automatically reversible because they violate a litigant’s right to jury trial. In our view, if a civil litigant was permitted to introduce evidence, cross-examine witnesses, and present argument before a fairly selected jury that rendered its honest verdict on the trial record, there has been no “structural [defect] in the constitution of the trial mechanism” that might call for automatic reversal of a civil judgment without consideration of actual prejudice. Obviously, any substantial “error which occurred during the presentation of the case to the jury” distorts or impairs the jury function to some degree. That fact cannot turn every such civil trial error into a fundamental, structural denial of the right to a jury.
Nor can we accept the traditional rationale that certain forms of instructional omission in civil cases are “inherently” prejudicial. Cases that automatically applied that theory without reference to the actual record “lost sight of the principal purpose and significance of . . . California’s constitutional provision explicitly addressing the matter of reversible error. . . .” (Cahill, supra, 5 Cal.4th at p. 503.)
*580Erroneous civil instructional omissions, like the criminal evidentiary error at issue in Cahill, may be more or less likely to cause actual prejudice, depending on their nature and context. Particularly serious forms of error might “almost invariably” prove prejudicial in fact. But it does not follow that courts may “automatically and monolithically” treat a particular category of civil instructional error as reversible per se. Article VI, section 13 of the California Constitution requires examination of each individual case to determine whether prejudice actually occurred in light of the entire record. (See Cahill, supra, 5 Cal.4th at p. 503.)
Finally, we may not blindly endorse traditional rules of automatic reversal or “inherent” prejudice in order to preserve doctrinal stability. As in Cahill, our adherence to such principles would undermine the important and still-vital requirements and policies of article VI, section 13 of the California Constitution. No form of civil trial error justifies reversal and retrial, with its attendant expense and possible loss of witnesses, where in light of the entire record, there was no actual prejudice to the appealing party.
We therefore conclude that there is no rule of automatic reversal or “inherent” prejudice applicable to any category of civil instructional error, whether of commission or omission. A judgment may not be reversed for instructional error in a civil case “unless, after an examination of the entire cause, including the evidence, the court shall be of the opinion that the error complained of has resulted in a miscarriage of justice.” (Cal. Const., art. VI, §13.) Contrary implications in prior decisions such as Self, supra, 42 Cal.App.3d 1, are disapproved and overruled.
Instructional error in a civil case is prejudicial “where it seems probable” that the error “prejudicially affected the verdict.” (See Pool v. City of Oakland, supra, 42 Cal.3d 1051, 1069; LeMons v. Regents of University of California, supra, 21 Cal.3d 869, 875; People v. Watson, supra, 46 Cal.2d 818, 836.) Of course, that determination depends heavily on the particular nature of the error, including its natural and probable effect on a party’s ability to place his full case before the jury.
But the analysis cannot stop there. Actual prejudice must be assessed in the context of the individual trial record. For this purpose, the multifactor test set forth in such cases as LeMons and Pool, both supra, is as pertinent in cases of instructional omission as in cases where instructions were erroneously given. Thus, when deciding whether an error of instructional omission was prejudicial, the court must also evaluate (1) the state of *581the evidence, (2) the effect of other instructions, (3) the effect of counsel’s arguments, and (4) any indications by the jury itself that it was misled.11
Here, GM does not even suggest that the refusal of its causation instruction caused it actual, as opposed to “inherent,” prejudice. Nonetheless, we examine the error by the standards we have set forth above. Our evaluation convinces us that the error was harmless.
At the outset, we note that the omission of GM’s proposed language did not cause an entire absence of instructional support for GM’s causation defense. (Cf., e.g., Hasson v. Ford Motor Co., supra, 19 Cal.3d 530, 548 [contributory negligence instruction refused].) The trial court instructed that plaintiff could not recover for a design defect unless the defect was a “substantial” factor in producing her “enhanced” injuries. In general terms, the instructions thus encompassed GM’s causation theory, and they did not foreclose a defense verdict on that theory.
What GM failed to obtain was a further explanation of how general principles of causation related to GM’s specific claim that plaintiff’s ankle injuries were caused by the force of the accident, not by any design defect in the Camaro. In essence, the omitted language was thus similar in function and purpose to “pinpoint” instructions. It is well settled that the erroneous refusal of “pinpoint” instructions may be deemed harmless in appropriate cases. (See, e.g., People v. Wright (1988) 45 Cal.3d 1126, 1144-1152 [248 Cal.Rptr. 600, 755 P.2d 1049].)
The fact that no other instructions covered GM’s well-supported theory with the required specificity may weigh in favor of a finding that prejudice *582occurred in this case. However, other factors show there is no reasonable probability the jury was misled or the verdict affected.
It was obvious at trial that GM’s theory of causation was a major aspect of its case. Without objection, GM produced voluminous expert evidence in support of its hypothesis that even if plaintiff’s Camaro was defective, the force of the collision was the sole “substantial” cause of plaintiff’s ankle injuries.
Again without objection, GM’s counsel devoted significant argument to this theory, and counsel articulated it very clearly. At length, counsel urged the evidence showed that the toe pan’s deformation could not have broken plaintiff’s ankles and that the actual cause of injury was plaintiff’s own inertial impact against the floorboard. Moreover, counsel emphasized, “[i]f the crash is so severe that you think [plaintiff’s] ankles would have broken anyway,” then any defect was not a cause of plaintiff’s injuries.12
In turn, plaintiff also devoted substantial attention to the causation issue raised by GM. Plaintiff’s counsel presented contrary evidence, cross-examined defense experts, and argued that GM’s theory was not persuasive on the facts. However, neither plaintiff’s counsel nor the court ever suggested that it was legally irrelevant.
Thus, the evidence and argument uniformly supported the reasonable inference that the general causation instruction allowed GM to escape liability if plaintiff’s injuries would have occurred regardless of any defect. Hence, there seems little chance the jury was actually misled. The jury itself gave no indication it was confused on the point, or that its deliberations were affected accordingly. We therefore find no reasonable probability that the *583error in refusing GM’s special instruction on causation affected the jury’s verdict. Accordingly, we conclude, the error was harmless.
Conclusion
The trial court erred when it instructed on the consumer expectations test for design defect, and when it refused GM’s special instruction on causation. However, neither error caused actual prejudice. Accordingly, the judgment of the Court of Appeal, upholding the trial court judgment in favor of plaintiff, is affirmed.
Kennard, J., George, J., Werdegar, J., and Boren, J.,* concurred.