Justice MEDINA
delivered the opinion of the Court,
in which Chief Justice JEFFERSON, Justice HECHT, Justice O’NEILL, Justice GREEN, Justice WILLETT, and Justice GUZMAN joined, and in Part III of which Justice WAINWRIGHT joined.
In this wrongful death and survival action, stemming from a multi-fatality vehicular accident, we consider the reliability of an accident reconstruction expert’s testimony, the legal sufficiency of the evidence supporting the verdict, and whether the admission of evidence concerning the illegal immigrant status of one of the parties to the accident was harmful error. The court of appeals, in a divided decision, concluded.that the expert’s testimony was reliable and therefore legally sufficient to support the plaintiffs’ verdict. 224 S.W.3d 870, 888. The court also held that the driver’s illegal status was relevant impeachment evidence or, alternatively, its admission was harmless error. Id. at 897. We agree that the trial court did not abuse its discretion in admitting the expert’s testimony. However, we do not agree that evidence of the driver’s illegal status was either relevant or harmless. Accordingly, we reverse the court of appeals’s judgment and remand the cause for a new trial.
I. The Litigation
Several members of the Hughes family were killed when their vehicle collided with an eighteen wheel tractor-trailer rig heavily loaded with gravel. The accident occurred outside the city of Paradise on [234]*234Highway 114, a two-lane highway. At the time of the accident, Kimberly Hughes was driving west toward Paradise with four other family members in her GMC Yukon. Ricardo Rodriguez, who was driving the gravel truck for TXI Transportation Company (“TXI”), was traveling east in the opposite direction. For reasons in dispute, the Yukon crossed the center line into the eastbound lane, collided with the gravel truck and careened down the length of its trailer. At the gravel truck’s tail end, the Yukon spun sideways into the path of an eastbound Ford pickup. The resulting collision killed everyone in the Yukon except Hughes’s infant grandson.
Hughes’s husband and other family members sued Rodriguez and his employer, TXI, for the deaths. After a seven-day trial, a jury found that Rodriguez’s and TXI’s negligence proximately caused the accident, and awarded compensatory and exemplary damages. The trial court rendered judgment on the verdict. The court of appeals set aside the award of exemplary damages, but otherwise affirmed the judgment against Rodriguez and TXI.1 224 S.W.3d at 881.
What caused the Yukon to cross the center line into Rodriguez’s eastbound lane was the critical issue at trial. Both sides relied on accident-reconstruction experts to explain their respective theories. Hughes’s accident-reconstruction expert opined that the gravel truck crossed the center line first, forcing Hughes to steer defensively into the eastbound lane where the collision occurred.
TXI sought to exclude Hughes’s expert, objecting that his opinion was unreliable. TXI also objected to evidence regarding Rodriguez’s status as an illegal immigrant on grounds of relevance and prejudice. Because the trial court overruled both objections, the jury learned Rodriguez had previously been deported and had made several misrepresentations regarding his immigration status to obtain his Texas commercial driver’s license and his employment with TXI. The dissent in the court of appeals concluded that the trial court had erred by admitting the expert testimony of Hughes’s accident reconstruc-tionist and the evidence of Rodriguez’s illegal immigrant status. Id. at 922 (Gardner, J., dissenting). We granted TXI’s petition for review to consider these issues.
II. The Accident-Reconstruction Expert
TXI argues the trial court erred by overruling its timely objection to Hughes’s reconstruction expert, Dr. Kurt Marshek, whom it contends expressed an unreliable opinion that Rodriguez caused the accident by crossing the center line first.
A. The Standard of Review
For an expert’s testimony to be admissible, the expert witness must be qualified to testify about “scientific, technical, or other specialized knowledge,” Tex.R. Evid. 702, and the testimony must be relevant and based upon a reliable foundation. Exxon Pipeline Co. v. Zwahr, 88 S.W.3d 623, 628 (Tex.2002). An expert’s testimony is relevant when it assists the jury in determining an issue or in understanding other evidence. Tex.R. Evid. 702. But, expert testimony based on an unreliable foundation or flawed methodology is unreliable and does not satisfy Rule 702’s relevancy requirement. E.I. du Pont de Nemours & Co. v. Robinson, 923 S.W.2d [235]*235549, 556-57 (Tex.1995) (discussing TEX. R. EVID. 702).
When the reliability of an expert’s testimony is challenged, courts “ ‘should ensure that the [expert’s] opinion comports with the applicable professional standards.’ ” Helena Chem. Co. v. Wilkins, 47 S.W.3d 486, 499 (Tex.2001) (quoting Gammill v. Jack Williams Chevrolet, Inc., 972 S.W.2d 713, 725-26 (Tex.1998)). To aid in that determination, we have suggested several factors to consider when assessing the admissibility of expert testimony under Rule 702.2 We have emphasized, however, that these factors are non-exclusive, and that they do not fit every scenario. Gammill, 972 S.W.2d at 726. They are particularly difficult to apply in vehicular accident cases involving accident reconstruction testimony. Ford Motor Co. v. Ledesma, 242 S.W.3d 32, 39 (Tex.2007) (citing Cooper Tire & Rubber Co. v. Mendez, 204 S.W.3d 797, 802 (Tex.2006)); see also Gammill, 972 S.W.2d at 727. Nevertheless, the court, as gatekeeper, “must determine how the reliability of particular testimony is to be assessed.” Gammill, 972 S.W.2d at 726. Rather than focus entirely on the reliability of the underlying technique used to generate the challenged opinion, as in Robinson, we have found it appropriate in cases like this to analyze whether the expert’s opinion actually fits the facts of the case. Volkswagen of Am., Inc. v. Ramirez, 159 S.W.3d 897, 904-05 (Tex.2004). In other words, we determine whether there are any significant analytical gaps in the expert’s opinion that undermine its reliability. Id.
B. The Expert’s Testimony
Dr. Kurt Marshek, an emeritus professor of mechanical engineering at the University of Texas, testified for Hughes. In preparing for his testimony, Marshek reviewed the police accident report and photographs from the accident scene, visited and took measurements at the accident site, specifically measured the gouge and scrape marks created by the accident, ran skid tests with an exemplar vehicle and measuring device to determine the roadway’s coefficient of friction, inspected and photographed the Yukon, collected data on the Yukon’s speed and braking during the five seconds before impact from the vehicle’s “black box,” performed a time-distance analysis, and reviewed the accident scene witnesses’ statements and depositions. Employing this data, Marshek rendered drawings of the accident site to illustrate his theory of the accident. Mar-shek’s theory was that Rodriguez left his lane of travel, crossed over the center line into the westbound lane, and partially reentered his eastbound lane before the initial impact with the Yukon. Marshek further concluded Kimberly Hughes steered sharply left into the eastbound lane to avoid Rodriguez’s gravel truck, which then at least partially occupied her lane, resulting in the collision in Rodriguez’s eastbound lane.3
[236]*236Using the physical evidence, Marshek described his version of the initial collision and each vehicle’s subsequent movements. The first impact occurred with the gravel truck’s second axle, creating downward pressure on the Yukon’s tire and forcing the rim to carve a gouge in the eastbound lane six inches from the center line. Reddish paint and rubber marks on the gravel truck’s tires revealed where the Yukon made contact with the tires at the second, third, fourth, and fifth axles. Rim and axle damage to the second and fourth axles demonstrated more substantial contact. After the initial collision, the gravel truck’s significant mass dictated the Yukon’s direction, forcing the Yukon’s rear end to move clockwise and adopt the gravel touch’s trailer’s angle. While following this angle, the Yukon’s front left rim first gouged and then scraped the concrete at an angle to the center line. After hitting the fourth axle, the Yukon’s left rear rim moved back toward the centerline creating a scrape mark. As it cleared the trailer’s end, the Yukon was fully in its westbound lane, moving slightly sideways before it reentered the eastbound lane, colliding with the Ford pickup. Meanwhile, the gravel truck applied its brakes 128 feet after the point of impact, leaving toe marks on the road until the truck rested 486 feet away.
C. TXI’s Reliability Complaints
TXI complains Marshek’s testimony is no evidence that Rodriguez proximately caused the collision. Marshek was the only witness to suggest the gravel truck crossed the center line, but TXI assails his testimony, arguing that (1) Marshek incorrectly assumed that the gouge mark pinpointed the place on the road where the Yukon collided with the gravel truck’s second axle; (2) Marshek incorrectly assumed the gouge mark indicated the angle of the gravel truck at the moment the Yukon struck it; (3) Marshek calculated the gravel truck’s position based on an imprecise witness time estimate contrary to proper protocol; and (4) Marshek selectively relied on eyewitness line-of-sight testimony.
TXI claims Marshek’s theory — that the Yukon’s collision with the second axle created the gouge mark — lacks any factual foundation.4 However, some facts do support Marshek’s theory. The gravel truck’s second and fourth axles were the most heavily damaged, and thus may signify the most likely collision points capable of creating the gouge. Marshek acknowledged that the severe damage to the fourth axle could indicate where the Yukon gouged the road, but rejected the possibility based on the additional scrape marks present in the eastbound lane after the gouge. Marshek matched these scrapes with subsequent impacts at the third and fourth axles. While disputing other points, all experts agreed the Yukon began moving counterclockwise back into the westbound lane after colliding with the fourth axle. As Marshek testified, had the fourth axle collision caused the gouge, there would have been no further event in the eastbound lane to create the additional scrape marks before the Yukon re-entered the westbound lane.
TXI also claims Marshek admitted during cross-examination that the gouge mark did not signify the initial collision with the [237]*237second axle. Marshek testified that the Yukon would have traveled eleven feet after colliding with the second axle, assuming it took one-eighth of a second for its wheel damage to create the gouge mark. However, Marshek estimated that the actual time from initial impact to the rim gouging the pavement would normally be one-tenth to one-twentieth of a second, and that here the impact between the Yukon and second axle created “extra drag” with the larger truck tire applying a downward force on the Yukon’s wheel, inhibiting its lateral movement. Thus, contrary to TXI’s claim, Marshek did not concede that the gouge mark would have been made eleven feet from the point of initial impact with the second axle.
TXI next argues Marshek’s conclusion that the gouge mark reflects the gravel truck’s angle during the collision with its second axle is unreliable because Marshek did not rule out the possibility the gouge mark might have been created during subsequent impacts with the gravel truck’s tires and axles. An expert’s failure to rule out alternative causes of an incident may i’ender his opinion unreliable. Merrell Dow Pharm., Inc. v. Havner, 953 S.W.2d 706, 720 (Tex.1997). However, that is not the case here. Marshek pointed to scrape marks and other physical evidence to conclude the gouge mark occurred during the collision with the gravel truck’s second axle, which effectively eliminated other causes of the gouge mark. He also testified the additional scrapes were created by the Yukon and angled in roughly the same direction as the gouge mark.
Marshek’s gouge-mark-angle theory finds support in the physical evidence. As Marshek explained, the Yukon weighs one-sixteenth of the gravel truck, making the collision analogous to a fly hitting a boulder. The gravel truck’s weight was distributed along the trailer, so when the Yukon impacted the gravel truck’s tires and axles it conformed to the trailer’s angle, gouging and scraping the road at an angle to the center line. Further, Mar-shek found additional support in the angle of the gravel truck’s brake marks. He testified the direction and length were consistent with the gouge mark angle and consistent with the gravel truck re-entering its eastbound lane. Moreover, Mar-shek tried to line- up the gouge mark and the brake marks using the assumption that the gravel truck remained in its eastbound lane. He concluded the brake marks would not line up unless Rodriguez executed a dangerous steering maneuver likely resulting in a rollover or spillage that did not occur.
TXI also contends Marshek incorrectly estimated the gravel truck’s position by distorting Rodriguez’s testimony and ignoring accepted accident reconstruction protocol. Rodriguez testified that he turned the gravel truck to the right in an attempt to avoid the collision, but his estimates of how long he turned varied from “probably one second or less” to “two or three seconds, I think.” TXI argues Mar-shek distorts Rodriguez’s testimony by relying on these statements, yet rejecting Rodriguez’s assertion that he never crossed the center line. Further, it contends Marshek violated accident reconstruction protocol by relying primarily on Rodriguez’s time estimates instead of physical data.
Marshek’s reliance on Rodriguez’s statements does not distort Rodriguez’s testimony. In City of Keller v. Wilson, we said that “evidence cannot be taken out of context in a way that makes it seem to support a verdict when in fact it never did.” 168 S.W.3d 802, 812 (Tex.2005) (citing Bostrom Seating, Inc. v. Crane Carrier Co., 140 S.W.3d 681, 684-85 (Tex.2004)). We provided an example: “If a witness’s state[238]*238ment T did not do that’ is contrary to the jury’s verdict, a reviewing court may need to disregard the whole statement, but cannot rewrite it by disregarding the middle word alone.” City of Keller, 168 S.W.3d at 812. Rodriguez made two statements: (1) he did not move out of his lane, and (2) he turned right immediately before the collision. Rather than cherry-picking parts of Rodriguez’s testimony or twisting its meaning, Marshek simply illustrated a possible inconsistency in Rodriguez’s testimony based on his review of the physical evidence.
Marshek’s use of Rodriguez’s testimony also did not violate accepted accident reconstruction protocol. According to TXI’s testifying expert, John Painter, an accident reconstruction specialist uses witness statements to help fill gaps after the specialist analyzes the physical data. Painter acknowledged eyewitness statements assist in reconstructing an accident, but implied such statements cannot be an expert’s primary data source. As discussed above, Marshek based the gravel truck’s position on the physical evidence— the gouge mark angle, the subsequent scrapes’ angles, and the gravel truck’s brake marks—using Rodriguez’s testimony solely to bolster his theory. Although his time estimates changed, Rodriguez consistently maintained that he turned to the right before the collision. Given the gravel truck’s speed, Marshek concluded that even with only one second of movement (Rodriguez’s lowest estimate), Rodriguez would have started the turn from the Yukon’s lane.
TXI similarly complains Marshek distorts another witness’s testimony by crediting the witness’s statement that he did not see the Yukon until it passed the gravel truck’s trailer while ignoring the same witness’s assertion that he never saw the gravel truck cross the center line.5 However, Marshek discussed the witness’s testimony only in response to questions regarding Painter’s line-of-sight analysis.6 When asked whether the possibility that the gravel truck re-entering the eastbound lane blocked the witness’s view of the Yukon until it cleared the truck’s trailer supported Marshek’s theory, he responded, Wes, it would.” However, Marshek did not ground his theory upon the witness’s testimony, but instead based it on other evidence.
Lastly, TXI asserts that Marshek conceded his theory to be speculation when he admitted that “nobody knows what the steering was ... it’s all total speculation.” Read in context, however, this comment was directed at Painter’s use of a computer simulation, and its inability to consider [239]*239the vehicles’ specific steering angles. Rodriguez testified that he turned to the right immediately before the collision, and Mar-shek confirmed that angle from the physical evidence and Rodriguez’s testimony.
D. Conclusion
Expert testimony is unreliable when “ ‘there is simply too great an analytical gap between the data and the opinion proffered.’” Ledesma, 242 S.W.3d at 39 (quoting Gen. Elec. Co. v. Joiner, 522 U.S. 136, 146, 118 S.Ct. 512, 139 L.Ed.2d 508 (1997)). Expert testimony is also unreliable if it is not grounded in scientific methods and procedures, but is rather based upon subjective belief or unsupported speculation. Coastal Transp. Co. v. Crown Cent. Petrol. Corp., 136 S.W.3d 227, 232 (Tex.2004). Expert testimony lacking a proper foundation is incompetent, City of Keller, 168 S.W.3d at 813, and its admission is an abuse of discretion. Cooper Tire, 204 S.W.3d at 800. The court’s ultimate task, however, is not to determine whether the expert’s conclusions are correct, but rather whether the analysis the expert used to reach those conclusions is reliable and therefore admissible. Zwahr, 88 S.W.3d at 629 (citing Gammill, 972 S.W.2d at 728).
Reliability may be demonstrated by the connection of the expert’s theory to the underlying facts and data in the case. Two recent cases illustrate the point. Compare Ledesma, 242 S.W.3d at 40-41 (concluding that a complaint about expert’s testimony went to its weight, not its admissibility) with Ramirez, 159 S.W.3d at 906 (concluding an expert’s testimony was unreliable because it was based on a subjective interpretation of the facts rather than scientific analysis). Both cases involved auto accidents allegedly caused by the failure of a defective mechanical part. The question in both cases was whether the failure of the part caused the accident or resulted from it.
In Ledesma, a metallurgical and mechanical engineer testified extensively about his theory of how a u-bolt came to be under-torqued on the rear leaf spring and axle assembly of a Ford truck. 242 S.W.3d at 37-38. He further explained how this defect caused the axle assembly to come apart which, in turn, caused the drive shaft to separate from the transmission. Id. at 37. The expert supported his theory with observations and measurements from the physical evidence and the manufacturer’s own specifications. Id. at 37-38. Although there was some question as to when the part failed, the expert pointed to other physical evidence to support his theory regarding the u-bolt’s failure as the triggering event for the accident. Id. at 38. We concluded that the manufacturer’s complaints about the expert testimony ultimately went to its weight and not its admissibility. Id. at 40-41.
In Ramirez, the expert’s theory was that a bearing defect in the left rear wheel assembly of a Volkswagen Passat caused a loss of control when that wheel became detached from its axle. 159 S.W.3d at 904. Although detached from the stub axle, the wheel was found under the rear wheel well after the accident. Id. at 902. Critical to the expert’s theory was the assumption that the detached wheel remained pocketed in the wheel well throughout a turbulent and high-speed accident sequence, involving a grass and concrete median and another vehicle. Id. at 904. The expert proposed the “laws of physics” explained his assumption, but did not connect his theory to any physical evidence in the case or to any tests or calculations prepared to substantiate his theory. Id. at 904-06. We concluded the expert’s testimony was unreliable because it was “not supported [240]*240by objective scientific analysis” but rather rested upon the expert’s “subjective interpretation of the facts.” Id. at 906. As we have repeatedly said, “‘a claim will not stand or fall on the mere ipse dixit of a credentialed witness.’ ” City of San Antonio v. Pollock, 284 S.W.3d 809, 818 (Tex.2009) (quoting Burrow v. Arce, 997 S.W.2d 229, 235 (Tex.1999)).
Marshek’s testimony here, however, was neither eonclusory nor subjective. His observations, measurements, and calculations were, as in Ledesma, tied to the physical evidence in the case which likewise provided support for his conclusions and theory. Marshek’s expert testimony thus meets our standard for reliability, and the trial court therefore did not abuse its discretion by admitting the testimony.
III. The Illegal Immigrant Issue
TXI next argues that it was error to admit evidence of Rodriguez’s immigration status and his misi-epresentation of that status in order to live and work in this country. TXI complains that Rodriguez’s status as an illegal immigrant was irrelevant to any issue in the case. TXI asserts instead that Rodriguez’s status was imper-missibly used to inflame the jury and impeach Rodriguez’s credibility. In sum, TXI submits that repeated questions on this subject prejudiced its defense and effectively denied it a fair trial.
Hughes argues, however, that Rodriguez’s misrepresentations about his qualifications and experience as a commercial truck driver were relevant to claims of negligent hiring and negligent entrustment. In particular, he relies on the Federal Motor Carrier Safety Regulation Act (FMCSRA), which defines mandatory employment checks motor carriers must make when hiring new drivers. Under these regulations, a carrier must ensure that prospective drivers have a commercial license, have a working knowledge of English, and possess the training or experience to safely operate a commercial vehicle. 49 C.F.R. §§ 383.23, 391.11(b)(2)-(7), 391.15.
A. The Negligent-Hiring/Negligent-Entrustmenl Claim
In a negligent-hiring or negligent-entrustment claim, a plaintiff must show that the risk that caused the entrustment or hiring to be negligent also proximately caused plaintiffs injuries. See Fifth Club, Inc. v. Ramirez, 196 S.W.3d 788, 796 (Tex.2006) (stating “[n]egligence in hiring requires that the employer’s ‘failure to investigate, screen, or supervise its [hirees] proximately caused the injuries the plaintiffs allege’ ” (quoting Doe v. Boys Clubs of Greater Dallas, Inc., 907 S.W.2d 472, 477 (Tex.1995))); Schneider v. Esperanza Transmission Co., 744 S.W.2d 595, 596-97 (Tex.1987). To sustain such a claim based on a failure to screen, a plaintiff must show that anything found in a background check “would cause a reason able employer to not hire” the employee, or would be sufficient to put the employer “on notice that hiring [the employee] would create a risk of harm to the public.” Fifth Club, 196 S.W.3d at 796-97. The plaintiff must also prove that the risk that caused the entrustment or hiring to be negligent caused the accident at issue. Schneider, 744 S.W.2d at 597. Therefore, a plaintiff will not succeed on a negligent entrustment or hiring claim where an investigation would not have revealed the risk. See, e.g., Doe, 907 S.W.2d at 477 (finding the failure to prove negligent hiring as a matter of law because screening would not have indicated a specific risk); Fifth Club, 196 S.W.3d at 796 (noting a background check might only have shown that the employee was violating terms of [241]*241employment with another employer, but not the employee’s proclivity for violence).
We have said a claim for negligent hiring or entrustment cannot lie if “[t]he risk that caused the entrustment to be negligent did not cause the collision,” Schneider, 744 S.W.2d at 597, and if a “defendant’s negligence did no more than furnish a condition which made the injury possible.” Doe, 907 S.W.2d at 477. Here, Rodriguez’s immigration status did not cause the collision, and was not relevant to the negligent entrustment or hiring claims-— even if TXI’s failure to screen, and thus its failure to discover his inability to work in the United States, “furnished [the] condition” that made the accident possible. Id. We agree with the court of appeals “that neither Rodriguez’s status as an illegal alien or his use of a fake Social Security number to obtain a commercial driver’s license created a foreseeable risk that Rodriguez would negligently drive the gravel truck.” 224 S.W.3d at 914.
B. Use of Immigration Status as Impeachment Evidence
The court of appeals concluded, however, that the evidence of Rodriguez’s immigration status was nevertheless admissible “to impeach his contrary trial testimony.” 224 S.W.3d at 897. This impeachment apparently related to Rodriguez’s trial testimony that he never lied to get a driver’s license and did not know whether he had a legal right to work in the United States. Id. at 897 n. 32. Relying on Texas Rule of Evidence 801(e)(2)(A), the court concluded that Rodriguez, as a party, could be impeached “with evidence of his own prior verbal statements.” Id. at 897. The court further concluded that, because the statements of a party are not hearsay, see id., it was unnecessary to “address complaints that Rodriguez’s immigration status was not relevant and was more prejudicial than probative.” Id. at 897 n. 32. We fail to see the connection.
Rule 801(e)(2)(A) provides that a party admission is not hearsay. Whether impeachment evidence is hearsay, however, has nothing to do with the relevancy requirement in Rules 401 and 402, or Rule 403’s requirement that evidence should be excluded if its prejudicial effect substantially outweighs any probative value. See Bay Area Healthcare Group, Ltd. v. McShane, 239 S.W.3d 231, 235 (Tex.2007) (stating that, “subject to other Rules of Evidence that may limit admissibility [citing, among other evidentiary rules, Rules 402 and 403], any statement by a party-opponent is admissible against that party”); Willover v. State, 70 S.W.3d 841, 846 n. 9 (Tex.Crim.App.2002) (noting that non-hearsay “still must meet other requirements for admissibility, such as relevance”). Thus, the observation that Rodriguez’s statements are not hearsay neither establishes them admissibility nor explains why other witnesses were permitted to be questioned about Rodriguez’s immigration status, or why extrinsic evidence was admitted on the subject.
C. The Error
Although Rodriguez’s statements about his immigration status may have been offered for impeachment as pri- or inconsistent statements, they were not admissible for at least two different reasons. First, Rodriguez’s immigration status was clearly a collateral matter, that is, a matter that was “not relevant to proving a material issue in the case.” Poole v. State, 974 S.W.2d 892, 905 (Tex.App.-Austin 1998, pet. ref'd). Rodriguez’s immigration status clearly was not a material part of the plaintiffs’ case; it was not something the plaintiffs had to prove to prevail. See Bates v. State, 587 S.W.2d 121, 133 (Tex.Crim.App.1979) (stating that the “test as [242]*242to whether a matter is collateral is whether the cross-examining party would be entitled to prove it as a part of his case”). As a collateral matter — not relating to any of plaintiffs’ claims on the merits, and merely serving to contradict Rodriguez on facts irrelevant to. issues at trial — it was inadmissible impeachment evidence. See Ramirez v. State, 802 S.W.2d 674, 675 (Tex.Crim.App.1990) (stating parties may not impeach on collateral or immaterial matters); Delamora v. State, 128 S.W.3d 344, 363 (Tex.App.-Austin 2004, pet. ref'd) (noting “[a]party may not cross-examine a witness on a collateral matter, then contradict the witness’s answer”).
The immigration-related evidence was also inadmissible under Texas Rule of Evidence 608(b). This rule provides that “specific instances of the conduct of a witness, for the purpose of attacking ... the witness’s credibility, ... may not be inquired into on cross-examination of the witness nor proved by extrinsic evidence.” Tex.R. Evid. 608(b); see Tex.R. Evid. 404(b) (governing admissibility of prior acts). The rule “reflects a general aversion in Texas to the use of specific instances of conduct for impeachment.” David A. Schlueter & Robert R. Barton, Texas Rules of Evidence Manual § 608.02[3][b] at 537 (8th ed.2009). For over 150 years, “Texas civil courts have consistently rejected evidence of specific instances of conduct for impeachment purposes, no matter how probative of truthfulness.” Cathy Cochran, Texas Rules of Evidence Handbook 597 (7th ed.2007-08) (citing Boon v. Weathered’s Adm’r, 23 Tex. 675, 679 (1859) and other Texas cases). Courts in other jurisdictions have similarly held that a witness’s immigration status is not admissible to impugn the witness’s character for truthfulness.7
The only exception to this general prohibition is for certain criminal convictions. Texas Rule of Evidence 609 permits evidence of a criminal conviction for impeachment purposes if the conviction is not more than ten years old, is a felony or involves moral turpitude, and is more probative than prejudicial. Tex.R. Evid. 609(a). As the dissenting justice in the court of appeals observed, Rodriguez’s immigration conviction does not meet this criteria. 224 S.W.3d at 930 (Gardner, J. dissenting). It was therefore error to admit evidence of Rodriguez’s immigration status and deportation. The court of appeals nevertheless concluded that, even if it were error to admit this evidence, it was not harmful. 224 S.W.3d at 897.
D. The Harm
The erroneous admission of evidence is harmless unless the error probably caused the rendition of an improper judgment. TexR.App. P. 44.1. Probable error is not subject to precise measurement, but it is something less than certitude; it is a matter of judgment drawn from an evaluation of “the whole case from voir dire to closing argument, considering [243]*243the ‘state of the evidence, the strength and weakness of the ease, and the verdict.’ ” Reliance Steel & Aluminum Co. v. Sevcik, 267 S.W.3d 867, 871 (Tex.2008) (quoting Standard Fire Ins. Co. v. Reese, 584 S.W.2d 835, 841 (Tex.1979)).
Although the trial court initially granted a motion in limine on immigration matters, it later reversed that ruling, admitting extensive testimony and extrinsic evidence concerning Rodriguez’s immigration status, including that he:
• was an undocumented Mexican alien who had illegally entered the United States on multiple occasions;
• invented a false Social Security number, which he used to apply for a Texas commercial driver’s license;
• falsely answered “no” in his deposition when asked if he had ever lied to obtain a Texas driver’s license;
• falsely answered “yes” on his TXI employment application when asked if he had the legal right to work in the United States;
• pleaded guilty to and was convicted of a misdemeanor immigration violation, serving four months in jail; and
• was previously deported and ordered not to return to the United States for ten years.
Rodriguez was also Hughes’s first called witness, and the first questions posed to him concerned his immigration status. There followed over forty references to Rodriguez’s status, including thirty-five to his status as an “illegal immigrant” and seven to his prior deportation. TXI representatives were also cross-examined regarding whether they owed a “duty” to the public to prevent an “illegal” from driving a TXI truck:
• “Do you think he is entitled to drive here if he’s illegally here?”
• “And you don’t think you owe any duty ... to the public ... to the people who are driving up and down [Highway] 114 ... to decide whether he’s illegal or not?”
• “Mr. Rodriguez is still illegal in the United States, is he not? ... Will anybody ever turn him in, or will he just continue to drive for TXI?”
The investigating DPS trooper was asked whether she knew Rodriguez was “in this country illegally.” Additionally, there were thirty-two references to Rodriguez’s misconduct in using a “falsified” Social Security number, sixteen references to Rodriguez’s commercial driver’s license being “invalid” or “fraudulently obtained,” and seven references that Rodriguez was a “liar” who had lied on his TXI employment application. A TXI representative was pointedly questioned about whether Rodriguez might also have lied in denying responsibility for the accident:
• “Do you think Mr. Rodriguez lied to ... enter the United States?”
• “Are you telling this jury that you don’t know whether he lied to get into the United States?” “Now do you think that Mr. Rodriguez would lie when it relates to driving a rock truck?”
• “Did you ever consider ... and I want you to face this jury and tell this jury, did you ever consider whether Mr. Rodriguez might have lied about how this accident occurred?”
TXI complains that the repeated references to Rodriguez’s immigration problems and alleged misrepresentations were inflammatory and deliberately calculated to cause the jury to disbelieve Rodriguez.
TXI further objected to the trial court’s charge, complaining that the broad-form negligence question was misleading in this particular case and that the negligence question should instead include Hughes’s theory of the accident’s cause — that Rodri[244]*244guez caused the accident by first crossing over into the opposing lane of traffic. The trial court refused TXI’s requested substitutions, which TXI complains was harmful because it allowed Hughes to disguise his real claim — that Rodriguez was negligent for driving without a right to be in this country and that TXI was negligent for hiring an illegal alien. The dissenting justice in the court of appeals concluded that the Hughes’s “repeated injection into the case of Rodriguez’s nationality, ethnicity, and illegal-immigrant status, including his conviction and deportation, was plainly calculated to inflame the jury against him.” 224 S.W.3d at 931 (Gardner, J., dissenting). We agree.
Even assuming the immigration evidence had some relevance, its prejudicial potential substantially outweighed any probative value. Even in instances where immigration status may have limited probative value as to credibility, courts have held that such evidence is properly excluded for undue prejudice under Rule 403.8 The only context in which courts have widely accepted using such evidence for impeachment is in criminal trials, where a government witness’s immigration status may indicate bias, particularly where the witness traded testimony for sanctuary from deportation.9
IV. Conclusion
Hughes faced a difficult conceptual burden. He had to convince a jury that a collision involving on-coming traffic, that unquestionably occurred in the eastbound lane of Highway 114, was the fault of Rodriguez, the eastbound driver. The task was all the more difficult because Rodriguez possessed a clean driving record and commercial driver’s licenses from both Texas and Mexico. Hughes had some evidence of how Rodriguez might have been at fault for the collision in his lane, but the issue was hotly contested.
[245]*245The record indicates that Hughes sought to hedge his theory by calling attention to Rodriguez’s illegal immigration status whenever he could. Such appeals to racial and ethnic prejudices, whether “explicit and brazen” or “veiled and subtle,” cannot be tolerated because they undermine the very basis of our judicial process. Tex. Employers’ Ins. Ass’n v. Guerrero, 800 S.W.2d 859, 864 (Tex.App.-San Antonio 1990, writ denied); see also Moss v. Sanger, 75 Tex. 321, 12 S.W. 619, 620 (1889) (“Cases ought to be tried in a court of justice upon the facts proved; and whether a party be Jew or gentile, white or black, is a matter of indifference.”); Penate v. Berry, 348 S.W.2d 167, 168-69 (Tex.Civ.App.-El Paso 1961, writ ref'd n.r.e.) (reversing judgment against illegal alien in vehicle collision case because of “numerous remarks” about alien status). We conclude that the trial court erred by admitting evidence impugning Rodriguez’s character on the basis of his immigration status. Such error was harmful, not only because its prejudice far outweighed any probative value, but also because it fostered the impression that Rodriguez’s employer should be held liable because it hired an illegal immigrant.
For the reasons stated, the judgment of the court of appeals is reversed and the cause is remanded to the trial court for a new trial.
Justice WAINWRIGHT filed an opinion concurring in part and dissenting in part.
Justice JOHNSON did not participate in the decision.