delivered the opinion of the Court,

In this products liability case, the trial court erroneously admitted hundreds of reports of alleged accidents, almost all of which were hearsay and almost none of which were shown to involve defects like those alleged here. As the plaintiffs evidence and arguments at trial focused on the quantity of other accidents rather than the quality of the evidence regarding her own, we discount her new position on appeal that the improper admission of evidence of other accidents was unimportant. Accordingly, we reverse the judgment of the court of appeals,1 and remand for a new trial.

I

Marian Armstrong’s parents bought a 1986 Nissan 300ZX in January 1986, and transferred2 it to her five or six years later with more than 90,000 miles on the odometer. On a rainy day in October 1992, immediately after resigning from her job, Armstrong got into her car and shifted into reverse. After she “barely touched” the accelerator, the car “took off’ backwards and hit a brick building, though she was pressing the brake pedal as hard as she could. When she shifted into drive and again “barely touched” the accelerator, the car “shot forward” and struck a telephone pole, again despite application of the brakes. These collisions resulted in two broken bones in her foot and nerve *135damage, injuries the jury assessed at $900,000.

The Armstrongs had the car repaired, and about six months later another unintended acceleration occurred when a family Mend was driving the car. A service writer at a shop specializing in ZX cars told Armstrong’s father — over the telephone and without seeing the car — that the throttle cable might have been jammed by either a small rubber “boot” designed to keep dust out of the accelerator mechanism, or a white liner around the throttle cable. Armstrong’s father cut off both and discarded them.

Two years before the Armstrongs bought their car, the National Highway Traffic Safety Administration (NHTSA) received a number of complaints of unintended acceleration of 280ZX and 300ZX cars. NHTSA opened an investigation, as it has done with similar complaints involving many other vehicles. Over the next four years, the agency and two outside inspection firms conducted inspections and testing, looking at potential causes that included the fuel injection system, accelerator pedal, throttle linkages, computerized control system, cruise control, brakes, engine mounts, automatic transmission, and floor mats.

In its Closing Report, NHTSA made the following findings:

[Tjest and vehicle inspection results showed that the tested and inspected vehicles did not exhibit any condition that would cause the vehicle to accelerate at a high rate on its own, and the tested vehicles with the wide open throttle condition can be controlled by applying the brakes.

[[Image here]]

Available information indicates that inadvertent and unknowing driver application of the accelerator pedal when the driver intended to apply the brake appears to be the cause of many of the reported SA [sudden acceleration] related accidents under several ODI [Office of Defects Investigation] incident investigations, even though many of the drivers continue to believe that they had been pushing on the brake pedal.

[[Image here]]

During the course of this investigation, no safety-related defect has been detected. Further commitment of resources to determine whether such a trend may exist does not appear to be warranted.

Investigations conducted by other agencies in the United States and Canada reached the same conclusions, as did Nissan in its own investigation. Armstrong contends these conclusions were incorrect because Nissan never reported to any agency that the boot or throttle cable was the problem.

In response to the finding of driver pedal-error, Nissan recalled 1979-1987 models to install a shift-interlock system that required drivers to step on the brake before shifting out of park.3 Armstrong’s car had such an interlock. After the recall, claims of unintended acceleration dropped dramatically.

Nissan made one more design change. In 1995 Nissan redesigned the throttle cable and incorporated a smaller boot on ZX cars, at a cost of $10.66 per unit. Nissan contends the change was made to standardize parts among its vehicles, but Armstrong argues the real reason was to eliminate the defects in the design.

*136Armstrong sued Nissan Motor Co. Ltd. and Nissan North America, Inc. for products liability, negligence, gross negligence, breach of warranty, fraud, negligent misrepresentations, and violations of the Texas Deceptive Trade Practices — Consumer Protection Act.4 In support of her claim that the throttle cable was defective, Armstrong presented expert testimony from a mechanical engineer, factual testimony from four other owners who had experienced unintended acceleration in 300ZX cars, 16 written reports of unintended acceleration, and Nissan’s database of 757 consumer complaints. Nissan objected unsuccessfully to most of this evidence on grounds of hearsay, relevance, and incompetence.

The jury returned a verdict favorable to Armstrong on each of her fifteen theories of liability. The trial court rendered judgment on the jury’s findings of design, manufacturing, and marketing defects, negligence, and gross negligence, but rendered judgment notwithstanding the verdict on the other theories. Although Armstrong and Nissan had stipulated to $2 million in punitive damages if the jury found gross negligence (which it did), the trial court remitted this amount to $1.2 million.

Both sides appealed. When the court of appeals affirmed, both sides petitioned this Court for review.

II

A

This is not the first lawsuit in Texas alleging that a vehicle accelerated on its own. This Court and others have considered several such cases involving different vehicles, different asserted defects, and different alternative causes.

In Gammill v. Jack Williams Chevrolet, Inc., we affirmed summary judgment for the manufacturer of a 1988 Isuzu Trooper in a suit alleging that the accelerator pedal became entangled with wiring below the dashboard.5 The Gammills pointed to a scrape on the wiring sheath as evidence of entanglement and expert testimony that a different design could have avoided the defect.6 But while their expert on unintended acceleration was a licensed mechanical engineer with substantial experience working on cars, he had never designed them.7 Accordingly, we held the trial court did not abuse its discretion in finding him unqualified to testify about design defects; as this was the only testimony supporting the plaintiffs’ claims relating to the car’s accelerator system, we affirmed the trial court’s summary judgment finding no defect.8

In General Motors Corp. v. Hopkins, we affirmed a judgment against the manufacturer of a 1970 Chevrolet pickup based on allegations that a lockout pin jammed a carburetor in the open position, causing an influx of fuel and unintended acceleration.9 We held the verdict was supported by evidence that (1) after the accident, the lockout pin was found in the jammed position, (2) a similar malfunction had occurred with the same vehicle earlier, and (3) GM files contained a picture and correspon*137dence among GM engineers addressing the problem.10

But in Henderson v. Ford Motor Co., we reversed a judgment against the manufacturer of a 1968 Lincoln Continental based on allegations that a piece of rubber from a gasket jammed the carburetor’s throttle blades.11 Though the carburetor was found in that condition some years later, no expert testified it had been caused by an unreasonably dangerous design or that an alternative would have prevented it.12

In all of these cases, it was not enough that a vehicle accelerated when claimants swore they had done nothing. Instead, we have consistently required competent expert testimony and objective proof that a defect caused the acceleration. The courts of appeals have done the same, holding liability cannot be based on unintended acceleration alone,13 on lay testimony regarding its cause,14 or on defects not confirmed by actual inspection.15 Courts elsewhere do too.16

These requirements are not peculiar to unintended acceleration cases. We recently held in Ford Motor Co. v. Ridgway that an engine fire in an older vehicle was not evidence that the vehicle was defective; there were simply too many potential causes to assume from the one that the other must have been the culprit.17 Instead, we held that a specific defect must be identified by competent evidence and other possible causes must be ruled out.18

These requirements are especially compelling in unintended acceleration cases. Not only are there many potential causes (from floor mats to cruise control), but one of the most frequent causes (inadvertently stepping on the wrong pedal) is untraeea-ble and unknown to the person who did it.19

Accordingly, we again affirm that the mere occurrence of an unintended acceleration incident is no evidence that a vehicle is defective. While it was up to the jury in *138this case to decide what caused Armstrong’s accident, she had to present evidence that her ZX car was defective, not just that other owners experienced unintended acceleration.

B

Nor is this the first time Texas courts have considered the admissibility of evidence of similar accidents.

We addressed this question some years ago in two cases involving evidence of other accidents offered to show a railroad crossing was extra-hazardous (thus requiring additional warnings).20 In one case, we held it was reversible error to admit evidence of two previous accidents occurring during the daytime in a case involving an accident occurring on a foggy night.21 But in another, we held it was reversible error to exclude evidence of six previous nighttime accidents in a case involving the same conditions,22 as in all of them the motorists’ headlights could not illuminate the depression where the crossing was located until it was too late.23

Just as proof of other accidents may show a crossing is extra-hazardous, proof of other accidents may be introduced to show a product is unreasonably dangerous.24 But several restrictions apply.

First, as in the train-crossing cases, the other incidents must have occurred under reasonably similar (though not necessarily identical) conditions.25 The degree of similarity required depends on the issue the evidence is offered to prove.26

Second, evidence of similar incidents is inadmissible if it creates undue prejudice, confusion, or delay.27 Proof of what happened in a previous accident does not, without more, prove what happened in a current one.28 Further, prolonged proof of what happened in other accidents cannot be used to distract a jury’s attention from what happened in the case at hand.

Third, the relevance of other accidents depends upon the purpose for offering them. Other accidents are admissible for some purposes and not for others. For example, other accidents may be relevant to show whether:

*139• a product was unreasonably dangerous; 29

• a warning should have been given;30

• a safer design was available;31 or

• a manufacturer was consciously indifferent toward accidents in a claim for exemplary damages.32

Conversely, a defendant may want to introduce evidence of other accidents as part of a state-of-the-art defense,33 or the absence of other accidents to rebut claims that a product was dangerous.

Several of these uses place logical limits on what accidents are relevant, such as when the accidents occurred. Whether a product was defective must be judged against the technological context existing at the time of its manufacture.34 Product design and product warnings can take into account accidents occurring before production and sale, but not unforeseeable accidents occurring thereafter.35 Nor can exemplary damages be assessed based on hindsight; deciding whether a defendant ignored an extreme risk “requires an examination of the events and circumstances from the viewpoint of the defendant at the time the events occurred, without viewing the matter in hindsight.”36

Of course, post-sale events may sometimes be quite relevant and thus admissible. For example, Nissan’s evidence showing the dramatic drop in ZX acceleration incidents after shift-interlock systems were installed — even though occurring long after the sale to the Armstrongs— certainly seems probative of what might have been causing them all along.

We do not attempt here to categorize every instance when other accidents are admissible or every occasion when they are not. But in exercising discretion regarding admissibility, trial courts must carefully consider the bounds of similarity, prejudice, confusion, and sequence before admitting evidence of other accidents involving a product.

C

Several Texas courts have also considered the admissibility of third-party complaints that an accident occurred.

Complaint letters in a manufacturer’s files may be true, but they also may be accusatory and selfserving; they are rarely under oath and never subject to cross-examination.37 As they are necessarily out-of-court statements, they are hearsay if offered to prove the truth of the assertions therein — that the incidents com*140plained of occurred as reported.38 While a compilation of such complaints by a manufacturer might constitute a business record that such claims were (or were not39) received, they could not be a business record that such claims are true unless the employee making the record had personal knowledge of each.40 Thus, consumer complaints in a company’s files are generally hearsay within hearsay, and require their own exception in addition to that for business records generally.41

Of course, accident complaints may be admissible if offered to prove something other than truth. In Uniroyal Goodrich Tire Co. v. Martinez, we held that evidence of 34 previous lawsuits in which claimants admitted mounting a 16-inch tire on a 16.5-inch rim was admissible to show that a manufacturer knew users were not heeding its warnings.42 Though the manufacturer in that case denied the validity of those claims, it acknowledged that they resulted from mismatched tires and rims;43 as the truth of that part of each claim was admitted, they were not admitted for that purpose.

However, a few courts of appeals (including the one involved here) have gone much further, concluding that mere claims of similar accidents are generally admissible because “[t]he fact that causation in the other complaints has not been proven with absolute certainty goes to their weight, not their admissibility.”44 But a hearsay objection goes to admissibility, not weight; oaths and cross-examination are required not to make evidence absolutely certain, but to make it admissible in the first place.

Again, we do not attempt here to categorize every instance when consumer complaints may or may not be admissible. But we have never held that mere claims of previous accidents can prove a product is defective, and we decline to do so here. A number of complaints may require a prudent manufacturer to investigate, and may presage liability if those complaints are substantiated and the manufacturer does nothing. But a large number of complaints cannot alone raise a fact question that a defect exists.

Ill

Nissan argues that most of Armstrong’s evidence admitted at trial violated these standards. We agree.

A

The trial court admitted Nissan’s database of 757 complaints it had received of unintended acceleration in ZX cars. Nissan unsuccessfully objected to the database on relevance and hearsay grounds.

The database was made up entirely of hearsay reports received by Nissan. Some reports attributed unintended aecel-*141eration to a particular cause, and others did not; but none pointed to the throttle cable. More than 200 of the complaints involved cars that did not even have the allegedly defective throttle mechanism.

The court of appeals cited four reasons for admitting the database. First, the court held these accidents occurred under reasonably similar circumstances.45 As noted above, unintended acceleration can have many causes, and we have always required competent evidence of a specific defect. There is nothing in the database to suggest that the defect, if any, causing those 757 incidents was similar to any of the defects alleged here. This is not similar enough.

Second, the court concluded that “the sheer number and nature of reported incidents raise an inference that the unintended acceleration or stuck throttle was caused by something other than driver error.”46 In other words, because many people claimed their cars took off on their own, it is less likely that any of them were mistaken. This, of course, is not true. The sheer number of auto accidents in America raises no inference that most have nothing to do with drivers’ mistakes. Moreover, it is using hearsay for the truth asserted — the more hearsay there is, then the more likely it must be right.

Third, the court held that Nissan waived any objection to admission of other accidents,47 despite its specific objection to the database as hearsay and irrelevant. Armstrong asserts waiver occurred when Nissan “opened the door” to evidence of consumer complaints by offering the reports by NHTSA and others that mentioned them as well.

But the reporter’s record does not reflect that Nissan offered these exhibits before Armstrong offered and used the database. While Armstrong assures us the exhibit was offered and admitted pretrial, there is no written or oral order in the record to that effect.48 Once the trial court admitted Armstrong’s evidence of other accidents, Nissan could offer similar evidence in rebuttal without waiver.49 Without proof that Nissan’s exhibit was admitted first, Nissan waived nothing by its subsequent reply.

But more important, the two kinds of proof here were not the same. Data, findings, and reports from government agencies are generally not excludable as hearsay;50 out-of-court complaints from unknown third-parties generally are. The NHTSA report listed the numbers of reported acceleration incidents as background to indicate what and why it was investigating; Armstrong offered the database to show that ZX cars were defective. Even if the NHTSA report was admitted first, it did not hold the door open for the database.

Fourth and finally, the court of appeals indicated that the database was relevant to show “Nissan’s knowledge of the dangerous condition in its vehicles.”51 In other words, the hearsay was admissible not for the truth of the matters asserted, but to *142show Nissan’s knowledge of the truth of the matters asserted. The hearsay rules cannot be avoided by this kind of circular reasoning.

To sum up, product defects must be proved; they cannot simply be inferred from a large number of complaints. If the rule were otherwise, product claims would become a self-fulfilling prophecy — the more that are made, the more likely all must be true. Without proof that a hearsay exception applied or that any of the reported incidents were due to a defect similar to those alleged by Armstrong, the trial court erred in admitting the database of complaints.

B

The trial court also admitted approximately sixteen52 reports of incidents of unintended acceleration, three from NHTSA’s files and the remainder apparently from Nissan’s.

Nissan objected to the reports from NHTSA’s files on grounds of hearsay. Even though these were collected by a government agency, like the database they contained hearsay within hearsay. Though Armstrong claims they were admissible to show “notice,” all were reported to NHTSA eight to ten years after the Armstrongs bought their car, and as she claims not to assert any post-sale duty or duty to recall, it is unclear how the notice so much later would have made any difference to her.

Unlike the database, Nissan objected to eight of the remaining reports on relevance grounds only; there was no hearsay objection, perhaps because some of them include observations by personnel at Nissan dealerships. As four of these reports contain no mention of a defect relating to the throttle cable, the objection should have been sustained. As with the database, reports of unintended acceleration due to an unknown or some other cause would be irrelevant to Armstrong’s case.

Eight of the reports (including those in NHTSA’s files) point to a defective throttle cable or boot as the cause.53 Of the eight that might involve similar defects, only two appear to involve rapid acceleration (similar to Armstrong’s claim) rather than a stuck throttle that simply would not decelerate; the latter would be relevant only if Armstrong pressed the accelerator more than she admitted.

Nevertheless, we agree with Armstrong that on the record before us the trial court did not abuse its discretion in admitting the eight reports of allegedly similar defects. Nissan presented evidence that the possibility that a deteriorated boot might jam the throttle cable was about as likely as “being struck by lightning”; this claim was made not regarding what Nissan knew in 1986, but as a fact regarding all ZX cars sold before or after. Evidence in some of the narrative reports suggested that it could and did occur, and thus they were admissible to rebut Nissan’s claim.54

*143Of course, this requires assuming the truth of the matters asserted in those reports. But Nissan only objected to two on hearsay grounds, making no objection at all to two others. Although Nissan filed a request for a limiting instruction, no ruling appears in our record, so the remaining four were properly admitted for all purposes.55

In sum, Nissan failed to object to six of the narrative reports; four others were properly admissible under the facts of this case. We hold the trial court erred in admitting the remainder.

C

Armstrong also offered, over Nissan’s objection, the testimony of four witnesses who said they had experienced unintended acceleration events in 300ZX cars.

The first, Paul Roupinian, claimed that in 1987 he was backing his 300ZX when the car accelerated for no reason and would not stop. When Roupinian sued Nissan but could produce no evidence that his incident was caused by anything Nissan did, the court found his suit frivolous and sanctioned him $125,000.

The second, Linda Faust, testified she had an acceleration incident in May 1987. She said she shifted into neutral, the car made an “incredible noise,” and she pumped the brake before finding that ineffective. She sued Nissan, but her expert and mechanics found nothing wrong with the car, and her suit was also dismissed.

The third, Martha Ham, said that she put a 300ZX into drive and it “took off’ until she hit a ditch. She testified that she “stood” on the brake pedal with both feet, but the car accelerated anyway. She did not know the cause of her incident, and the Texas Motor Vehicle Commission denied her complaint that her car was defective.56

The fourth witness, Gary Lysdale, was the only one to mention the throttle cable, testifying that the boot on his 100,000-mile car was loose but not jamming the throttle. However, as his problems continued after installation of the redesigned cable (which Armstrong alleged would have eliminated the defect causing her accident), it could not have been caused by the same defect.

None of the four lay witnesses could verify a defect as the cause of their acceleration incidents, much less a defect similar to that alleged by Armstrong. Other than having accidents they described as unintended acceleration, these owners could show no similarity between their experiences and those involved in Armstrong’s suit. As with documentary evidence, testimonial evidence of unintended acceleration is no evidence of a defect. Thus, the trial court erred in allowing these witnesses to testify.

D

Finally, Nissan objects that Armstrong’s expert witness was not qualified to testify. The court of appeals rejected Nissan’s objection on the basis that the defect alleged here “did not require an explanation from a ‘rocket scientist,’ ”57 apparently overlooking that the expert we rejected in Gammill arguably was a rocket scientist.58 But as Nissan’s pretrial motion *144to exclude the expert’s testimony asserted not that he was unqualified but that his opinions were unreliable, we hold this complaint was not preserved for appeal.59

IV

Having found the trial court erred in admitting the database, the testimony of four witnesses, and several narrative reports regarding other incidents of unintended acceleration, we must consider whether these errors require reversal. Erroneous admission of evidence requires reversal only if the error probably (though not necessarily) resulted in an improper judgment.60 We review the entire record, and require the complaining party to demonstrate that the judgment turns on the particular evidence admitted.61

Clearly, erroneous admission is harmless if it is merely cumulative.62 But beyond that, whether erroneous admission is harmful is more a matter of judgment than precise measurement. In making that judgment, we have sometimes looked to the efforts made by counsel to emphasize the erroneous evidence63 and whether there was contrary evidence that the improperly admitted evidence was calculated to overcome.64

In this case, Armstrong’s trial attorneys clearly believed her case turned on the evidence of other incidents. At every opportunity, they emphasized the large number of general complaints of unintended acceleration rather than the small number of those involving a similar defect. For example, her trial counsel began voir dire by stating:

This case involves a defective car that injured Marian Armstrong, our client. Marian’s tragedy was not an isolated incident. She was one of many whose lives have been tragically affected by Nissan’s 300 ZX car.

He continued throughout voir dire in the same vein:

There’s a bunch of Nissan dealers here in Houston. Nissan got a bunch of complaints about these 300ZX with sudden acceleration, you expect Nissan to tell the dealers that this is a problem. Can we all agree with that?

[[Image here]]

If Nissan admits there is a defect in their car, with lots of complaints, accidents, and even deaths, what happens to *145the company? ... They are responsible.

Co-counsel for Armstrong promised the following in his opening statement:

We are going to bring you a lot of incidents involving the ZX. Bring you the government report on it.... We are going to bring you evidence about what Nissan did to try and fix these cars. Similarly, Armstrong’s sole expert witness supported his opinions by again relying on many incidents of similar acceleration without proof that they involved similar defects:

Q: [By plaintiffs counsel] Do you have an opinion, sir, whether or not there was a design defect in Marian Armstrong’s 1986 3800ZX that was a cause of her accident on October 30, 1992?

A: Yes, I do.

Q: What is your opinion?

A: My opinion is that the defect presence [sic] was well known to Nissan. They have enumerable [sic] reports that this very thing was happening.

[[Image here]]

Q: [By plaintiffs counsel] Do you have an opinion as to whether or not any of those acts would have been a cause of Marian Armstrong’s occurrence on October 30,1992?

A: Yes, I do.

Q: What is your opinion?

A: My opinion is that first Nissan B I have seen many documents from Nissan to show that they were aware of this hazard. I have also seen documents where [sic] Nissan submitted to NHTSA where they denied any problem. They hides [sic] behind the excuse, driver error. Must have pushed the gas. We found no fault with this vehicle. They would say that even at the same time them records are full of people saying my cable got stuck. That’s deceitful in the very ultimate sense.

[[Image here]]

Q: [By defense counsel] I think you told this jury yesterday that in your opinion Nissan was negligent when they didn’t fix the problem of unintended acceleration after the first report they received?

A: I did not say that. What I did say was that they had many reports of sudden acceleration and they were derelict in not responding to that situation.

[[Image here]]

Q: [By defense counsel] Your whole opinion about whether Nissan should have warned with respect to the throttle cable, assumes that the throttle cable is defective, right?

A: I would rather say that it is based upon the repeated reports by Nissan personnel that the cables had frayed. They were told this by so many people, they should have realized there is a problem here and responded to it. I do not view that as an assumption, but more a reaction to the material I’ve looked at.

Q: Maybe I am confused. I thought you just told us a second ago that you hadn’t seen a single document where anybody had suggested that the cable was frayed.

A: They said it was stuck and failed. They did not use the words frayed. And I answered your specific question. We have those documents and many more that show that Nissan personnel, customers, a whole wide range of people had informed Nis*146san that the cable failure was causing their car to accelerate inappropriately.

Armstrong’s closing argument began similarly:

What is the essence, the real essence of this case? It’s about a machine that fails without warning. It’s about a company that knows that but will not admit that.

[[Image here]]

It’s a company in essence that will say to you, what is happening with these cables ... it happens on the same level as getting struck by lightning.... Gary Larson was struck by lightning in '88, '89, '89, '89, several times throughout the year; '94 again Gary Lysdale was struck by lightning. Martha Ham was struck by lightning. Paul Roupinian was struck by lightning and now faces 125 thousand dollar judgment over his head, a 29 year old man. All these families are struck by lightning.

Nissan’s counsel complained about this emphasis, beginning his closing argument with the following observation:

I have noticed a trend in this case from the beginning to now at the very end. The majority of the plaintiffs case has focused on things other than her accident, other than the facts of her accident. Now why would that be? Why would they focus on other incidents, other accidents and not hers? What is this case about, folks?

In rebuttal, Armstrong’s trial counsel replied:

[Defense counsel] said, you know, I only saw two incidents of throttle cable problems. You know, I have a problem with that because I keep seeing these reports .... Those are just some of the reports. You can look in here and see how many more there are. There are lots of them.

Finally, Armstrong’s counsel closed his argument with the following:

Remember Linda Faust and these [other] people [who] came here. Nobody is paid. Everybody is here because they have had a problem and they’re angry that they are called idiots and stupid for putting their foot on the gas or having the mat tie up in their cars.

As Armstrong’s trial attorneys spent much of the trial trying to persuade the judge and jury that other incidents of unintended acceleration were of great significance, we reject her new argument on appeal that they were not.

Though we reach a different conclusion from the court of appeals on admissibility, we agree with its assessment of the crucial nature of this evidence:

We agree with Nissan’s argument that the admission of such a high volume of other unintended acceleration incidents was prejudicial to its claims asserted at trial. However, we disagree with Nissan’s arguments that the trial court abused its discretion in deciding to admit the evidence.65

Three times in its opinion, the court of appeals relied on the quantity of other accidents rather than their quality to find the jury’s verdict factually and legally sufficient:

The jury could reasonably have concluded from this “mountain” of related incidents that it was more probable than not the unintended acceleration had been caused by some mechanical malfunction, *147not by some error on the part of the driver.66

[[Image here]]

Although most of the incident reports did not attribute a technical cause for the unintended acceleration, the sheer number and nature of reported incidents raise an inference that the unintended acceleration or stuck throttle was caused by something other than driver error.67

[[Image here]]

In sum, the evidence showed that Nissan had within its possession and knowledge numerous documented reports of incidents of unintended acceleration and stuck throttles caused by throttle cable boot failure and that Nissan never advised NHTSA or its own customers about these potential safety issues.... Based on this evidence, the jury could reasonably have concluded that Nissan, despite knowledge of a defect, had failed to take any action or sufficient action to correct the problem or to warn its customers or the public about the danger.68

The court’s reliance on the “high volume,” “mountain,” “sheer number,” and “numerous” reports of other incidents fails to acknowledge that at most a handful of them related specifically to the throttle cable defects Armstrong alleged. As the experienced justices of the court of appeals made this mistake, it is probable that the jurors did as well.

Armstrong argues that because eight reports of throttle cable boot problems were properly admitted, the remainder were cumulative and thus harmless. But there is a difference in the context of this case between cumulative and overwhelming; a jury trying to decide whether ZX cars were unreasonably dangerous might need to know whether the true number of similar problems was closer to eight or eight hundred. And as shown above, Armstrong’s attorneys did not treat the large number of hearsay complaints as merely cumulative; it was the thrust of her case.

Finally, this case was not like Hopkins or Henderson, as no one ever observed either the boot or liner actually jamming the throttle cable. Armstrong’s father did not, and he was the only one who saw them. While he testified both were worn and loose, he cut both off and threw them away before anyone else saw them, taking the photographs admitted at trial only after he had done so. A service writer for an auto repair shop said he frequently saw loose throttle cable boots, but did not testify he had ever seen one actually jam a throttle cable, and rendered no expert opinions as the trial court found he was unqualified to do so.69

Armstrong’s expert only testified that the cause of her accident could have been a cruise control defect, the throttle cable boot, or the liner, but that he had “no way of knowing which.” Because the alleged boot and finer defects could hold a throttle open but not open it wider, he did not try to explain how Armstrong’s car could have accelerated wildly when she “barely touched” the accelerator; he simply assumed she did otherwise.

Of course, there was other evidence to support Armstrong’s claim. As noted, there was some evidence that the boot could jam a ZX cable, and that Nissan may *148have known that before 1986. Further, the redesigned cable and boot presented some evidence of a feasible alternative. Armstrong’s expert testified that, though he was not a doctor, he did not believe she could have broken her foot where she did if her foot was on the accelerator rather than the brake (as she said it was).

The question before us is not whether the jury’s verdict was right or wrong; we only must decide whether the erroneously admitted evidence probably made the judgment improper.70 In this ease, the evidence of other incidents was far more than cumulative: it was emphasized at every opportunity; it was used to prove a defect when the actual evidence had been destroyed; and it was calculated to show Nissan was malicious rather than mistaken in suggesting the accident was Armstrong’s fault.

It also violated the long-standing rule in Texas that proof of unintended acceleration is not proof of a defect. Under that rule, proof of many instances of unintended acceleration cannot prove a defect either; 71 a lot of no evidence is still no evidence.

For all these reasons, we hold that the evidence improperly admitted in this case probably resulted in an improper judgment. Accordingly, we reverse and remand for a new trial without it.72

V

As we must remand for a new trial, we address the remaining arguments only briefly.

First, Nissan asserts there was no evidence to support the trial court’s award of exemplary damages, pointing out that the court of appeals relied heavily on matters Nissan learned only after the vehicle here was designed, manufactured, and sold.73 In Transportation Ins. Co. v. Morid, we prohibited post-hoc gross negligence, holding that whether a defendant’s acts or omissions involved an extreme risk “requires an examination of the events and circumstances from the viewpoint óf the defendant at the time the events occurred, without viewing the matter in hindsight.”74 But as there was also testimony that could be construed to indicate Nissan knew of acceleration claims relating to the boot from 1984, we remand the exemplary damage claim with the remainder of the case.

Next, Nissan argues that Armstrong’s claims assert a failure to take post-sale remedial measures, and thus conflict with our decision in American Tobacco Co. v. Grinnell.75 But Armstrong denies she is *149making such a claim, and the court of appeals never purported to create such a duty.76 Nissan’s complaint under this point instead appears to be that the trial court admitted reports of post-sale incidents, complaints we have addressed in part III above.

Third, Nissan argues that the trial court erred in charging the jury that Nissan was negligent per se in failing to notify the United States Government and Armstrong “when it learn[ed] the vehicle ... contained] a defect and decid[ed] in good faith that the defect [was] related to motor vehicle safety.”77 Nissan did notify the government of the complaints of unintended acceleration in ZX cars, and the subsequent investigation was not concluded until three years after the Armstrongs bought their car. Even assuming Nissan’s notification was incomplete and that a violation of the statute could be asserted in an individual citizen’s negligence case, there is no evidence that any report to the government would have resulted in a change in ZX cars before the Armstrongs bought theirs. Thus, no violation of the statute caused their injury.78

In her appeal, Armstrong argues that the trial court erred in remitting part of the exemplary damages. Because we have concluded inadmissible evidence tainted the former trial and requires a new one, this point is moot.

Armstrong also argues the trial court erred in rendering judgment notwithstanding the verdict on her claims asserting fraud, negligent misrepresentation, breach of warranty, and violation of the DTPA. Armstrong obtained her car from her parents six years after they bought it from Nissan; Nissan had no involvement or pecuniary interest in the transaction.79 There was no evidence of any specific warranty or representation Nissan ever made to her, and no duty to disclose either.80 The trial court did not err in dismissing these claims.

Accordingly, we reverse the judgment of the court of appeals, and remand the case for a new trial on Armstrong’s remaining claims in accordance with this opinion.

concurring.

I agree with the Court that judges must be especially vigilant in assuring that evidence of other incidents involving a product to show a specific product defect must be sufficiently similar to be admissible. *150Other-incidents evidence like much of that presented in this case poses a significant risk of confusing the jury and creating prejudice, while its probative value is relatively low. Accordingly, a high degree of similarity to the specific defect in issue must be shown to meet the threshold for admissibility. I also agree with the Court that the number of other incidents improperly admitted in this case to show that the plaintiffs vehicle was defective as alleged resulted in reversible error. I write separately, though, out of concern that the breadth of the Court’s opinion could cause the trial court to perceive that its hands are tied upon retrial.

The Court categorically deems much of the plaintiffs’ evidence inadmissible, when such evidence might prove to be admissible for other purposes as the retrial unfolds. Nissan could, for example, present evidence that opens the door to other-incidents evidence. And some of the evidence that the Court dismisses out of hand could prove to be the proper subject of rebuttal, depending upon how the case develops. We simply will not know until the case is retried, and the experienced trial judge in this case should be free to reevaluate such evidence as. it is presented under the parameters the Court describes.

I am also concerned with the Court’s view that the “sheer number” of other incidents could never raise an inference that unintended acceleration was caused by something other than driver error. 145 S.W.3d at 147. If a particular vehicle model had experienced, say, thousands of unintended-acceleration incidents, while other comparable models had experienced only two, an inference might be raised that driver error was not the cause.

In sum, I agree with the similarity standard the Court sets and its conclusion that harmful error occurred in this case. But for the foregoing reasons I cannot fully join the Court’s opinion, though I concur in its judgment.