OPINION
delivered the opinion of the Court,
in which MEYERS, PRICE, WOMACK, JOHNSON, and HOLCOMB, JJ., joined.
In this manslaughter/criminally negligent homicide trial, the visiting judge granted a defense-requested mistrial when the prosecutor asked the defendant’s expert, “Are you aware that her [the defendant’s] insurance carrier found her at fault?” The defendant then filed a habeas corpus application, claiming that a second trial was barred by federal and Texas constitutional double-jeopardy principles. The trial court denied relief. The court of appeals held that “the prosecutor intentionally or recklessly caused the trial to end in a mistrial,” so it reversed the trial court and dismissed the case with prejudice.1 We conclude that the court of appeals misapplied the standard of review by failing to assess the objective facts “in the light most favorable to the trial judge’s ruling.”2 Under that standard, the trial judge did not abuse her discretion in deny*320ing Ms. Wheeler’s double-jeopardy claim. We therefore reverse the judgment of the court of appeals and remand this case for further proceedings in the trial court.
I.
On the afternoon of July 21, 1999, Herman West was driving down a rural road in southwest Tarrant County when he saw Dr. David Mitchell standing near the rear of his parked truck, ready to walk across the road to his mailbox. A few seconds later, Mr. West saw a white Ford Mustang pass him in the opposite direction. Eighteen-year-old Kristin Wheeler was driving that Mustang. Mr. West heard the squeal of brakes, and, looking in his rear-view mirror, saw Dr. Mitchell flying through the air above the Mustang. He immediately turned his truck around and drove back to the accident scene. Dr. Mitchell later died from his injuries, and Ms. Wheeler was charged with manslaughter and criminally negligent homicide. The primary contested issue at trial was, ‘Who was at fault in causing this accident and Dr. Mitchell’s death?” The State’s theory was that Ms. Wheeler caused the accident because she was speeding and failed to exercise proper control of her car. The defense theory was that Dr. Mitchell failed to look before he walked into the road, and he failed to yield the right-of-way to the oncoming car. The trial was a battle between the State and defense accident-reconstruction experts.
After the State put on its case-in-chief, the defense called Alan B. Weckerling, an expert in accident-reconstruction analysis. His measurements, calculations, methodology, and conclusions differed significantly from those offered by Tim Lovett, a certi-fled peace officer and the State’s expert accident investigator who had testified extensively over four different days.3 Much of Mr. Weckerling’s opinion was based on a tape-recorded statement given by Mr. West a few days after the accident. This tape recording of Mr. West’s eyewitness account and his actions immediately after the accident was played to the jury. What the attorneys knew, but the jury apparently did not, was that the tape recording had been made by Ms. Wheeler’s insurance adjuster, who conducted his own investigation of the accident. Based on Mr. West’s recorded recollection of time, distance, and other factors, Mr. Weckerling concluded that it would not have mattered whether Ms. Wheeler was speeding-there was not time for her to avoid hitting Mr. Mitchell once he appeared out from behind his truck and started across the road in front of her. After Mr. Weckerling’s lengthy direct testimony over two days, the State cross-examined him for three hours.
The defense then conducted redirect and clarified Mr. Weckerling’s testimony in several respects. The State briefly recross-examined. Then, the defense, on re-redirect examination, asked a summing-up question: “Sir, based on everything that you have heard, all of the cross-examination, every exhibit you have looked at, going to the scene, reviewing every document, autopsy, this, that and the other, what caused the accident?” Mr. Wecker-ling responded: “The pedestrian walking in front of the Mustang.” The defense passed its witness for the third time. The record then shows the following:
Court: Anything else?
State: Yes, Your Honor—
*321Court: Thank you, sir. You may stand down.
State: I have one more question.
Court: I’m sorry I misunderstood you.
State: Are you aware that her insurance carrier found her at fault?
Defense: Your Honor, may we approach?
Court: You don’t have to approach. Send the jury out.
(Jury not present)
Court: Is there a motion in limine on that?
State: Only if she ever paid, Judge—
Defense: Your Honor they filed a motion in limine not to go into any of the insurance reports. They now have made a statement unsupported in bad faith to create a mistrial in this case.4
Court: Do you want a mistrial?
Defense: Yes, sir—
Defense:5 With prejudice, Your Honor.
Defense: With Prejudice.
Court: I’m going to take it under consideration. We will recess until Monday morning at 9:00 o’clock.
At the Monday conference, the visiting judge stated that he had researched the issues over the week-end. He listened as both the defense and State set out, at great length, their respective positions concerning the granting of a mistrial; he questioned the prosecutor concerning his legal theory for admission of the insurance investigator’s conclusion that Ms. Wheeler was at fault; and, finally, he asked: ‘Why did you say the very last question, Mr. [prosecutor]? You cross-examined him for three hours.” The prosecutor responded: “Judge, I’ve got to ask it sometime.” The judge then said,
I know. I’m just thinking I don’t think I need any more argument. It is a very troublesome ease. I made notes for myself. We [have] been at this a month, nine or ten days of actual testimony. This jury has been in and out of the courtroom. You all have fought over everything from pictures of a telephone pole to marking someone else’s exhibit....
There is one issue in this case and that’s who is at fault, The defendant as alleged or the deceased. Both sides knew as of Thursday[,] I told you all I was going to instruct the jury on concurrent cause.... So the only issue in this case was fault. This question goes right to that. It is not a collateral question, not a collateral issue. I am going to grant the motion for mistrial.
The defense later filed an application for a pretrial writ of habeas corpus and a plea of double jeopardy, which was heard by the presiding judge of the court, who stated that she had been in daily contact with the visiting judge during the trial, so she understood the issues. She explained:
I also have the benefit of a letter as well as legal case law that he left me when he left town stating the reasons for his rulings, and quite frankly, some opinions that he seemed to have on the issue before us today.
The defense offered various exhibits into evidence and then rested on its pleading. The State did not call any witnesses, but the trial judge questioned the prosecutor extensively about his reasons for asking the offending question, why he thought the question was a proper one, why he thought *322his question did not violate Rule 411 of the Texas Rules of Evidence,6 and why he thought that the defense had opened the door to the results of the insurance investigation. The trial judge then questioned the defense about whether it was clear that Rule 411 prohibits the type of question that the prosecutor asked, as there were no Texas criminal cases dealing with Rule 411. After both sides had fully explained their positions, the trial judge ended the hearing without ruling on the defense motion, saying that she wanted to see the transcript of Mr. Weckerling’s entire testimony first.
The trial court later denied the defendant’s double jeopardy motion and pretrial writ application, and the defense appealed. The court of appeals reversed the trial court and dismissed the charges.7 This Court then granted the State’s PDR8 and remanded the case for reconsideration under Ex parte Peterson,9 the newly announced clarification of the Texas double-jeopardy analysis set out in Bauder v. State.10 The court of appeals, using the Peterson analysis, once again concluded that the trial judge abused her discretion in denying Ms. Wheeler’s double-jeopardy claim based on manifestly improper prose-cutorial conduct which forced the defense to request a mistrial.11
II.
The double jeopardy provisions of the federal and Texas constitutions protect a citizen from repeated attempts at prosecution for the same criminal offense.12 However, if a defendant requests a mistrial, double jeopardy normally does not bar reprosecution.13 Under the federal double jeopardy clause, a retrial is prohibited after the defendant requests and is granted a mistrial only if the prosecution intentionally commits manifestly improper conduct with the intent to provoke that mistrial.14 In Bauder v. State, this Court held that, under the Texas Constitution, double jeopardy principles bar reprosecution when the prosecution acts, not only with the intent to goad the defendant into requesting a mistrial, but also when the prosecutor’s reckless misconduct requires a mistrial.15 Under either standard, the constitutional violation is that of depriving *323the defendant of his chosen jury, a jury presumably poised to acquit him.
In Ex parte Peterson, this Court explained that, under Kennedy,
the prosecutor’s goal is to terminate a first trial which is going badly. [Under Bander ], the prosecutor’s goal is to “win at any price,” either by mistrial and a subsequent retrial, or by using manifestly improper means to obtain a conviction in the first trial that he likely would not have achieved otherwise[,] and the prosecutor is aware that his conduct requires a mistrial if the defendant should request it.16
Peterson explained that Texas courts analyze a federal or state constitutional double-jeopardy claim under a three-part test:
1) Did manifestly improper prosecutorial misconduct provoke the mistrial?
2) Was the mistrial required because the prejudice produced from that misconduct could not be cured by an instruction to disregard? And
3) Did the prosecutor engage in that conduct with the intent to goad the defendant into requesting a mistrial (Kennedy standard) or with conscious disregard for a substantial risk that the trial court would be required to declare a mistrial (Bander standard)? 17
The distinction between the federal and state standards is that the former requires a finding that the prosecutor intended to “goad the defendant into requesting a mistrial” because the first trial was likely to end in an acquittal, whereas the latter is satisfied by a finding that the prosecutor acted recklessly and played “foul” to “win at any cost” because the first trial was likely to end in an acquittal were it tried fairly.18
Under either standard, the trial court must assess the prosecutor’s state of mind: Is it a culpable one-an intentional “deep-sixing” of the defendant’s chosen factfinder by an act of manifest impropriety or a reckless “win at any cost” act of manifest impropriety. Culpable intent or recklessness is not always easy to discern in this context. The adversarial “rough and tumble” of a hotly contested trial conducted by zealous advocates results in much unplanned, inadvertent, or impulsive rule-violating.19 In Peterson, we quoted Justice Powell’s concurring opinion in Kennedy, which stressed the importance of relying “ ‘primarily upon the objective facts and circumstances of the particular case’ in determining the prosecutor’s subjective intent, which may often be unknowable.”20 We therefore set out a list of non-exclusive objective factors to assist trial and reviewing courts in assessing the *324prosecutor’s state of mind. Those factors include:
1) Was the misconduct a reaction to abort a trial that was “going badly for the State?” In other words, at the time that the prosecutor acted, did it reasonably appear that the defendant would likely obtain an acquittal?
2) Was the misconduct repeated despite admonitions from the trial court?
3) Did the prosecutor provide a reasonable, “good faith” explanation for the conduct?
4) Was the conduct “clearly erroneous”?
5) Was there a legally or factually plausible basis for the conduct, despite its ultimate impropriety?
6) Were the prosecutor’s actions leading up to the mistrial consistent with inadvertence, lack of judgment, or negligence, or were they consistent with intentional or reckless misconduct?21
We, like the Supreme Court in Kennedy, are mindful that appellate judges “will not inexorably reach the same conclusion on a cold record at the appellate stage that they might if any one of them had been sitting as a trial judge[,]” thus “[i]t seems entirely reasonable to expect ... that appellate judges will continue to defer to the judgment of trial judges who are ‘on the scene’ in this area[.]”22 In Peterson, we stressed the importance of deferring to the trial court’s assessment of the facts, including the prosecutor’s state of mind. Here, as in other contexts, “appellate courts review the facts in the light most favorable to the trial judge’s ruling and should uphold it absent an abuse of discretion.”23
III.
In the present case, the court of appeals found that the prosecutor’s question — “Are you aware that her [the defendant’s] insurance carrier found her at fault?” — was manifestly improper.24 We agree. Whatever else it might be, this question was clearly barred by Rules 401-403 of the Texas Rules of Evidence. What Ms. Wheeler’s insurance carrier concluded has very little probative value on the question of whether she caused the fatal accident. There might be many reasons why the insurance carrier would not dispute liability or would ultimately concede it. This is akin to asking the testifying defendant if she was aware that her own attorney-her agent, just as an insurance carrier is a driver’s agent-admitted that she should accept a plea bargain. Lawyers impeach witnesses with specific facts, not with someone else’s legal concession or ultimate factual decision.
The prosecutor could have called the insurance investigator as a witness25 and asked him about his investigation-what measurements did he take, to whom did he speak, what calculations did he perform, *325what opinions, if any, did he reach about the cause of the accident. But the insurance company’s ultimate act of conceding liability is barely (if at all) probative of appellant’s “fault” in causing the accident. And it is extremely prejudicial, likely to confuse and mislead the jury, and lead to time-consuming collateral litigation. This is precisely the type of evidence that Tex.R. Evid. 403 is designed to exclude.
The court of appeals also concluded that the visiting judge granted the mistrial because the prejudice produced from the prosecutor’s improper question could not be cured by an instruction to disregard.26 Again, we agree. This was, put charitably, a hair-raising question, and the visiting judge immediately recognized it as such. The defense did not even request a mistrial before the judge asked whether the defense wanted one. The judge reasonably concluded that once this skunk was in the jury box, there was no way to hide the smell. No instruction to disregard would still the reverberations (“even her own insurance carrier thought she was at fault; it wouldn’t stick up for her”). This was a statement of fact, with a question mark appended at the end, which went to the very heart of the single, hotly disputed, issue in this trial. The court of appeals correctly concluded that the second prong of the Peterson test was met.27
We part company with the court of appeals’s analysis, however, on the third prong-whether the prosecutor asked this question with the intent to goad the defense into requesting a mistrial (Kennedy) or with conscious disregard of a substantial risk that the trial court would be required to grant a mistrial (Bauder ).28 On this third prong, the court of appeals failed to view the objective facts in the light most favorable to the trial judge’s ruling. The court of appeals explicitly stated that it was employing a de novo review of the facts, rather than a deferential one, because the trial judge who conducted the habeas hearing was not the same judge who presided over the trial and granted the mistrial.29
First, as a matter of law, reviewing courts defer to the trial court’s implied *326factual findings that are supported by the record, even when no witnesses testify and all of the evidence is submitted in written affidavits.30
Second, in this case, the trial judge personally presided over the habeas hearing and was well-positioned to make credibility decisions. Although she did not preside over the underlying trial, at the writ hearing she stated that she had been in “daily” communication with the visiting judge. “I mean, I can’t even begin to tell you how many times I talked to him during /all’s trial.... So I feel very comfortable that I understand the issues.” She also said that the visiting judge left her a letter “stating the reasons for his rulings” and “some opinions that he seemed to have on the issue” at the writ hearing. Further, the trial prosecutor stood in front of her, and, while he did not formally testify, she quizzed him extensively concerning his reasons for asking the offending question. She could gauge his credibility and demeanor during this face-to-face colloquy.31 She may have had prior knowledge of this prosecutor, and her assessment of his credibility could well be based upon those prior experiences as well as the cogency and genuineness of his explanation.
Third, the record shows that the trial judge did not make a hasty or ill-informed ruling. At the writ hearing, she had the transcripts of both the pertinent excerpts from the trial and the Monday mistrial hearing. She stated that she had also done some research on the double jeopardy issue, but she wanted to defer any ruling until she reviewed a transcript of the entirety of the defense expert’s testimony, as well as any other testimony that either side thought important. This was a trial judge who was exercising her discretion carefully and cautiously, not one who was making an “off-the-cuff’ ruling. Thus, she was the “Johnny-on-the-Spot” factfin-der at the habeas hearing; this Court must view the objective facts in the light most favorable to her ruling32 and uphold that ruling absent an abuse of discretion.
Here, the objective facts do not show that the trial judge abused her discretion in concluding that the prosecutor honestly (though misguidedly) thought that his question was permissible. We assess the *327objective facts in light of the six factors set out in Peterson.
1. Was the trial going badly for the State?
Both the State and the defense sponsored well-qualified, articulate expert witnesses. They came to opposing conclusions on the hotly disputed issue of who caused the accident: Ms. Wheeler or Dr. Mitchell. At the time of the mistrial, it was an evidentiary “neck-and-neck” horse race. However, at the mistrial motion hearing, the prosecutor stated that the insurance investigators were under subpoena and “we can bring them down here.” Indeed it could and should have done so.33 Then the State could have pointed to an independent expert-working for neither the State nor the defense attorney-who had reached the same factual conclusion (assuming that he did) as that reached by the State’s expert.34 This prosecutor was in an enviable evidentiary position. In fact, the visiting judge asked the prosecutor why he did not call the adjuster to testify to his investigation during the mistrial hearing, and the prosecutor responded, “That’s one possible way.” But then he argued, “As far as how I can prove it, I can prove it through asking questions of Mr. Westerlink [sic].”35 On that score he was resoundingly wrong, but there is no indication that this was a trial that the prosecution was losing.36
Ms. Wheeler argues that the prosecutor “clearly saved the [offending] question for the very last instant. This was the last question, for the last witness, for the last day of trial. The timing of this question is proof that the prosecutor intentionally *328saved this ‘zinger’ for last.” Well, it was the last question of the trial since the trial ended because of this question, but no one knows what might have occurred had this question not been asked. Perhaps the State would have called the insurance adjuster to testify to his investigation. Furthermore, the record does not support any suggestion that the State “saved” this “zinger” question. It had ended its initial cross-examination of Mr. Weckerling without asking this question, and it could not predict whether the defense would re-direct. It had ended its re-cross-examination of Mr. Weckerling without asking this question, and, again, it could not predict whether the defense would re-re-direct. It was only when the defense asked its summing-up question, including a reference to “all” of the materials that Mr. Weckerling had reviewed, that the State, on its third cross-examination, asked its fatal question. In context, the prosecutor appears more rash than Machiavellian.
2. Did the State repeat its misconduct despite admonitions from the Court?
This factor focuses on the wilfulness of repeated misconduct by the prosecutor. There was no repeated misconduct. Ms. Wheeler argues that the prosecutor violated its own motion in limine by asking this question and that conduct shows wilfulness. But the State’s motion in limine barred only the mention of a “monetary settlement,” not any mention of an insurance carrier’s investigation. The defense had filed a motion in limine barring “[a]ny statements of any witness that the defendant is guilty of any offense being tried,” and that motion, though close, is not a clear reference to the insurance carrier’s finding of “fault.” Furthermore, a single violation of a motion in limine is insufficient to show repeated, wilful violations made despite judicial admonitions.
3. Did the prosecutor provide a reasonable “good faith” explanation for the conduct?
At the mistrial hearing, the prosecutor offered two explanations for his question: First, he thought (incorrectly) that Mr. Weckerling “had seen the materials from an insurance investigation,” thus “this issue is already before the jury as to an insurance investigation.” Second, he noted that
the last question by defense counsel pri- or to passing a witness for recross [was] in reference to causation and fault of the victim. I certainly think it is fair impeachment of any expert who comes in here and gives an opinion as to, one, things they have reviewed, and, two things they have reviewed that are contrary. ... First of all, I did ask him did he review material from insurance investigations. He said he looked at it; therefore, anything he’s looked at is open for me to impeach him or cross-exam him on. The rules specifically allow for that.
There is no indication anywhere in the record that the prosecutor did not sincerely, but mistakenly, believe that Mr. Weck-erling had, in fact, reviewed the insurance report.37 The prosecutor repeatedly explained, both at the mistrial hearing and at the writ hearing, that he based his ques*329tion upon this fact. It turns out that this “fact” is wrong. The prosecutor may have misheard the witness or confused the testimony of his own expert (who had reviewed the insurance investigation report) with that of the defense, but that does not make him either wilful or reckless. Of course, even if Mr. Weckerling had reviewed the entire file, the prosecutor could not ask the precise question that he did ask, but he could have asked about specific facts and data, calculations, and whether and why the defense expert thought the other investigator’s opinion about the cause of the accident was inaccurate.
The court of appeals stated that, toward the end of the mistrial hearing, the visiting judge asked the prosecutor why he had asked that particular question, and the prosecutor responded, “Judge, I got to ask it sometime.”38 Only those present in the courtroom at the time could determine whether this statement was made in a flippant or earnest manner, but, on its face, the statement suggests that the prosecutor thought his was a proper question. At any rate, this ambiguous response came long after the prosecutor had explained his reasons for the question.
4. Was the conduct “clearly erroneous”?
The prosecutor’s question brought an immediate halt to the proceedings. There seems no doubt that the visiting judge instantly thought it was a “clearly erroneous” question that could not be cured by an instruction to disregard. There seems little doubt that the trial judge also thought this question was “clearly erroneous,” and during the writ hearing she was simply seeking an explanation from the prosecutor as to his reasoning for asking the question. Although the prosecutor, defense attorney, and trial judge all focused on whether the question was prohibited or allowable under Rule 411 of the Texas Rules of Evidence, that rule does not really speak to the propriety of the question.
Rule 411 prohibits the parties from inquiring about the existence of insurance offered to show that the insured was negligent or otherwise at fault.39 The prohibited inferences are that (1) one who has insurance is more likely than others to be negligent because he is protected by insurance; and (2) an insurer has “deep pockets” and thus could afford to pay any judgment against its insured.40 The prosecutor’s question did not go to the existence of insurance (a fact that had been mentioned several times during this trial), but rather to the insurance carrier’s apparent ultimate decision to accept fault by its insured, Ms. Wheeler. Thus, the prosecutor was technically correct that Rule 411 would not directly bar his questioning, but his question was far more prejudicial (and more misleading) than any mere mention of the existence of insurance as proof of Ms. Wheeler’s negligence or fault. The question stated the result of an insurance investigation and the conclusion that Ms. Wheeler’s own agent, her insurance carrier, had conceded her “fault.” Thus, the question was outside Rule 411, but even *330more prejudicial than the questioning that is prohibited by that rule.
5. Was there a plausible basis for the prosecutor’s conduct?
There was, as discussed above, a very reasonable basis for asking Mr. Wecker-ling whether he had reviewed any expert investigations or reports other than the one prepared by the police. There was a plausible basis for cross-examining him on the importance of reviewing all materials and sources connected to an accident investigation before reaching an opinion. This was a fruitful area of impeachment, and the State could have then called the “other” investigator during rebuttal to discuss his underlying data and opinion on the cause of the accident. Thus, the area of inquiry was entirely appropriate, but the specific question actually asked by the prosecutor was entirely inappropriate.
6. Were the prosecutor’s actions leading up to the mistrial consistent with inadvertence, lack of judgment or negligence, or were they consistent with intentional or reckless misconduct?
This was a hotly contested trial and, given its duration and various contretemps between the advocates, tempers were hot on both sides. As the visiting judge plaintively concluded, both sides acted “play-groundish.” The objective record bears out this assessment. Both the State and defense were very zealous in putting forward their respective factual and legal positions. The advocates clashed repeatedly. The visiting judge assessed no intentional or reckless fault against either side, but he might well have taken into account the rising tempers and trial fatigue by both sides as contributing to the prosecutor’s final, fatal question.
IV.
In sum, we agree with the court of appeals that the visiting judge did not abuse his discretion in granting the mistrial when the prosecutor asked a manifestly improper question.41 But, viewing the objective facts in the light most favorable to the trial judge’s ruling, we cannot say that the trial judge abused her discretion in denying relief on Ms. Wheeler’s double-jeopardy claim. Factually, this is a close case, and, had the trial judge ruled that double jeopardy did bar any retrial, we would, of course, uphold that ruling as well. Rational people can differ about whether, given these facts, it is a reasonable inference that the prosecutor was intentionally goading the defense into asking for a mistrial (the federal Kennedy42 standard) or acting with conscious disregard of a substantial risk that the trial court would be required to declare a mistrial (the state Peterson/Bauder43 standard). That is precisely why we, as a reviewing court, must defer to the trial judge’s factfinding.
From all appearances, the prosecutor could well have thought that the trial was going fine. It is reasonable to conclude that he was, in his own mind, just giving the coup de grace to the defense expert. We cannot disagree with the trial judge’s implicit conclusion that this was a question that was asked in good faith, albeit an impetuous, perhaps even stupid, question. *331The trial judge saw the prosecutor and could judge his credibility and integrity; she was entitled to conclude that the prosecutor acted with unwarranted zeal rather than malice or reckless disregard for the defendant’s rights. We cannot say that the trial judge abused her discretion in finding that neither Kennedy nor Peterson would bar the defendant’s retrial.
We therefore reverse the judgment of the court of appeals and remand this case to the trial court for further proceedings.
KELLER, P.J., filed a concurring opinion.
KEASLER, J., filed a concurring opinion, in which KELLER, P. J., and HERYEY, J., joined.