Defendant Bernard Lipka appeals from a judgment of conviction, based on a jury verdict, of two counts of sexual assault of a minor in violation of 13 V.S.A. § 3252(a)(3), and one count of lewd and *379lascivious conduct with a child in violation of 13 V.S.A. § 2602. Defendant contends the trial court committed reversible error by: (1) granting a special seating arrangement for a child witness in violation of defendant’s Sixth Amendment right to confrontation; (2) admitting evidence of prior sexual misconduct; and (3) refusing to admit the videotape of defendant’s police interview. We agree with defendant’s first claim of error and find that the error was not harmless. We also agree that the evidence of prior sexual misconduct was not admissible on the grounds asserted by the State. Accordingly, we reverse and remand.
The sequence of events in the investigation and trial was as follows: Defendant and J.C. became lovers in the early fall of 1997. J.C. lived in an apartment in Burlington with her twin daughters, M.C. and K.C., who were seven years old at the time. When an apartment below J.C.’s became vacant later in the fall, defendant rented it. The girls were' sometimes permitted to visit defendant in his apartment, and defendant would occasionally babysit for them.
On Halloween night 1997, J.C. refused to allow M.C. to go trick-or-treating because she had gone to a friend’s house after school without permission. Defendant offered to watch M.C. while J.C. took her other daughter trick-or-treating. J.C. testified that the next day, while in defendant’s car, she “heard the girls talking... about a shower.” When questioned, M.C. told her mother that she had taken a shower with defendant. M.C. said there had been no inappropriate touching — she washed defendant’s back and he washed hers. J.C. asked defendant about the incident that evening. According to J.C., defendant claimed that he and M.C. may have passed each other while he was getting into the shower and M.C. was getting out, or vice versa, but that they had not actually showered together.
Following this conversation, J.C. did not permit defendant to be alone with her children. She stopped seeing defendant romantically several days after New Year’s Eve 1998. The following July, J.C. heard one of her daughters make a comment about “sucking on private parts.” Describing herself as “stunned,” she talked with M.C. alone. As J.C. recalled, “I let her know that she wasn’t in trouble____And she reluctantly told me that they were talking about sucking on [defendant’s] private parts. That this was something that had happened.”
A few days later, according to J.C., M.C. further revealed that sexual contact had occurred between her and defendant on a number of occasions, and that she and defendant “had taken pictures of their private parts.” M.C. told her mother that these events had occurred “when she was supposed to be at the Boys & Girls Club.” J.C. called the police to *380report the allegations of sexual misconduct. As a result, M.C. was interviewed at an office of the Department of Social and Rehabilitation Services (SRS) by an SRS worker and a policeman, at which time she described her sexual activity with defendant.1
M.C. stated in the interview that between the time defendant moved into his apartment and July of 1998, she sucked on defendant’s penis five or six times, he placed his fingers in her vagina many times and she rubbed his penis with her hands more than ten times, possibly as many as seventy times. M.C. described an incident in which she and defendant showered together: First, she played with defendant’s penis outside the shower. Then, she entered the shower while defendant was in it waiting for her. They soaped each other, and she again played with his penis. M.C. also recalled one occasion on which she and defendant took nude photographs of each other in defendant’s laundry room. She stated that the photograph she took of defendant did not come out. M.C. also stated that she and defendant had touched each other’s private parts while they were lying on a couch in her apartment with a blanket covering them, while her mother and sister were in another room.
Defendant was later questioned by three police detectives, including the officer who interviewed M.C. They talked for approximately one hour in the detectives’ vehicle. The detectives recorded the conversation without defendant’s knowledge. The officer who interviewed M.C. testified that defendant denied having oral sex with M.C., but admitted that it was possible that M.C.’s hand had accidentally touched his penis while she was getting into his shower and he was getting out. Defendant also told the detectives that he had gone to the store with M.C. to buy her candy “before we took the shower.” Defendant described another incident in which he was playing with the girls when they allegedly pulled down his pants and M.C. reached up and grabbed his penis.2 Defendant denied that he and M.C. had taken photographs of each other. A search warrant executed that same day resulted in the discovery of an under-exposed Polaroid photograph depicting a man wearing camouflaged pants, such as those habitually worn by defendant, displaying his erect penis.3
The State charged defendant with three counts of sexual misconduct, all involving M.C. and all involving conduct occurring between September 1997 and June 1998. The first two counts charged defendant with child *381sexual abuse, a crime for which the maximum punishment is life in prison, for (1) inserting his fingers in the vagina of M.C., and (2) bringing his penis in contact with the mouth of M.C. The third count charged him with lewd conduct with a child by having contact between the hand of M.C. and the penis of defendant. Although this charge also involved a felony, the maximum punishment was only five years in prison, far less than that of the first two charges.
The State moved to introduce evidence of uncharged sexual conduct with M.C. over the expanded time frame. The trial court granted the State’s motion, finding that the prior bad act evidence was “admissible to show the ‘situational context’ for the abuse of M.C. by her mother’s boyfriend.”
The State also moved to admit evidence that defendant had similar sexual contact with his six-year-old daughter, R.L., and a third victim, A.G., the seven-year-old daughter of another girlfriend.4 The court granted the motion in a written order, finding that the evidence of defendant’s prior bad acts was “clearly relevant to the State’s theory of the case, not only to show absence of mistake, intent and motive, but also to demonstrate opportunity to commit the offenses.” The court noted that when interviewed by the police, defendant “suggested that there might have been accidental sexual contact between [him] and M.C. during a shower incident.” Hence, the court found that the prior bad act evidence was “directly related to a disputed issue in the case” and “clearly bears on Defendant’s claims of accident or mistake.” Additionally, the court found that “[t]he repetitive series of sexual abuse of young girls in the same or similar settings shows a common scheme or plan by Defendant and therefore is highly probative as to his intent to commit the acts alleged.”
The evidence consisted of the testimony of six persons: the officer who interviewed M.C., J.C., defendant, R.L., R.L.’s mother (defendant’s ex-wife) and defendant’s father. M.C. testified by video deposition, and a videotape of her statement to the police officer and the SRS worker was also admitted. The remaining exhibits consisted of the Polaroid camera, a film box, the Polaroid photo described above, camouflaged pants, and the search warrant under which the pants were obtained. M.C.’s description of the incidents alleged in the charge was graphic and largely consistent with her earlier statement.
Defendant testified at trial. He denied all of the sexual acts alleged by M.C. He did not claim accident or lack of intent as a defense. He did admit *382the contact as he left the shower on Halloween, but his description of the shower incident was not remotely similar to the incident described by M.C. Neither in his opening nor in his closing did defendant’s counsel argue that defendant committed any of the acts alleged but did not have the criminal intent. Nothing in his examination or cross-examination of witnesses supported such a claim.
Over objection, R.L. testified at trial. She described an incident that occurred in her paternal grandparents’ home about eighteen months before the charged offenses in which defendant touched her vagina while they were watching television with R.L.’s older brother. The court later instructed the jury that R.L.’s testimony, if believed, was to be considered “for a limited purpose only... that defendant is capable of taking the chance of having a lewd contact with a child in the presence of another person... [or] in determining whether any contact between the Defendant’s penis and [M.C.’s] hand was accidental or intentional on Defendant’s part.” In connection with R.L.’s testimony, her mother, defendant’s former wife, testified that she learned about the incident when R.L. disclosed it at the day care center, that the incident brought her in contact with the police, and that, since the incident, defendant has had no visitation with his daughter. Also, defendant’s father testified that he was in the room when the alleged touching occurred and did not see it happen.
The jury returned a guilty verdict on all counts. Defendant’s motion for new trial was denied. This appeal followed.
I.
A.
Defendant first contends that his Sixth Amendment right to confrontation was violated by a special seating arrangement at trial in which R.L. was seated facing away from defendant during her testimony. We discuss this issue briefly because the State has essentially conceded it, briefing and arguing instead that the error was harmless.
The trial court granted the State’s pretrial motion for the special seating arrangement, noting that R.L. was nine years old, and that she had not seen defendant since disclosing that he had inappropriately touched her two years earlier. As described by defense counsel in noting his objection, R.L. sat at a table placed in front of the jury with her back to defendant. The prosecutor sat at the table facing R.L. during questioning. Defense counsel was offered the same opportunity to sit facing R.L., but chose not to cross-examine her.
Prior to R.L.’s testimony, the court instructed the jury that the special seating arrangement had been made “because it’s a child and we feel the *383child would be more at ease if she’s sitting [in] this kind of configuration rather than sitting on the witness stand.”
The United States Supreme Court has expressly held that the Confrontation Clause “guarantees the defendant a face-to-face meeting with witnesses appearing before the trier of fact,” Coy v. Iowa, 487 U.S. 1012, 1016 (1988), but has also recognized that the right is “not absolute” and “ “must occasionally give way to considerations of public policy and the necessities of the case.’ ” Maryland v. Craig, 497 U.S. 836, 849 (1990) (quoting Mattox v. United States, 156 U.S. 237, 243 (1895)). Accordingly, the Court in Craig upheld Maryland’s statutory procedure for receiving the testimony of a child witness in a sexual molestation case by closed circuit television, a procedure similar to that used to admit M.C.’s testimony in this case under V.R.E. 807. The Supreme Court noted that the procedure otherwise
preserved] all of the other elements of the confrontation right: The child witness must be competent to testify and must testify under oath; the defendant retains full opportunity for contemporaneous cross-examination; and the judge, jury, and defendant are able to view (albeit by video monitor) the demeanor (and body) of the witness as he or she testifies.
Id. at 851. In so holding, the Court stressed the necessity of “case-specific” findings that the special procedure is necessary to protect the welfare of the particular child witness. Id. at 855. Without purporting to define the minimum showing of emotional distress required for such a procedure, the Court nevertheless required a determination by the trial court that the child witness would be traumatized by the presence of the defendant. Id. at 856.
We agree with defendant that his right to confrontation as defined by the Supreme Court in Craig was violated by the seating arrangement that prevented defendant from seeing R.L. while she testified, as well as by the trial court’s failure to make adequate findings concerning the necessity of such an arrangement. Although Craig approved the use of closed circuit television to prevent a child witness from seeing the defendant, it required — at a minimum — that defendant retain the opportunity “to view... the demeanor (and body) of the witness as he or she testifies.” Id. at 851; see also People v. Lofton, 740 N.E.2d 782, 794 (Ill. 2000) (defendant’s confrontation clause rights violated by barricade erected by trial court that blocked child witness from defendant’s view). The seating arrangement devised by the trial court in this case indisputably deprived defendant of the opportunity to observe the witness’s demeanor during' *384her testimony, contrary to this requirement.5 Furthermore, although the State argued that the seating arrangement was necessary because R.L. had not seen defendant since her allegations of misconduct several years earlier, there was no specific evidence — or finding — that the seating arrangement was necessary to prevent emotional distress or trauma to the child witness. The court simply ruled that it would “permit the State to proceed along those lines.” This was plainly inadequate under Craig, which requires specific and particularized findings of necessity. 497 U.S. at 855.
B.
The State argues, however, that the error was harmless in view of the strength of the State’s case and the limited weight of R.L.’s testimony. We can uphold a criminal conviction, despite a confrontation clause error, if we find that the error was harmless beyond a reasonable doubt. See Coy, 487 U.S. at 1021; State v. Lynds, 158 Vt. 37, 42, 605 A.2d 501, 503 (1991). The way the standard is to be applied is set out in Coy, the leading decision of the United States Supreme Court:
An assessment of harmlessness cannot include consideration of whether the witness’ testimony would have been unchanged, or the jury’s assessment unaltered, had there been confrontation; such an inquiry would obviously involve pure speculation, and harmlessness must therefore be determined on the basis of the remaining evidence.
Coy, 487 U.S. at 1021-22. As we said in Lynds, “[t]o determine whether the error is harmless, we must posit a trial without any evidence by [the witness who testified in violation of defendant’s confrontation rights].” Lynds, 158 Vt. at 42, 605 A.2d at 503.
Thus, the question before us is whether we can say beyond a reasonable doubt that the jury would have convicted defendant if R.L. had never testified. In considering this question, we must also ignore the testimony of defendant’s father and the testimony of defendant’s ex-wife that R.L.’s accusations brought her in contact with the police and resulted in defendant never seeing R.L. again before the trial. We conclude that we cannot *385say beyond a reasonable doubt that this evidence had no effect on the jury’s verdict.
The two most important factors in the harm equation we must employ are the strength of the prosecution’s case without the offending evidence and the strength of the offending evidence. In approaching the first of these factors, it is important to understand that it is not the role of this Court to determine whether defendant is guilty. See Kotteakos v. United States, 328 U.S. 750, 763 (1946). The main danger in our analysis of the strength of the evidence in determining whether an error is harmless was described by Chief Judge Harry Edwards of the United States Court of Appeals for the District of Columbia Circuit:
The problem with harmless error arises when we as appellate judges conflate the harmlessness inquiry with our own assessment of a defendant’s guilt. This approach is dangerously seductive, for our natural inclination is to view an error as harmless whenever a defendant’s conviction appears well justified by the record evidence. However, the seductiveness of this approach is its chief defect, for, drawn in by its attractions, we have applied the harmless-error rule to such an extent that it is my impression that my colleagues and I are inclined to invoke it almost automatically where the proof of a defendant’s guilt seems strong.
H. Edwards, To Err is Human, But Not Always Harmless: When Should Legal Error Be Tolerated?, 70 N.Y.U. L. Rev. 1167,1170 (1995). The issue before us is what the jury might have done without the offending evidence, not what we would do if we were the factfinder.
At base, this was a classic swearing contest. Both M.C. and defendant testified. M.C. made detailed accusations. Defendant denied each of these detailed accusations. There were discrepancies in the testimony of both, and cross-examination pointed out these discrepancies. There was no corroboration for M.C.’s accusations. Although she testified to many incidents of sexual abuse — up to 100 or more, over an eight month period — no one observed any of the incidents or even her presence in defendant’s apartment that enabled the incidents. Apparently, because she had not observed any conduct incident to the alleged abuse, M.C.’s twin sister did not testify.
The only event that emerged before M.C.’s detailed disclosures was the “shower incident.” Both M.C. and defendant described a shower incident, but these descriptions were entirely different. M.C. testified that she and defendant showered together with sexual activity occurring before and *386during the shower. Defendant testified that he showered alone, and that M.C. tried to enter the shower as he exited. In his description, there was no joint shower or sexual activity; any touching was brief and accidental and occurred outside the shower as he and M.C. passed each other.
Two pieces of evidence may have helped the State to support M.C.’s story, but we cannot say that either was so strong as to determine our conclusion. M.C. testified that she and defendant photographed each other while they were nude, and the State offered a poor-quality Polaroid photograph of defendant dressed in camouflaged pants with his penis exposed. Defendant denied the photography incident and stated the picture had been taken by M.C.’s mother, J.C. Defense counsel argued that the photograph could not support M.C.’s testimony because defendant was not nude in the photograph..
Perhaps the State’s most probative supporting evidence was defendant’s answer to a question during the police interrogation that he took M.C. to a store to buy candy “before we took the shower.” This answer occurred, however, during the questioning in which defendant consistently denied that he and M.C. had showered together but agreed that each took a shower. Thus, the State claims, and the defendant denies, that defendant admitted that he and M.C. showered together despite his continuous denial of that accusation.
Although these two pieces of evidence favored the State and the State particularly focused on them during the trial, we cannot determine how the jury viewed them. For each, defendant had a different explanation, one inconsistent with his guilt.
We also note a consistency difficulty with the State’s argument that its ease was so strong that the erroneous admission of R.L.’s testimony was harmless beyond a reasonable doubt. As we discuss in detail infra, R.L.’s testimony was prior bad act evidence admitted under V.R.E. 404(b) and found to pass the balancing test of V.R.E. 403. To demonstrate conformance with these rules, the State had to show a need for the prior bad act evidence. See State v. Winter, 162 Vt. 388, 400, 648 A.2d 624, 631 (1994) (State’s need for evidence of prior bad acts is major factor in balancing process); State v. Catsam, 148 Vt. 366, 382-83, 534 A.2d 184, 195 (1987) (prosecution’s duty is to demonstrate a need to introduce evidence). It did so, in part, by arguing that the prior bad act evidence was “highly probative” because there was no direct evidence of the sexual abuse other than the victim’s testimony and this Court had labeled such testimony from a victim as not “entirely reliable.” See State v. Johnson, 158 Vt. 344, 353, 612 A.2d 1114, 1119 (1992) (“evidence of related sexual acts against victims whose testimony is not considered entirely reliable has substantial *387probative value”); see also State v. Sullivan, 576 N.W.2d 30, 41 (Wis. 1998) (admission of prior bad act evidence not harmless where state conceded need for evidence in arguing for its admissibility). Thus, the evidence that the State labeled as not entirely reliable to persuade the court to admit R.L.’s testimony is now called overwhelming to persuade us that the admission of R.L.’s testimony was harmless.
We have one particular precedent, State v. Lynds, that is helpful in examining this first factor, the strength of the properly-admitted evidence in support of the conviction. In Lynds, we reversed a conviction because the trial court violated defendant’s confrontation clause rights by allowing admission of a deposition of an expert witness who testified generally to issues of delayed reporting of child sexual abuse and family dynamics, patterns and effects of such abuse. Lynds, 158 Vt. at 42, 605 A.2d at 504. The witness had never met the victim. The State’s evidence consisted of the testimony of the victim, who described in detail regular sexual abuse by defendant over a four-year period, and the deposition of the expert witness. The defendant testified that the alleged conduct had not occurred, and his testimony was supported by three of the victim’s brothers, who testified that they had not observed the abuse although they lived in the same house as defendant and the vietim and the house lacked privacy. Observing that “[t]he trial was a credibility contest between the defendant and the victim with the defendant having the advantages of the presumption of innocence, the State’s high burden of proof and the supporting testimony,” we concluded that the error was not harmless. Id. at 43-44, 605 A.2d at 503.
The only differences between this case and Lynds are the nature of the erroneously-admitted evidence and the fact that supporting witnesses, to the extent they existed, were part of the State’s case here and the defendant’s case in Lynds. Given that the only direct evidence of the sexual abuse came from M.C., we do not believe that the alignment of the supporting witnesses can be determinative under the high standard for harmless error. Consistent with Lynds, we must conclude that the first harmless error factor does not support a conclusion that the error in admitting R.L.’s testimony was harmless.
The second factor — the effect of the erroneously admitted evidence — weighs heavier against finding harmless error. If we could view this evidence only with respect to the logical relevance that supported its admission, we could agree with the State that it had very limited probative value. As we discuss below in regard to the admission of this evidence, however, this was evidence of prior child sexual abuse with a high danger *388of prejudicial impact on the deliberations of the jury. We said of such evidence in State v. Catsam, 148 Vt. at 383, 534 A.2d at 195: “One can think of no potentially more inflammatory evidence than similar prior sexual contacts.” We amplified this statement in State v. Forbes, 161 Vt. 327, 330, 640 A.2d 13, 15 (1993):
We must be vigilant in reviewing the admission of evidence of uncharged misconduct, because once jurors learn of uncharged misconduct, they tend to use an entirely “different... calculus of probabilities” in deciding whether to convict. See Imwinkelried, Uncharged Misconduct, 1986 A.B.A. Sec. Crim. J. 6, 8 (Summer) (quoting H. Kalven & H. Ziesel, The American Jury 179 (1966)). In fact, several empirical studies tend to confirm prosecutors’ beliefs that the introduction of a defendant’s uncharged misconduct can “easily tip the balance against the defendant.” Imwinkelried at 8 (referring to studies conducted by the Chicago Jury Project, the London School of Economics, and the National Science Foundation Law and Social Science Project); see State v. McCarthy, 156 Vt. 148, 155-58, 589 A.2d 869, 873-75 (1991) (misconduct evidence can have an “incendiary” impact on the jury).
The evidence in this case was particularly explosive and prejudicial because the victim of defendant’s prior sexual misconduct was his own daughter. In State v. McCarthy, 156 Vt. at 155, 589 A.2d at 873, we labeled evidence of uncharged misconduct as the “ ‘most prejudicial evidence imaginable.’ ” (quoting People v. Smallwood, 722 P.2d 197, 205 (Cal. 1986)). It is particularly telling that this Court has never before found that improper admission of prior bad act evidence was harmless beyond a reasonable doubt.6
We find our decision in McCarthy particularly relevant in our analysis of this second harmless error factor. In McCarthy, the defendant was charged with lewd and lascivious conduct with his daughter, and the jury found him guilty. During the presentation of evidence, the prosecutor showed that in the past the defendant had been investigated by the SRS for having oral sex with his son. Defense counsel failed to object to this *389evidence, which came in through numerous witnesses, and the question on appeal was whether the admission of the prior bad act evidence was plain error. We held that it was because of the highly prejudicial effect of prior sexual misconduct evidence and the fact that “[t]he case could have gone either way, depending upon whether the jury believed complainant beyond a reasonable doubt or believed defendant, who testified, or could not resolve the conflict in the testimony to convict.” McCarthy, 156 Vt. at 157, 589 A.2d at 875. We summarized our holding as follows:
In summary, the court in this case erroneously admitted evidence of another alleged sexual assault on another of defendant’s children, and the prosecution exploited the evidence unfairly to suggest that defendant generally committed acts of sexual misconduct on his children. In a case that depended ultimately on the credibility of defendant and the complainant, such evidence was incendiary. Because there is a substantial likelihood that defendant was convicted by this incendiary device and not by properly admitted evidence, we reverse.
Id. at 158, 589 A.2d at 875 (footnotes omitted).
It is very rare that we find plain error, and the burden to show it is extremely high. See id. at 154, 589 A.2d at 873. Although the prosecutor exploited the improper evidence to a much greater degree in McCarthy than the prosecutor did here, the nature of the evidence and the case are essentially the same. We cannot justify finding plain error because of this evidence in one case and in a very similar case find any error is harmless.
For the above reasons, we conclude that the two most significant factors in a harmless error analysis weigh heavily against a conclusion that the admission of R.L.’s testimony in this case was harmless error. We cannot conclude beyond a reasonable doubt that the jury would still have convicted defendant if it had never heard the testimony of R.L., her mother and grandfather. We must, therefore, reverse defendant’s convictions and give him a new trial because of the confrontation clause violation in the seating of R.L.
II.
Because the issues are likely to recur on retrial, we address defendant’s remaining issues on appeal. We start with defendant’s argument that the trial court erred in admitting R.L.’s testimony under V.R.E. 404(b). Consistent with our purpose in considering this issue, we address the primary grounds selected by the trial court for admitting the testimony— that R.L.’s testimony shows that defendant’s conduct with M.C. was not *390accidental — recognizing that other grounds were raised during the course of the argument on the motion in limine and at trial.7
Defendant contends the trial court erred in admitting R.L.’s testimony because it did not relate to an element of the charged offenses or to his defense at trial. He argues that he neither claimed accident, nor placed his intent specifically at issue, and therefore the evidence was inadmissible under V.R.E. 404(b). In reviewing the trial court’s admission of prior bad act evidence under Rule 404(b), we determine whether the admitted evidence was relevant and material to the cause of action, and if so, whether its admission was so prejudicial as to outweigh its probative value. State v. Parker, 149 Vt. 393, 398, 545 A.2d 512, 515 (1988). “We will reverse the trial court’s decision to admit this evidence only if the court withheld or abused its discretion... and a substantial right of [the] defendant was affected by the alleged error.” State v. Kelley, 163 Vt. 325, 328, 664 A.2d 708, 710 (1995) (internal citations omitted).
Vermont Rule of Evidence 404(b) provides:
Evidence of other crimes, wrongs, or acts is not admissible to prove the character of a person in order to show that he acted in *391conformity therewith. It may, however, be admissible for other purposes, such as proof of motive, opportunity, intent, preparation, plan, knowledge, identity, or absence of mistake or accident.
We have explained that Rule 404(b) “allows evidence of uncharged misconduct for any purpose other than proving the defendant’s bad character.” Forbes, 161 Vt. at 332, 640 A.2d at 16. To be admissible, however, prior bad act evidence must be relevant to “an element of the offense or the defense that is genuinely in issue.” Winter, 162 Vt.. at 393, 648 A.2d at 627. “The State has the burden to show precisely how the proffered evidence is relevant to the theory advanced, how the issue to which it is addressed is related to the disputed elements in the case, and how the probative value of the evidence is not substantially outweighed by its prejudicial effect.” Id.
The State argues that absence of accident, or more generally defendant’s intent, was genuinely in issue for purposes of Winter because: (a) a defendant’s intent is always in issue because it is an element of the State’s proof once defendant pleads not guilty to the offense; and (b) intent was specifically in issue in this case because defendant claimed in his interview with the police that M.C.’s touching of his penis during the shower incident was accidental. We conclude that the evidence did not meet Winter’s “genuinely in issue” requirement and reject both of the State’s rationales as to why it did.
We note at the outset that the State’s first rationale would essentially negate the general prohibition of Rule 404(b) that prior bad act evidence is not admissible to show character or propensity. The State’s logic is (1) the prior bad act was intentional; (2) if the prior bad act event was intentional, it is more likely that the charged similar conduct is intentional; and (3) because the State must show intent as an element of the offense, the evidence is relevant to an element genuinely in issue. Since the State would not offer evidence of a prior bad act unless it appeared to be intentional, the practical result of the acceptance of the State’s argument is that all prior bad act evidence is admissible if it shows conduct similar to that involved in the charged offense.
The weakness in the State’s argument is that, while intent is an element of the offense, it is frequently not “genuinely in issue.” Indeed, if we were to accept the State’s rationale, we would turn “genuinely in issue” into “possibly in issue.” The general rule on the question before us is described in the leading treatise on prior bad act evidence:
*392In most of these cases, the defendant has implicitly conceded the existence of the material fact. Suppose, for example, that the defendant denies his identity as the criminal. By completely denying identity or raising alibi, the defendant in effect admits the mens rea of the perpetrators of the crime, at least when the mental element is general mens rea. The prosecution’s eyewitness testimony usually describes a crime that was obviously committed voluntarily. The jurors will readily infer the general mens rea from the testimony, and mens rea is indisputed in the sense that it is implicitly or practically conceded.
2 E. Imwinkelried, Uncharged Misconduct Evidence § 8:13, at 69 (1999). Thus, in State v. Blackey, 623 A.2d 1331, 1333-34 (N.H. 1993), the Supreme Court of New Hampshire held in a child physical abuse case in a day care center that the state could not put in evidence of prior physical abuse of other children at the day care center to show absence of accident where defendant defended on the basis that she did not commit the alleged act. Blackey is in accord with the majority of the decisions on the point. See, e.g., Clark v. State, 953 P.2d 159, 162 (Alaska Ct. App. 1998) (intent must be seriously disputed for prior bad act evidence to be admissible); State v. Ives, 927 P.2d 762, 771 (Ariz. 1996) (prior bad act evidence inadmissible where there was no genuine issue of whether defendant accidentally molested victim); State v. Goodrich, 432 A.2d 413, 417 (Me. 1981) (where defendant alleged that rape did not occur, evidence of other crimes not admissible to prove intent or identity); Emory v. State, 647 A.2d 1243, 1255-56 (Md. Ct. Spec. App. 1994) (prior bad act evidence not relevant where defendant never claimed mistake); State v. Rogers, 992 P.2d 229, 235 (Mont. 1999) (prior bad act evidence of defendant’s intent irrelevant when only issue was victim’s consent to sexual intercourse); State v. G.V., 744 A.2d 137, 142 (N.J. 2000) (prior bad act evidence not admissible where there was no genuinely disputed issue that intercourse with child was result of accident or mistake); State v. Picklesimer, 1996 WL 599425, at *4 (Ohio Ct. App. 1996) (state cannot use other acts evidence to negate accident or mistake when defendant has not raised that defense); State v. Smith, 617 N.E.2d 1160, 1172 (Ohio Ct. App. 1992) (evidence of other bad acts admissible to show mistake by defendant, not by state’s witnesses, and only if mistake is at issue); Blaylock v. Commonwealth, 496 S.E.2d 97, 103 (Va. Ct. App. 1998) (where defendant’s intent genuinely uncontested, probative value of prior bad act evidence outweighed by prejudicial value).
*393Putting aside the shower incident for the moment, the Imwinkeiried rationale exactly fits the circumstances of this case. Defendant was charged with inserting his fingers into the vagina of M.C., bringing his penis in contact with the mouth of M.C., and causing contact between the hand of M.C. and his penis. The charges are based on the testimony of M.C. that these acts occurred up to 100 times or more from the fall of 1997 through the spring of 1998.
Defendant consistently denied these allegations, and his denial was clear during the arguments on the motion in limine. He testified at trial and denied all of the sexual acts alleged by M.C. He did not claim accident, or lack of intent, as a defense. Neither in his opening nor in his closing did defendant’s counsel argue that defendant committed any of the acts alleged but did not have the criminal intent. Nothing in his examination or cross-examination of witnesses supported such a claim.8
As Professor Imwinkeiried explains, by relying on a defense that he did not commit the acts, a defendant has “implicitly or practically conceded” that he acted with criminal intent if the jury found that he did commit the acts. Imwinkeiried, supra, at 69. Certainly, no juror who believes the events testified to by M.C. could conclude that the acts were accidental. This case was about whether the acts occurred.
The State responds with its second argument, that defendant placed intent in issue by claiming that the touching during the “shower incident” was accidental. That argument might have some credence if defendant’s version of the shower incident bore any relation to the shower incident described by M.C. In fact, the only element similar in their descriptions is that both versions involved a shower. M.C. described the shower as a part of sexual activity, preceded by her rubbing defendant’s penis with fondling of sex organs continuing in the shower. Defendant described brief incidental contact occurring outside the shower as he left the shower and M.C. entered it. Defendant denied that there was any sexual contact *394before he took a shower and that he and M.C. showered together, the central points of M.C.’s description.9
The State argues this issue as primarily one of timing — that is, when issues appear in the criminal investigation, the State must respond to them even if they are not actually part of the trial. The State’s argument would make more sense in a case where the State had no idea what the defense would be until defendant put on his evidence and under criminal procedure rules that did not identify or narrow the issues.10 Here, the nature of the evidence and the actual issues in the trial were fully explored in the hearing on the motion in limine, and defense counsel made an opening presentation that identified defendant’s defense. At the time R.L. testified, the State and the trial judge knew that defendant would deny all of M.C.’s accusations including her version of the shower incident.
There are other reasons to reject the State’s argument. Defendant was charged with three counts, two of which have a maximum punishment of life in prison. Under the State’s theory, the accident issue did not relate to the more serious offenses, but only to the less serious charge of lewd and *395lascivious conduct. Significantly, the conduct which R.L. alleged — rubbing of her vagina — was not the conduct for which defendant arguably claimed accidental touching — having M.C. rub his penis.
We find two decisions from other states particularly helpful. In the New Jersey case of State v. G.V., the defendant was charged with sexually abusing his younger daughter over a five-year period. The trial court admitted evidence that he had similarly sexually abused his older daughter to show intent or absence of mistake. In language approved by the Supreme Court, the Appellate Division rejected this theory stating:
The supposed “possible defense” [of accident or mistake] was never raised by the defendant. Moreover, to do so in the context of this case would have been absurd. If we were dealing with an isolated incident, or even a few separate occasions, of allegedly improper touching, the “possible defense” might have been an issue. But this case involves an horrendous course of patent sexual depravity which continued for years. No reasonable defense, under these circumstances, would rely on the theory that these atrocious acts were simply misinterpreted expressions of fatherly affection.
Nor can it be fairly said that if the defendant committed the acts in question, there was a material factual dispute with regard to whether he was seeking sexual gratification. As stated in State v. Stevens, 115 N.J. 289, 301, 558 A.2d 833 (1989), a “necessary corollary to the principle that other-crime evidence can be admitted to prove any fact in issue... is the requirement that the “issue” be genuine, and that the other-crime evidence be necessary for its proof.” Neither of these requirements were satisfied here.
744 A.2d at 141-42.
In State v. Ives, 927 P.2d 762 (Ariz. 1996), the defendant was charged with four counts that he had directly or indirectly touched the private parts of three young female victims. The trial court admitted the testimony of a fourth young female who stated that the defendant had touched her vagina over her clothing on a number of occasions when she was four years of age. The state argued that the evidence was admissible to show intent and lack of accident or mistake. In support of its argument the state emphasized that the defendant admitted touching two of the victims, but argued that there was no sexual intent. The court rejected this argument:
*396[Defendant did not claim that he touched the girls in the proscribed maimer but did so by accident or mistake. Instead, defendant consistently maintained he did not touch the girls in any illegal way.... The state has repeatedly characterized portions of defendant’s testimony as admissions of touching the girls----[Defendant only admitted that he touched L.M. to help her up to a cabinet and that he touched T. J. to dry her off with a towel at her parents’ request. It seems likely that many defendants charged with child molestation will have had occasions for legal touching of the alleged victims.
There is simply no issue in this case as to whether defendant “accidently” or “mistakenly” rubbed the victims’ private parts. Instead, the issue is whether defendant did the acts at all. As the trial transcripts very clearly demonstrate, the issue at trial was one of credibility; did the jurors believe the girls or did they believe defendant?
Id. at 771 (internal citations omitted).
Under any reasonable construction of Winter’s “genuinely in issue” requirement, a defense of accident was not genuinely in issue in this case. The trial court erred in admitting the prior bad act evidence based on this rationale. While the trial court must be accorded discretion in its evidentiary rulings, we find that this ruling was an abuse of discretion.
We also address the trial court holdings that R.L.’s testimony met the requirements of V.R.E. 403 because its probative value was not substantially outweighed by its prejudicial effect. Again, this is a discretionary ruling of the trial court, reviewable here for abuse of that discretion. See Winter, 162 Vt. at 399, 648 A.2d at 631. In making this comparison, we assume the probative value of the evidence argued by the State.11 Given the dissimilarity of the prior bad act and the event for which *397defendant claimed the touching was accidental, we cannot find that the legitimate probative value of R.L.’s evidence was strong.
On the other hand, even a marginally relevant assertion that defendant had sexually abused his own daughter, and got away with it, was highly prejudicial. Any evidence of a similar past unpunished crime creates a great risk of unfair prejudice. See id. As we discussed above, we have in other decisions labeled such evidence as extremely prejudicial.
The State responds that the “relevant portion” of R.L.’s testimony covers only two transcript pages, and that the State did not even mention it in closing argument. These considerations go to whether the court’s Rule 403 rulings were harmless since they emerged only after the rulings were made. We recognize that R.L.’s testimony of her father’s sexual abuse of her was brief because it involved only one event. On the other hand, this accusation became a significant part of the trial. Almost half of the live witnesses testified about this event. Perhaps more damaging than the event itself was R.L.’s mother’s testimony that the police responded to the accusation and that defendant has never seen his daughter since the accusation. The State did not mention the testimony in closing argument, but defense counsel did address it, apparently believing it was too damaging to be ignored.
There is another point about R.L.’s evidence which bears on the Rule 403 analysis. R.L.’s mother testified that the police became involved as a result of her accusation. The jury heard that the police were involved, but did not hear, as the prosecutor conceded in a sidebar conversation to the judge, that defendant was not charged with a crime as a result of R.L.’s disclosure. We held in Winter, 162 Vt. at 401, 648 A.2d at 632, that in conducting a Rule 403 balancing, where the prior bad act is a crime that was investigated but not charged, the court must “weigh against admission the State’s failure to show why an allegation, that was not prosecuted... years ago, should be presented before the jury now.” The trial judge did not consider this factor, and the jury heard only the accusation and not the result of the police investigation.
In summary, we conclude that the prior bad act evidence was not admissible under V.R.E. 404(b) on the basis that it showed absence of accident. Even if it passed the Rule 404(b) test on this theory, its probative value was far outweighed by the prejudicial effect, and it did not meet the standard of V.R.E. 403.
*398III.
Finally, defendant contends the court abused its discretion, and violated his constitutional right to present a defense, by denying his request to have the jury hear the audiotape of his police interview. During the State’s cross-examination of defendant, the prosecutor asked whether the police officer had accurately testified that defendant appeared to be “cocky” and had laughed throughout his interview. Defendant denied that he had been cocky or laughed, and the prosecutor noted that “we can play the tape.” Defense counsel, in response, requested that the tape be played in its entirety to demonstrate defendant’s demeanor. Although the State did not oppose the request, it noted that the tape was about fifty minutes in length, and contained other potentially prejudicial statements which the court had earlier expressed a disinclination to get into and which would be difficult to edit out. The court ruled that it would admit a redacted transcript of the interview but denied the request to play the tape, noting that defendant could address the demeanor issue on redirect. Thereafter, on redirect examination, defendant testified that he was “nervous, uptight, and just a little bit distraught” during the police interview. The court subsequently observed with respect to the demeanor issue that counsel had “put that in... context on your redirect.”
Although not explicit in its ruling, the trial court apparently based its decision to exclude the audiotape on V.R.E. 403 because it would cause “undue delay, waste of time, or needless presentation of cumulative evidence.” V.R.E. 403. As noted, the State had expressed concern about the length of the audio recording and the need to redact irrelevant and potentially prejudicial statements throughout the tape.
The trial court enjoys broad discretion in ruling on the admissibility of evidence under Rule 403, and this Court reviews its rulings solely for abuse of that discretion. State v. McElreavy, 157 Vt. 18, 23, 595 A.2d 1332, 1334-35 (1991). Here, we note that the trial court had the benefit of hearing both the officer’s and defendant’s accounts of the interview, and of measuring their relevance and potential prejudicial impact in the overall context of the trial. The trial court’s judgment that the demeanor issue could be adequately addressed through redirect examination of defendant, and defendant’s subsequent explanation that he was nervous and distraught during the interview, leave no basis to conclude that the court abused its discretion. Accordingly, we discern no error.
Reversed and remanded.
The interview was videotaped and later shown to the jury at trial.
Although defendant testified to this incident of accidental contact, the State ignored it, neither making any charge based upon it nor offering it as a prior bad act.
At trial, defendant admitted that the photograph was of him, but testified that it was taken by J.C.,notM.C.
The alleged acts with A.G. consisted of his touching her vagina while they watched cartoons, while A.G.’s mother was elsewhere in the house. The State did not seek to admit this evidence at trial.
This Court upheld a seating arrangement in which the defendant and the child witness had a somewhat obstructed view of each other, State v. Dunbar, 152 Vt. 399, 404-06, 566 A.2d 970, 973-74 (1989), but that decision preceded the Supreme Court’s decision in Craig. In Dunbar, the court made case-specific findings of necessity based on evidence of trauma to the child. Id. at 405, 566 A.2d at 974.
The State cited State v. Plante, 164 Vt. 350, 355, 668 A.2d 674, 677 (1995), as a case in which this Court found that improper admission of prior bad act evidence was harmless error. Plante was decided before State v. Carter, 164 Vt. 545, 555, 674 A.2d 1258, 1265 (1996), in which we held that the beyond-a-reasonable-doubt standard applied to both constitutional and nonconstitutional errors alleged to be harmless, and did not use the harmless beyond-a-reasonable-doubt standard.
The State also argued from the beginning that the evidence was admissible to show a common scheme or plan. This rationale was not picked up initially by the judge who considered the motion in limine, but the court ruled the evidence admissible to show absence of accident, the State’s primary rationale, and “to demonstrate opportunity to commit the offenses.” In determiningwhether the evidence met the standard of V.R.E. 403, however, the court added that the evidence showed a common scheme or plan by defendant.
The issue was revisited at trial. The trial court ruled that R.L.’s evidence “would ... suggest, if not demonstrate, a common plan and scheme to take advantage of young girls.” The judge suggested an alternative rationale — that the evidence was admissible to respond to defendant’s claim that it was incredible that defendant would rub M.C.’s vagina in J.C.’s apartment when she was in the next room. The court noted that R.L. claimed that defendant did the same thing to her, and this showed that defendant was bold enough to engage in this conduct despite the high risk of discovery.
The trial judge charged the jury that it could consider R.L.’s evidence as bearing on whether defendant was capable of taking a chance of having a lewd contact with a child with another person present or whether defendant’s conduct was accidental or intentional, the primary rationale argued by the State. We cannot determine what rationale the jury might have used in considering the evidence.
Neither defendant nor the State has addressed the State’s first rationale of common scheme or plan or the motion judge’s rationale of demonstration of opportunity. Both defendant and the State have addressed the trial judge’s alternative rationale, but they have done so only briefly, and the trial court record related to this rationale is very limited. We conclude that we cannot, based on the state of the record and briefing, effectively address any of the possible alternative grounds for admission. If the State seeks to admit R.L.’s testimony during a retrial based on a rationale different from that considered in this opinion, the trial court can develop an appropriate record for review.
The dissent suggests he made such a claim by cross-examining M.C. during her video deposition to show that defendant was in the shower and was unaware of M.C.’s subsequent entry into the shower. The inference the dissent is trying to draw is obscure at best. M.C.’s testimony was that she started the shower, momentarily left the bathroom with the shower running, and returned to find defendant in the shower. The notion that defendant would enter a running shower, but be unaware of the person who turned it on, is incredible. The point of the cross-examination was to find discrepancies in M.C.’s testimony, not to setup a defense of accidental touching in the shower because defendant was unaware that M.C. was in the shower with him. Defendant’s consistent position was that he was never in the shower with M.C.
Although the dissent does not say so directly, we believe that much of the dissent’s position is based on the misconception that M.C. and defendant described the same shower incident, but disagreed on the intent involved in the sexual touching. Thus, the dissent talks about a juror “who believed defendant’s touching of M.C. in the shower was accidental.” (Emphasis added.) There is no evidence of accidental touching in the shower: M.C. described intentional touching in the shower, and defendant described accidental touching outside the shower.
This same misconception pervades reliance on many of the precedents from other jurisdictions cited by the dissent. For example, in Butcher v. State, 627 N.E.2d 855 (Ind. Ct. App. 1994), defendant admitted to being found nude lying on top of his minor daughter, after fondling her breasts, but defended on the basis of lack of intent because either he “(1) could not resist his daughter’s advances or (2) he was forced, against his will, to touch his daughter’s breasts.” Id. at859. In Day v. State, 643 N.E.2d 1 (Ind. Ct App. 1994), the defendant admitted improperly touching two ten-year-old girls who were staying with his daughter, as they testified, but “denied any improper intent to arouse or satisfy his sexual desires.” Id. at 2. In these cases, both the victim(s) and the defendant described the same incident, but the defendant claimed lack of intent.
That was exactly the situation in Michigan when the Supreme Court of Michigan decided People v. Vandervliet, 508 N.W.2d 114 (Mich. 1993), opinion amended by 520 N.W.2d 338 (Mich. 1994), one of the cases relied upon by the dissent. In the absence of a procedure to narrow the issues and require the defendant to articulate his defense, the court ruled that it was required to allow admission of the prior bad act evidence as bearing on the defendant’s intent. Id. at 127-28. The main holding of Vandervliet is, however, the prospective adoption of a procedure, like that existing in Vermont, to require the prosecution to give notice of the intent to offer prior bad act evidence and to authorize the trial judge “to require the defendant to articulate his theory or theories of defense.” Id. at 133. Under this procedure, the court prospectively adopted the Winter rale that an issue must be truly contested, and the point of Professor Imwinkelried that a defendant who contests only the doing of the illegal act implicitly concedes intent. See id.
The State’s theory, as adopted by the trial judge, was that the fact that defendant once touched his daughter’s vagina made it more likely that M.C.’s contact with defendant’s penis in or near the shower was not accidental. As Wigmore explains, logical relevance is present because of “the improbability of a like result being repeated by mere chance.” 2 J. Wigmore, Evidence § 302, at245 (Chadboumrev. 1979), quoted in In re S.G., 153 Vt. 466, 471, 571 A.2d 677, 680 (1990). We indicated reluctance to accept this rationale in In re S. G., where the State offered evidence of the physical abuse of an older sibling to show that the broken bone of a younger sibling was not accidental because it was indistinguishable from a character rationale. In re S.G., 153 Vt. at 471, 571 A.2d at 680.
Here, the reason to find that there was no logical relevance is even greater because the nature of the conduct was quite different on the two occasions. See, e.g., Clark v. State, 953 P.2d 159, 162 (Alaska Ct. App. 1998) (the evidence of other misconduct must be similar to the *397act of misconduct with which defendant is charged). The real point of the evidence in this case was to show that defendant is a child sexual abuser.