"Dr. Johnson is going to be testifying to his general qualifications, which are as a psychologist, an experienced psychologist. A psychologist who has been involved in a lot of assessments of children, a lot of research connected to assessments of children, interviewing of children, formation of questions regarding interviews of children, suggestibility research that has applied to adolescents, not just young children.
**582"There was a suggestion in this case earlier that we only worry about suggestibility when it applies to very young children, so he can address that topic from a standpoint of research. He can comment on the fact that he has reviewed the [CARES2 ] interview of [GP]. He has reviewed [Detective Massey's] interview of [JN].
"What's already pretty much established. [He can testify about the] absence of exploration of alternative theories or secondary gain[3 ] in the interview of [GP] relative to [JN].
"The fact that the methodology used by Detective Massey [in his interview of JN] involved not only leading questions, but suggestive questions, and to some degree, what an emotionally coercive question is.
"He will not be offering testimony on any bottom lines. He will not be opining on the credibility of any witness or any victim or the defendant. He will not be talking *1124about the results of any psychosexual evaluation."
The state objected, taking the position that Johnson's testimony about established protocols would be admissible but that testimony about whether those protocols had been followed would provide a "comment on the method of an interview" and would not be admissible. To permit Johnson to make that connection, the state argued, would allow him to impermissibly comment on the credibility of GP and JN, and therefore would violate the vouching rule.
Defendant sought to clarify that Johnson would not be commenting on whether the detective had engaged in an honest interview but rather would be testifying as to whether the detective's interview of JN, for example, raised "concerns for suggestibility." Notwithstanding that clarification, the trial court agreed with the state's position and ruled that Johnson was not "going to be talking about any of the interviews":
"I agree with [defendant] that Dr. Johnson can absolutely come in here and talk about interviews and how interviews **583should be conducted and-and suggestibility and what can be suggested, you know, leading questions and how they can, dah dah dah dah. I'm with you on that.
"He's just not going to get in to talk about any of the specific interviews in this particular case, because that's-that's just too close to comment on the credibility.
"* * * * *
"I could care less if he says that [the detective] did one of the worse interviews I've ever seen possible in the case here.
"It's the flip side of that is in so doing then, he's suggesting there that the credibility of the witness who made those statements has been affected and is not credible. And, therefore, it's commenting on the credibility of a witness.
"* * * * *
"I think you can certainly have Dr. Johnson testify to all the information that he knows and that all the things that would make an interview bad and what can happen as a result of those bad interviews in a generic sense and these good jurors will be able to tie the two together if they so desire."
Johnson testified in accordance with the trial court's ruling. He explained that appropriate protocols include asking open-ended questions and avoiding leading, suggestive, and emotionally coercive questions. He also testified that proper lines of inquiry are those that do not encourage particular responses and explore alternative hypotheses, including the potential for secondary gain. Johnson did not testify about whether the interviewers in this case followed those protocols or asked appropriate questions when they interviewed GP and JN. Nor did he testify about the victims' answers to the interviewers' questions or expressly state an opinion about whether the victims' statements about what had happened to them were truthful. Following the presentation of evidence and the parties' closing arguments, the jury found defendant guilty.
Defendant appealed. He argued that the trial court had erred when it sustained the state's vouching objection and prohibited Johnson from testifying that aspects of the **584interviews at issue were not conducted in accordance with established standards. The Court of Appeals affirmed, concluding that the testimony defendant wanted to offer violated the vouching rule. Black , 289 Or.App. at 258, 407 P.3d 992.
The court explained that the rule prohibiting vouching applies to both a witness's direct comments on the credibility of another witness and to comments that are "tantamount" to such direct comments. Id. at 261, 407 P.3d 992. The court took the position that a determination of whether testimony is tantamount to a direct comment on credibility requires a two-fold inquiry: (1) whether the testimony is a " 'commonly understood way[ ] of signaling a declarant's belief that a witness is telling the truth' or, instead, is relevant for a reason other than indicating that a witness may or may not be telling the truth"; and (2) whether the testimony is " 'sufficiently beyond the ordinary experience of a lay finder of fact' such that the expert testimony would help the jury make its own informed decision in evaluating a witness's credibility." Id. at 263, 407 P.3d 992 (quoting State v. Beauvais , 357 Or. 524, 543, 545, 354 P.3d 680 (2015) ).
*1125Applying that test, the Court of Appeals first determined that Johnson's proposed testimony was a commonly understood way of signaling his belief that GP and JN were not telling the truth. In the court's view, that testimony would suggest to the jury that the interviews "did not lead to truthful answers" and would not be "relevant for an independent reason." Id. at 264-65, 407 P.3d 992. At the second step in its analysis, the court determined that the proposed testimony would not "provide information that was 'sufficiently beyond the ordinary experience of a lay finder of fact' such that the expert testimony served an additional purpose in helping the jury make an informed decision about credibility." Id. at 264, 407 P.3d 992 (quoting Beauvais , 357 Or. at 545, 354 P.3d 680 ). Rather, the court concluded, defendant hoped to have his expert testify to "conclusions that the jury could adequately draw on its own without further witness assistance." Id.
Defendant filed a petition for review in this court, which we allowed. The questions before us are whether the trial court correctly sustained the state's vouching objection to Johnson's proffered testimony and, if not, whether the **585trial court's error requires that we reverse defendant's judgment of conviction.
The vouching rule is a judicially created rule of evidence, the exact contours of which can be "difficult to trace." State v. Chandler , 360 Or. 323, 331, 380 P.3d 932 (2016). The rule "developed largely in response to the use of expert psychiatric testimony to attack a witness's character," but it has come to apply to experts and lay witness alike. Id. at 330, 380 P.3d 932 ; see State v. Middleton , 294 Or. 427, 438, 657 P.2d 1215 (1983) (stating that rule applies to all witnesses). The purpose of the rule is to ensure that "the jury remains the sole arbiter of witness credibility and that the jury's role in assessing witness credibility is not usurped by another witness's opinion testimony." Chandler , 360 Or. at 330, 380 P.3d 932. To that aim, the rule against vouching prohibits a witness from making a direct comment, or one that is tantamount to a direct comment, on another witness's credibility. Beauvais , 357 Or. at 545, 354 P.3d 680. The rule "applies to credibility opinions about statements that a witness made either at trial or on some other occasion." Chandler , 360 Or. at 331, 380 P.3d 932 ; see also State v. Brown , 297 Or. 404, 443, 687 P.2d 751 (1984) ( OEC 608, which permits a party to attack or support a witness's credibility in the form of reputation or opinion evidence, "complements and supports the long-standing position of this court that no witness may pass upon the credibility of another witness in respect to specific conduct.").
As the parties note, and as the Court of Appeals determined, Johnson's proposed testimony would not have been a direct comment on the credibility of any witness. Defendant did not propose that Johnson would state directly that GP or JN were lying, thereby expressly conveying his opinion that their statements were not credible. See State v. Isom , 306 Or. 587, 591-92, 761 P.2d 524 (1988) ("We now inform all trial counsel that this type of cross-examination [-i.e. , if a witness says 'x,' would he be lying?-] will not be tolerated in any court in this state."); see also State v. Charboneau , 323 Or. 38, 42-43, 47, 913 P.2d 308 (1996) (state's introduction of a plea agreement that provides that "the State believes this charge accurately reflects the role [the witness] played in the death" of the victim and that **586"the state has reason to believe [that the witness's version of events] is true" analogous to a witness making an impermissible comment on the credibility of another witness). The question this case presents is whether Johnson's proposed testimony would have been tantamount to such direct statements.
We have stated before that "it is not always easy to draw the line between an inadmissible statement that is tantamount to a direct comment on the credibility of a witness and an admissible statement that is relevant for a different reason but that tends to show that a witness is telling the truth." Beauvais , 357 Or. at 545, 354 P.3d 680. That is so, at least in part, because much evidence, especially expert testimony, will "tend to show that another witness either is or is not telling the truth." Middleton , 294 Or. at 435, 657 P.2d 1215. Both parties propose rules that, they argue, help separate statements that are tantamount to a direct comment on *1126the credibility of a witness from those that are not.
Defendant's rule is categorical: Testimony is tantamount to a direct comment on the credibility of a witness and must be excluded if it rests on the witness's opinion that another witness is truthful or untruthful and conveys that opinion to the factfinder; testimony that conveys information from which the factfinder can make an independent determination of truthfulness is not vouching, and, subject to the other rules of evidence, is admissible. Applied here, defendant argues that Johnson's proposed testimony was not predicated upon his belief or disbelief of GP's or JN's statements. In fact, defendant asserts, he did not intend to ask Johnson any questions about those victims' statements, conduct, or demeanor; instead, he intended to ask Johnson about whether the interviewers had appropriately conducted their interviews of GP and JN. That testimony, defendant urges, would have been relevant and helpful to the jury notwithstanding whether Johnson believed or disbelieved the statements that GP and JN made in response to the questions that they were asked.
In contrast, the state's rule casts the vouching inquiry as requiring a weighing process. The state proposes that, when an expert witness does not make a direct **587statement about the credibility of another witness, but instead makes a statement that reflects an implicit opinion about that witness's credibility, a court must assess both the helpful effect of the evidence and the implicit opinion that inheres in it. A court may permit the implicit opinion if the primary effect of the evidence is helpful and sufficiently remote from the implicit credibility opinion. That analysis, the state argues, implicates the trial court's discretionary authority under OEC 7024 to assess the helpfulness of expert testimony and to balance (consistently with the principles established in OEC 4035 ) the merits of the testimony against the risk that it will improperly influence the jury. Applied here, the state contends that the trial court did not err in deciding to preclude Johnson from providing the proffered testimony. The jury had been provided with the established protocols and could identify the breach of those protocols on its own, so Johnson's testimony was not particularly helpful. On the other side of the balance, permitting Johnson to connect the dots for the jury carried a risk of conveying an improper opinion about witness credibility. In that instance, the state contends, the trial court did not abuse its discretion in prohibiting the testimony because the risk of improperly influencing the jury outweighed the minimal assistance that Johnson could provide.
As we will explain, both parties identify relevant questions that a court must address when considering the admissibility of expert testimony, but a trial court's determination of whether testimony violates the vouching rule is distinct from a determination of whether that testimony is permitted by the Oregon Evidence Code. As the trial court correctly understood, testimony that constitutes vouching is categorically inadmissible. Whether proffered testimony **588constitutes impermissible vouching is measured by whether it conveys one witness's opinion of the truthfulness of another witness, or, instead, provides information that permits the jury to make that determination. A review of our prior cases demonstrates the distinction.
In State v. Milbradt , 305 Or. 621, 623, 756 P.2d 620 (1988), the defendant was accused of raping two mentally disabled women, both of whom the defendant and his wife occasionally had had over to their house for overnight visits. At the defendant's trial, the state introduced the testimony of a psychologist who had interviewed the two women. Id. at 625, 756 P.2d 620. During his testimony, the psychologist *1127testified that he had experience with detecting deception and that, in his opinion, one of the victims did not display any of the indicators of deception. Id. at 626, 756 P.2d 620. He further opined that the victim's "limited mental capacity" would make it difficult for her to lie without "trip[ping] herself up five minutes later" and would "impair her ability to betray someone she perceives to be a friend," like the defendant. Id. at 627, 756 P.2d 620. This court held that that testimony was tantamount to a direct comment on the credibility of the victim. Id. at 630, 756 P.2d 620.
This court reached the same conclusion about the testimony at issue in State v. Keller , 315 Or. 273, 275, 844 P.2d 195 (1993), where a medical doctor testified on behalf of the state in a child sex abuse case. The doctor explained that, in reaching the diagnosis that the child victim had been sexually abused, she had looked for whether the victim had "been coached, been led, is fantasizing, or is indeed describing something that happened to her own body." Id. at 277, 844 P.2d 195. The doctor explained that the victim "had given a clear history of an episode of sexual touching which had happened to her own body" and that there "was no evidence of leading or coaching or fantasizing." Id. This court held that the doctor's statements "amount[ ] to testimony that the child was credible" and should not have been allowed. Id. at 285, 844 P.2d 195.6
**589Statements by an expert witness that another witness was not "deceptive" or "fantasizing"-as the experts in Milbradt and Keller testified-are statements that disclose an opinion about the truthfulness of another witness's statements and that convey that opinion to the jury. See Beauvais , 357 Or. at 543, 354 P.3d 680 (describing examples of statements "tantamount" to a direct comment on credibility as "commonly understood ways of signaling a declarant's belief that a witness is telling the truth"). Such statements signal the expert's belief that another witness is telling the truth, and they invade the jury's role as the sole arbiter of witness credibility.
The same is not true of statements that, although bearing on credibility, are not statements by one witness giving an opinion about another witness's credibility. In Middleton , the 14-year-old victim told several people that her father (the defendant) had raped her. 294 Or. at 429, 657 P.2d 1215. After the defendant was indicted, but before his trial, the victim stated that she had lied about the rape so that she could get "out on her own." Id. At trial, the victim again testified that the defendant had in fact raped her. Id. She was impeached on cross-examination with her pretrial recantation. Id. Following the victim's testimony, both the state and the defendant presented expert testimony from two social workers who worked with abused children. Id. Both experts, over the defendant's objections, were asked by the state and permitted to answer whether the victim's behavior was typical of young rape victims. Id. at 432-33, 657 P.2d 1215. The defendant appealed, and one of his arguments before this court was that the experts' testimony that the victim had behaved consistently with other victims of child sex abuse was an impermissible comment on the victim's credibility. Id. at 432-34, 657 P.2d 1215. This court held that that argument was "not well founded."
**590Id. at 435, 657 P.2d 1215. The court acknowledged that "[i]t is true that if the jurors believed the experts' testimony, they would be more likely to believe the victim's *1128account." Id. However, the court continued, "[n]either of the experts directly expressed an opinion on the truth of the victim's testimony." Id. For that reason, the court held that it was not error to admit the testimony over the defendant's vouching objection.
State v. Viranond , 346 Or. 451, 212 P.3d 1252 (2009), provides another example. There, two witness who, along with the defendant, had participated in a robbery testified against the defendant at his trial. Id. at 454, 212 P.3d 1252. The defendant cross-examined the witnesses about the lenient treatment they had received from the state and suggested that they had fabricated their testimony to curry favor with the prosecution. Id. The state then introduced the testimony of the detective who had interviewed the witnesses as part of the robbery investigation and who was present at trial when those witness testified. Id. at 455, 212 P.3d 1252. Over the defendant's objection, the detective testified that the witnesses' trial testimony was "consistent" with their pretrial statements. Id. at 456-57, 212 P.3d 1252. On review in this court, defendant argued that the detective's testimony "serve[d] no real purpose other than indirectly to bolster the credibility" of the witnesses and thus constituted an impermissible comment on their credibility. Id. at 460, 212 P.3d 1252. In addressing that argument, this court did not "deny that the prosecutor's hope in offering [the detective's] testimony was that it would" rehabilitate the witnesses' credibility, as that is what the testimony "was supposed to do." Id. However, that intent did not mean that the testimony was inadmissible. Id. The court explained that "consistency has no necessary connection with veracity," as "a witness may repeat the same lie multiple times." Id.
The witnesses whose testimony was at issue in Middleton and Viranond did not provide the jury with their opinions about the truthfulness of other witnesses; instead, they provided jurors with information that they could use to form their own opinions on that issue. The expert in Middleton informed the jury that the fact that the victim, a witness at trial, previously had recanted her statements was not unusual, and the detective in Viranond informed the jury that two witnesses' trial testimony was consistent **591with their pretrial statements. Although those facts could suggest that those witnesses were being truthful, the testimony was not vouching.
That testimony is not vouching, and therefore not categorically inadmissible, does not mean, however, that the testimony is necessarily admissible. See Chandler , 360 Or. at 334, 380 P.3d 932 (noting principle). To be admissible, proffered testimony must comport with other rules of evidence as well. That notion was apparent in Beauvais , where this court discussed the various questions that may be presented when a trial court is considering the admissibility of expert testimony. There, the state sought to have an expert testify to a diagnosis of sexual abuse based, in part, on physical evidence. Beauvais , 357 Or. at 528-29, 354 P.3d 680. After determining that that diagnosis was admissible under OEC 702 and OEC 403, this court took up the defendant's argument that the expert's testimony with respect to the evaluative criteria used to make that diagnosis was nonetheless inadmissible under the vouching rule. Id. at 540, 354 P.3d 680. Resolution of that argument lead this court through a discussion of Milbradt , Keller , and Middleton . Id. at 543-44, 354 P.3d 680. After explaining that the testimony in Milbradt and Keller was inadmissible because the declarants' statements were "commonly understood ways of signaling a declarant's belief that a witness is telling the truth," and that the testimony in Middleton did not have that "primary effect," the court explained that that difference is "not all that distinguishes Middleton from Milbradt and Keller ." Id. at 544-45, 354 P.3d 680. The court then went on to discuss Middleton 's further inquiry into what is properly understood as a discrete inquiry under OEC 702 -whether testimony is "sufficiently beyond the ordinary experience of a lay finder of fact such that expert testimony would" be helpful to the jury. Id. at 545, 354 P.3d 680.7
*1129In Middleton , this court considered whether the experts' testimony would be helpful to explain "superficially **592bizarre behavior by identifying its emotional antecedents." 294 Or. at 436, 657 P.2d 1215. The court explained that the test to analyze that question,
" 'is not whether a jury is capable of drawing its own inferences from the evidence presented. Rather, the test is whether the expert's testimony, if believed, will be of help or assistance to the jury.' "
Id. at 435, 657 P.2d 1215 (quoting State v. Stringer , 292 Or. 388, 391, 639 P.2d 1264 (1982) ). The court determined that the experts' testimony regarding victim recantation would "help the jury better assess the witness's credibility." Id. at 436, 657 P.2d 1215.
We do not read the court's discussion of Middleton in Beauvais as doing anything more than noting that, in Middleton , the court had been required to contend with objections based on OEC 702 and on the prohibition against vouching. The court's subsequent discussion of State v. Lupoli , 348 Or. 346, 234 P.3d 117 (2010), also rests on the idea that, in considering the admissibility of an expert's testimony, a trial court must address objections under the rules of evidence in addition to the vouching rule. To understand why, we must start with State v. Southard , 347 Or. 127, 218 P.3d 104 (2009).
In Southard , this court held that a diagnosis of child sex abuse that is not based, at least in part, on physical evidence is inadmissible under OEC 403, as understood in light of OEC 702. Id. at 140-42, 218 P.3d 104. The court reasoned that, in that circumstance, the probative value of the diagnosis is slight because it is based not on "an abstruse matter of science" but on evidence within the competence of a lay finder of fact and is substantially outweighed by the danger of unfair prejudice. Id. at 135, 142, 218 P.3d 104. In Lupoli , this court built on Southard and held that the evaluative criteria that are used to make such a diagnosis are similarly inadmissible. 348 Or. at 362-63, 234 P.3d 117. We explained that testimony that might be admissible when considered independently must be considered in context and, when so considered in that case, could not be separated from the related diagnoses. Id. Because, under Southard , it is OEC 403 and OEC 702 that make a diagnosis of sex abuse inadmissible when it is not based on physical evidence, it is logical to conclude that it also is those rules **593of evidence that make testimony about the criteria used in such a diagnosis inadmissible.
This court's discussion in Beauvais is consistent with that analysis. In Beauvais , this court concluded that, when a diagnosis of sexual abuse is admissible because it is adequately supported by physical evidence, testimony about the evaluative criteria underlying that diagnosis and the characteristics of the child that led to the diagnosis generally comports with the rules of evidence and is admissible. 357 Or. at 545-46, 354 P.3d 680. Such testimony cannot, however, include "a direct comment or a statement that is tantamount to stating that the child is telling the truth." Id. at 546, 354 P.3d 680. Thus, we understand Beauvais as recognizing a distinction between the rules of evidence, particularly those identified in Southard , and the vouching rule. An expert's opinion that reflects on the credibility of another witness must satisfy all evidentiary requirements.
In sum, when a party objects to testimony as improper vouching, a court must determine whether the testimony provides an opinion on truthfulness or, instead, provides a tool that the factfinder could use in assessing credibility. That determination does not necessarily require an assessment of whether that specific tool is permitted under the rules of evidence. For example, a court need not consider, as part of a vouching inquiry, whether expert testimony is sufficiently beyond the experience of a lay juror to meet the test of admissibility under OEC 702 and OEC 403. If a court determines that testimony constitutes vouching because it provides an opinion about the truthfulness of another witness and not information that could be helpful to jurors in forming their own opinions about that subject, the court must prohibit the testimony. If the testimony does not provide such an opinion, the court must then consider any other objections to the testimony that are raised by the parties.
*1130Applying the foregoing to the evidence at issue here, Johnson's proposed testimony would not have provided jurors with his opinion on the truthfulness of GP or JN. Rather, Johnson's testimony would have identified the ways in which the interviews of GP and JN fell short of established interviewing protocols and would have provided information that **594would have been helpful to the jury in assessing the credibility of those witnesses. The trial court erred in concluding that the testimony would have been impermissible vouching and in prohibiting it under that rule.
As indicated, however, that does not necessarily mean that the testimony that the trial court precluded Johnson from relating was admissible. All testimony must meet the requirements of the Oregon Rules of Evidence, and the state contends that that portion of Johnson's testimony was inadmissible under OEC 702 and OEC 403. Specifically, the state argues that the trial court reasonably concluded that Johnson's proposed testimony about the interviewers' alleged shortcomings would provide little help to the jury while carrying a risk of conveying an improper opinion about witness credibility. For that reason, the state argues, the trial court did not abuse its discretion in limiting Johnson's testimony under those rules of evidence.
The record does not indicate, though, that the trial court ever exercised its discretion under OEC 702 and OEC 403.8 As an initial matter, the state's objection to Johnson testifying with respect to the specific interviews was that it would be an impermissible comment on the credibility of witnesses. And, the trial court made clear that it agreed with the state, ruling that Johnson could not talk about any specific interview because it would be "commenting on the credibility of a witness." The trial court did state, as it finished its explanation for its ruling, that Johnson could testify to the established standards of interviews and that "these jurors are smart enough-in fact, very smart folks, that they can-they can [tie] together whatever you need them to tie together." However, we cannot say that that amounted to a ruling that Johnson's proposed expert testimony was inadmissible under OEC 702. See **595Stringer , 292 Or. at 391, 639 P.2d 1264 ("The test is not whether a jury is capable of drawing its own inferences from the evidence presented. Rather, the test is whether the expert's testimony, if believed will be of help or assistance to the jury."). In fact, if anything, the trial court suggested that Johnson's testimony, at least generally, would be admissible under OEC 702 ; the court determined, and the parties agreed, that no OEC 104 hearing was necessary because, in the court's view, "there's [not] going to be any issues whether [Johnson's testimony] is scientific" and would meet all relevant requirements.
Moreover, although the record suggests that the trial court identified the prejudicial effect of allowing Johnson to testify to the specific interviews-i.e. , that it would get "too close to comment on the credibility"-the record does not support an inference that the trial court identified the probative value of that evidence or weighed it against the prejudicial effect it identified. See State v. Anderson , 363 Or. 392, 406, 423 P.3d 43 (2018) ("[A] court will make a sufficient record * * * if the trial court's ruling, considered in light of the parties' arguments, demonstrates that the court balanced the appropriate considerations."); State v. Mayfield , 302 Or. 631, 645, 733 P.2d 438 (1987) (in making a decision under OEC 403, a judge is to identify the probative value of evidence and balance it against its prejudicial effect; the judge errs "if the judge fails to exercise discretion, refuses to exercise discretion[,] or fails to make a record which reflects an exercise of discretion."). And, as we stated above, the state did not object to the evidence on OEC 403 grounds, and the parties did not make any argument with respect to that rule. See *1131Anderson , 363 Or. at 397-98, 423 P.3d 43 (objection and parties arguments concerned the probative value of evidence measured against its prejudicial effect); State v. Johanesen , 319 Or. 128, 138-39, 873 P.2d 1065 (1994) (same). Accordingly, we cannot conclude, as the state argues, that the court exercised its discretion and permissibly excluded Johnson's testimony under OEC 702 or OEC 403.
The final question for our consideration, then, is whether the trial court's error in excluding Johnson's testimony requires reversal. Errors that had "little likelihood of affecting the verdict" are harmless and do not warrant **596reversal. State v. Henley , 363 Or. 284, 307, 422 P.3d 217 (2018). "In making a determination of harmlessness, the court does not ask whether the evidence of guilt is substantial or compelling, but rather whether the trial court's error was likely to have influenced the verdict." Id. An error is less likely to be harmless where it relates to a central factual issue in the case. State v. Bement , 363 Or. 760, 779, 429 P.3d 715 (2018). "But even when evidence relates to a central factual issue, its exclusion may be harmless if it is 'merely cumulative of,' instead of 'qualitatively different than,' evidence presented to the factfinder." Id. (quoting State v. Davis , 336 Or. 19, 34, 77 P.3d 1111 (2003) ). We note that "scientific evidence or evidence perceived by lay jurors to be scientific in nature possesses an unusually high degree of persuasive power." Henley , 363 Or. at 307, 422 P.3d 217.
The state gives two reasons for its harmlessness argument-that Johnson's proposed testimony did not relate to a central factual issue, and that it was not qualitatively different than the evidence that the jury heard. With respect to its first reason, the state contends that defendant took the position at trial that the allegations against him were fabricated for reasons unrelated to flaws in forensic interviewing. Thus, the state asserts, the principal question before the jury was whether it should give any credence to the victims' allegations of abuse, including allegations that were made before any forensic interviews were conducted. In that context, the state urges, there is little likelihood that the jury's ultimate conclusion would have been affected by Johnson's identification of potential flaws in those interviews.
The state is correct that defendant questioned the credibility of the victims' initial allegations of abuse, but his case was not so limited. Throughout trial, defendant consistently raised an issue with the quality of the state's investigation in its entirety. Defendant argued in opening statements that, although it was conducted with the best intentions, the investigation was not objective and law enforcement settled on defendant's guilt as soon as the first disclosure by GP was made. Defendant thoroughly questioned witnesses on those issues and drew the jury's attention to them in closing argument:
**597"It's not about a conspiracy. It's about the government and the government's community partners not doing a good job, and it's spilling over into the very nature and quality of the evidence that's presented to you. And it's spilling over in impact on the evidence that's been presented to you in this case. * * *
"So what the State wants you to do is exercise your emotion, ignore the State's use of emotional coercion in the course of investigation, and even ignore the-appeal to your own emotions, ignore how the State sculpted the flow of evidence in this case. How the government sculpted the flow of evidence in this case through its conduct, and only look at selective evidence that was consistent with the State's theory of guilt, and then of course appeal to your emotions, and ask you to find the defendant guilty."
Johnson's testimony was crucial to defendant's assertion that the conclusions that the state reached were not reliable, and the state is incorrect in its argument that his testimony did not relate to a central issue in this case.
The second reason that the state advances to support its argument that the trial court's error was harmless also is unpersuasive. The state contends that Johnson's proposed testimony would have been cumulative of, and not qualitatively different than, evidence that defendant adduced from other witnesses. Specifically, the state argues that Johnson's general testimony about the appropriate interview protocols, including the kinds of questions that are permitted and the need to *1132explore secondary gain, fully equipped the jury to identify "potential flaws" in the forensic interviews of GP and JN. Having Johnson explicitly identify problematic portions of those interviews, the state posits, was unnecessary, especially considering that defendant repeatedly focused the jury's attention on those portions of the interviews throughout trial.9
The state is correct that there was evidence from which the jury could have identified flaws in the forensic interviews of GP and JN. Defendant thoroughly questioned **598the interviewers about questionable aspects of their respective interviews, and, in doing so, he highlighted their "potential flaws." But those examinations did not yield any admissions that those aspects of the interviews were in fact problematic. Rather, the interviewers explained away or justified their techniques, or, at best, they minimized the significance of their errors. Although defendant's offer of proof was abbreviated, it informed the court that Johnson would not only identify established interview protocols but also would specify the deficiencies in the interviews that were conducted in this case. The interviewers' testimony did not fit that bill.
To be sure, the detective who interviewed JN acknowledged that interviews must be carefully constructed to avoid influence on the answers given, and that he had begun his interview with JN by telling him that it was important to hold defendant accountable and that the information JN had was important. The detective did not concede error, however; instead, he offered a justification:
"[Beginning the interview that way] would be consistent in my behavior when I'm talking to any young person, or anybody whatsoever that is involved in a case, is just discussing how important what they have to tell me is, and reassuring them that, you know, I'm there to listen and what they have to say to me is important."
The detective also did the same thing in another exchange highlighted by the state. There, the detective acknowledged telling JN about "classic defense strategies" in these kinds of cases; but, when asked how defense strategies were relevant, he did not concede that that topic was improper. Instead, the detective again defended his approach:
"Well, if it was toward the end of the interview and I didn't know how much more contact I would have with them, I'd like to do everything I could to prepare them for the difficulties of what might be ahead of them."
Justification and defense of an interview is qualitatively different than expert testimony that the interview was flawed.
Perhaps the exchange referenced by the state that best supports its argument that Johnson's testimony would have been duplicative is the following:
**599Counsel: "Do you remember talking to [JN] that his coming forward and what he had to say could stop [defendant] from harming other kids?
Detective: "I don't specifically recall those words, but-
Counsel: "Would you agree that that concept or that kind of statement puts a lot of pressure and responsibility onto an 18- or 19-year-old?
Detective: "It may.
"* * * * *
Counsel: "And of course, one of the things that pressure can do is it can influence people, can it not?
Detective: "Depending on the circumstances, possibly.
Counsel: "Do you recall sharing with [JN] relevant to [defendant] that, 'It's really important that we stop him?'
Detective: "I don't recall those specific words, but it was very important for me, yes.
Counsel: "But you agree, you have a job, right?
Detective: "Yes.
Counsel: "This wasn't about [JN's] job, right? [JN], in terms of whether or not he's a victim of a crime, it's not about his employment obligations, correct?
*1133Detective: "That's correct.
Counsel: "It's about what has or has not happened to him, correct?
Detective: "That is correct.
Counsel: "It's not about what his social obligations or responsibilities are to society or the community, correct?
"* * * * *
Detective: "[JN]'s social responsibilities are completely up to [him]. I cannot answer that question, Sir.
Counsel: "And you would agree that it's not part of your job, when conducting a forensic interview or a first interview, to try and influence [JN] in terms of what his responsibilities are or are not, correct?
Detective: "No [that is not my job]."
**600But even the detective's acknowledgments in that discussion were not definitive. The detective testified only that such statements "may" put pressure on a teenager; it depended on the circumstances.10 Considered as a whole, the detective's testimony suggests that he did not think such circumstances were present here. For example, when questioned by defense counsel about the use of leading-questions in an interview, the detective again stated that it depended on the circumstances whether their use was appropriate:
Counsel: "You have training about how to conduct an interview, do you not?
Detective: "Yes.
"* * * * *
Counsel: "And * * * one of your goals is to ask reasonably open-ended questions so that you are not over-influencing an answer with your presuppositions or goals, correct?
Detective: "Based on circumstances, yes.
Counsel: "Well, is it ever appropriate for you to ask questions in a manner that shapes or sculpts the answer to be consistent with a presupposition or bias about what may have happened in your opinion? Is it ever appropriate to ask those kinds of questions of a critical witness as a detective?
Detective: "Based upon-depending on the circumstances, sometimes that can happen. It's not ideal. But, yeah, sometimes it can happen. If you're trying to recap something that somebody has already gone over with you briefly, it's a difficult moment for them to explain it, sometimes you do ask leading questions to help them get through it, you know, especially if it's something that you're already previously aware of."
**601The detective also offered a similar justification when the prosecutor questioned him about the same topic:
Prosecutor: "Open-ended questions, lots of talk about leading witnesses, you described some times in which leading witnesses would be appropriate. In your training and experience are-how might your concern be different about leading witnesses when you're dealing with teenagers, meaning 14 to 17-year-olds, as opposed to preschool ages?
Detective: "There's a significant difference in the cognitive abilities of teenagers versus preschoolers or grade school children.
Prosecutor: "So would you be more likely as a detective to use leading questions when having a conversation with a teenager or a preschooler?
Detective: "Much more likely with a teenager."
And the detective justified his approach to the investigation when defense counsel questioned him about his investigation of GP's allegations. Defendant noted that when GP first officially disclosed defendant's abuse, he told his primary care provider that he was *1134upset with defendant because defendant had imposed dating restrictions on his daughter after discovering that she and GP had been intimate. Although the report containing that statement was available to the detective, neither the detective nor the CARES interviewer had attempted to obtain it or were aware of the notation. When questioned about that report, the detective acknowledged that such documentation can be important but testified that in this case it was not:
Counsel: "[I]sn't it important, Detective, when you get contextual information about potential bias, secondary gain or motive on the part of a complaining witness or victim, isn't it important when you have documentation both from the suspect and the alleged victim that addresses the same topic or the same potential motive?
Detective: "In some cases it can be.
Counsel: "Okay. In this case you didn't consider it to be important?
Detective: "No."
**602In all, we do not consider the detective's testimony to be an adequate substitute for Johnson's proposed testimony. The detective did not acknowledge that his statements or questions were problematic, and, when pressed on certain statements, he stated that their use would not be problematic under some circumstances. Particularly in light of the fact that the detective had testified that he had received training in how to conduct interviews and had experience in doing so, the jury may have given substantial weight to his explanations and relied on the answers that JN and GP gave in their interviews. Had Johnson been permitted to testify, he would have been able to point out the specific ways in which the detective failed to follow standard protocols and to explain whether deviations from those protocols were warranted.
We also do not view Johnson's prospective testimony about GP's CARES interview as necessarily duplicative. The error that defendant identified in that interview was that the CARES interviewer did not fully explore the motives for GP's allegations, specifically that GP had expressed frustration with defendant's dating restrictions. Although the CARES interviewer acknowledged that she had not obtained the primary care provider's report containing that information and that she had not specifically explored whether GP was frustrated with defendant for interfering with GP's relationship with his daughter, she did not acknowledge that that mistake was significant. The CARES interviewer testified that she had generally explored the topic of secondary gain and that she was unsure that the primary care provider's report "would have changed any of [her] interview with" GP. Johnson could have testified about whether the interviewer should have proceeded differently in light of that report.
In sum, we hold that Johnson's proposed testimony related to central issues in the case and would not have been cumulative of the evidence that defendant was able to adduce from other witnesses. Johnson would have testified to the specific instances in which the interviewers failed to follow established protocols. The interviewers did not acknowledge those deficiencies, and we therefore conclude that the trial court's failure to admit the proffered evidence was not harmless.
**603The decision of the Court of Appeals is reversed. The judgment of the circuit court is reversed, and the case is remanded to the circuit court for further proceedings.
Garrett, J., dissented and filed an opinion, in which Balmer, J. joined.
CARES is a clinic that evaluates children who may have been sexually abused. State v. Lupoli , 348 Or. 346, 350, 234 P.3d 117 (2010).
Johnson explained this as "acting in a way in order to achieve some other end, another end that may not be obvious at-at the outset."
OEC 702 provides:
"If scientific, technical or other specialized knowledge will assist the trier of fact to understand the evidence or to determine a fact in issue, a witness qualified as an expert by knowledge, skill, experience, training or education may testify thereto in the form of an opinion or otherwise."
OEC 403 provides:
"Although relevant, evidence may be excluded if its probative value is substantially outweighed by the danger of unfair prejudice, confusion of the issues, or misleading the jury, or by considerations of undue delay or needless presentation of cumulative evidence."
The testimony at issue in Keller also would be inadmissible on a separate ground, based on a trio of this court's recent cases-Southard , Beauvais , and Lupoli . Through the course of those cases, this court has held that a diagnosis of child sex abuse that is not based, at least in part, on physical evidence is inadmissible under OEC 702 and OEC 403. State v. Southard , 347 Or. 127, 141-43, 218 P.3d 104 (2009). The same is true of the evaluative criteria that are used to make such a diagnosis, as they cannot be separated from the diagnosis itself. Lupoli , 348 Or. at 362-63, 234 P.3d 117. However, where the diagnosis is based on physical evidence and is admissible, the evaluative criteria that went into that diagnosis also are admissible. Beauvais , 357 Or. at 546, 354 P.3d 680. In Keller , the expert had diagnosed the victim with sex abuse in the absence of physical evidence, and her testimony with respect to "coaching" was a factor that she used to help her reach that diagnosis. See 315 Or. at 275, 844 P.2d 195 (noting that doctor had found no physical evidence of sexual abuse and that expert had reviewed that report in reaching diagnosis). Thus, even if that statement was not inadmissible vouching, it nonetheless would have been inadmissible.
The Middleton court understood the defendant to have made two independent objections: (1) the experts' testimony was an impermissible comment on credibility and (2) the experts' "answers would be opinions." 294 Or. at 432-34, 657 P.2d 1215. With respect to the latter, the court stated, "[w]e take this to mean opinions that are inadmissible because they are not helpful to the jury, for it is well settled that experts may offer opinions in some situations."Id. at 434, 657 P.2d 1215.
Not all questions that arise under OEC 702 are questions that permit an exercise of discretion. See State v. Jesse , 360 Or. 584, 598-600, 385 P.3d 1063 (2016) (reviewing for error of law trial court's ruling under OEC 702 that expert's testimony was not helpful because "it lacked a sufficient nexus to [the] defendant's theory of defense that his apparent admissions to [his wife] and the deputies were not actual confessions of guilt"); State v. Rogers , 330 Or. 282, 315, 4 P.3d 1261 (2000) ("[T]his court reviews without deference for errors of law whether a trial court properly applied OEC 702 to decide whether an expert is qualified to give testimony relative to a particular topic [.]" (Emphasis in original.)).
The state also argues that defendant identified the problematic parts of the interviews in his closing arguments. That argument is not well taken. What a party says in closing argument is not evidence, and the jury in this case was instructed on that point. See UCrJI 1004 (defining evidence).
The state notes that the CARES interviewer testified that, as a general matter, she would "try" to avoid telling a child that, if the abuse happened, then "we need to worry about the safety of other children." That testimony, the state argues, alerted the jury that the detective's similar statement to JN was not proper. The problem with that argument, however, is that the CARES interviewer testified to a hypothetical and as a general matter, whereas Johnson's testimony would have been based on the actual interview at issue and its attendant circumstances. In other words, the testimony of the CARES interviewer, coupled with the detective's acknowledgement that he made the statement (but not that the statement was improper), would not be the same as having Johnson opine whether the detective's statement was improper under the circumstances.