¶37 (dissenting) — A jury convicted Brandon Coristine of raping his new roommate on her first night at her new place of residence. I agree with the majority that giving an affirmative defense instruction over Coristine’s objection violated his Sixth Amendment right to control his own defense. But not every error merits reversal. This one certainly does not. I respectfully dissent.
*387¶38 I also write separately to observe that we have not established a clear framework of harmless error review in Washington. Until we build and consistently follow such a framework, our opinions in this area will be confusing and inconsistent and review for harmless error will continue to be “an arbitrary exercise of judicial authority.” Dennis J. Sweeney, An Analysis of Harmless Error in Washington: A Principled Process, 31 Gonz. L. Rev. 277, 323 (1996).
¶39 We can do better. A review of our case law reveals workable standards and suggests a framework that would foster clarity and consistency in this difficult area of the law. In my view, this court should explicitly require a three-step process for appellate review of constitutional errors at criminal trials. First, an appellate court should evaluate whether the error in question requires automatic reversal — what is sometimes called “structural” error. See, e.g., State v. Frost, 160 Wn.2d 765, 779-82, 161 P.3d 361 (2007). If the error is among these rarest of errors, the court should simply reverse. Second, if the error is not structural, the court should consider whether the error was plausibly relevant to the verdict. See, e.g., State v. Bobenhouse, 166 Wn.2d 881, 894, 214 P.3d 907 (2009). If the court is convinced, beyond a reasonable doubt, that the error was not plausibly related to the verdict, the court should affirm the conviction. Third, if the error is plausibly relevant to the challenged guilty verdict, the court should reverse unless it is persuaded, beyond a reasonable doubt, in light of the full record, that the error did not tip the scales from innocence to guilt. See, e.g., State v. Rice, 120 Wn.2d 549, 569, 844 P.2d 416 (1993); State v. Easter, 130 Wn.2d 228, 242, 922 P.2d 1285 (1996). This three-step process would provide our courts with a clear, workable, and effective way of conducting harmless error review and adjudicating whether a constitutional error at trial is harmless. It would foster clearer, more consistent, and more objective precedents. It would help “ ‘cleanse the judicial process of prejudicial error without becoming mired in harmless error.’ ” State v. *388Grenning, 169 Wn.2d 47, 59 n.8, 234 P.3d 169 (2010) (quoting Roger J. Traynor, The Riddle of Harmless Error 81 (1970)).
1. Harmless Error: Federal Standards
¶40 United States Supreme Court precedent has established the minimum requirements state courts must follow in evaluating constitutional error. See Chapman v. California, 386 U.S. 18, 20-21, 87 S. Ct. 824, 17 L. Ed. 2d 705 (1967). An otherwise valid conviction should not be set aside where a reviewing court is convinced beyond a reasonable doubt that the error was harmless. Errors are inevitable, and many errors, even constitutional ones, do not call the fundamental fairness of the trial into question or affect the final outcome. See Delaware v. Van Arsdall, 475 U.S. 673, 681, 106 S. Ct. 1431, 89 L. Ed. 2d 674 (1986). Upholding fair criminal convictions “promotes public respect for the criminal process by focusing on the underlying fairness of the trial rather than on the virtually inevitable presence of immaterial error.” Van Arsdall, 475 U.S. at 681. It avoids the myriad costs associated with retrials, both to the justice system and to the witnesses haled into court again and again. See United States v. Hasting, 461 U.S. 499, 509, 103 S. Ct. 1974, 76 L. Ed. 2d 96 (1983). In contrast, reversing for harmless errors “ ‘encourages litigants to abuse the judicial process and bestirs the public to ridicule it.’ ” Van Arsdall, 475 U.S. at 681 (quoting Traynor, supra, at 50).
¶41 There are a limited number of errors that warrant automatic reversal. See, e.g., United States v. Gonzalez-Lopez, 548 U.S. 140, 148-50 & n.4, 126 S. Ct. 2557, 165 L. Ed. 2d 409 (2006); Neder v. United States, 527 U.S. 1, 8, 119 S. Ct. 1827, 144 L. Ed. 2d 35 (1999). Such errors are sometimes referred to as “structural” errors. See, e.g., Gonzalez-Lopez, 548 U.S. at 148. An error may be deemed structural because it is intrinsically harmful, Neder, 527 U.S. at 7, because it necessarily brings into question the fundamental fairness or reliability of the entire trial process, id. at 8-9; Arizona v. *389Fulminante, 499 U.S. 279, 309-10, 111 S. Ct. 1246, 113 L. Ed. 2d 302 (1991), because assessing its effect in a given case is highly difficult or impossible, Gonzalez-Lopez, 548 U.S. at 149 n.4, 150, or because the error substantially implicates some weighty interest other than ensuring a reliable and accurate outcome, id. at 149 n.4; McKaskle v. Wiggins, 465 U.S. 168, 177 n.8, 104 S. Ct. 944, 79 L. Ed. 2d 122 (1984). I wholeheartedly agree with the majority that the error here was not structural.
¶42 For all other constitutional errors, the appellate court must consider whether the error actually might have affected the trial’s outcome in order to determine whether reversal is appropriate. See, e.g., Neder, 527 U.S. at 17-19. If the court is persuaded, beyond a reasonable doubt, that the error did not affect the verdict, the conviction should be upheld. Van Arsdall, 475 U.S. at 681; see also Chapman, 386 U.S. at 24. Most constitutional errors are subject to harmless error review, Gonzalez-Lopez, 548 U.S. at 148, and most such errors are indeed harmless, Hasting, 461 U.S. at 509. The court will consider any plausible effects that such an error might have had on a rational jury, and the error will be deemed harmless if the reviewing court concludes beyond a reasonable doubt that “the jury verdict would have been the same absent the error” and thus “the error ‘did not contribute to the verdict obtained.’ ” Neder, 527 U.S. at 17 (quoting Chapman, 386 U.S. at 24).
¶43 Neder was convicted of fraud for using wildly inflated real estate appraisals to secure loans, concealing his ownership interests in various shell corporations, and falsifying documentation. Id. at 4-5. The trial court failed to instruct the jury that the defendant’s false statements must have been “material [ ]” to convict, omitting an element of the charged crimes. Id. at 6. The evidence establishing the materiality of the false statements was uncontroverted and overwhelming. See Neder, 527 U.S. at 6, 18-20. Thus, the court had little difficulty determining the error was harmless beyond a reasonable doubt.
*390¶44 The United States Supreme Court requires that harmless error analysis be thorough and substantive. The appellate court must consider the error in the context of the trial as a whole. Neder, 527 U.S. at 19. The appellate court also must provide substantive reasoning and analysis in support of its conclusions. See Clemons v. Mississippi, 494 U.S. 738, 753-54, 110 S. Ct. 1441, 108 L. Ed. 2d 725 (1990) (vacating state court’s “cryptic holding” of harmlessness that was unsupported and “difficult to accept” without “a detailed explanation based on the record”). Requiring thorough review of the record and detailed explanation helps to ensure that harmless error review is conducted properly and consistently over time. State courts must comply with these minimum federal requirements and abide by the standards that the United States Supreme Court has established. Beyond these established federal standards and minimum requirements, states are left to develop their own concrete and detailed frameworks and standards, consistent with federal law, to govern harmless error review in state courts. See, e.g., State v. Guloy, 104 Wn.2d 412, 425-26, 705 P.2d 1182 (1985); State v. Tollardo, 2012-NMSC-8, 275 P.3d 110.
2. Harmless Error: Washington
¶45 Washington does not have a clear, workable, and effective legal framework to govern judicial review of constitutional errors at criminal trials. See, e.g., Sweeney, supra, at 287 (noting that our cases have done “little ... to provide the analytical framework necessary to apply the doctrine of harmless error”). As a result, our adjudications of harmlessness have been unpredictable. The closest this court has come to adopting a comprehensive framework to govern harmless error review was in Guloy, when we intended “to settle the question of what type of standard this court will use in its harmless error analysis” and with little discussion or explanation selected “ ‘the overwhelming untainted evidence’ ” test over “the contribution test.” 104 *391Wn.2d at 425-26; see also State v. Evans, 96 Wn.2d 1, 6-10, 633 P.2d 83 (1981) (Brachtenbach, C.J., concurring) (discussing these two tests). Specifically, we decided that harmlessness would not depend on whether the error could have played a part in the actual jury’s determination of guilt (the contribution test) and decided that harmlessness would instead depend on whether “the untainted evidence is so overwhelming that it necessarily leads to a finding of guilt” (the overwhelming evidence test). Guloy, 104 Wn.2d at 426. The contribution test has since been rejected by the United States Supreme Court. See Neder, 527 U.S. at 11-13, 17-18. Unfortunately, the overwhelming evidence test fares little better.
¶46 The overwhelming evidence test ignores a wide variety of potential trial errors with differing effects and is an overly broad standard that properly applies in only a limited set of circumstances. First, the standard ignores that some trial errors are plausibly relevant only to discrete elements or issues. In Neder, for example, the jury instructions omitted one element of the crime, and that error was properly found harmless because of uncontested and overwhelming evidence of the omitted element, not of guilt generally. 527 U.S. at 18. Second, the overwhelming evidence test ignores that some trial errors are not plausibly relevant to the verdict at all. For example, in State v. Kroll, 87 Wn.2d 829, 558 P.2d 173 (1976), the jury instructions erroneously shifted the burden of proof to the defendant as to a certain element of second degree murder but properly instructed the jury as to first degree murder. The jury convicted on first degree murder, and we rightly concluded that the error was harmless. 87 Wn.2d at 840-41. Our conclusion did not depend on any quantum of evidence because the error was not even plausibly relevant to the challenged verdict in the first place. Third, the overwhelming evidence test ignores that the amount of untainted evidence necessary to render an error harmless beyond a reasonable doubt depends upon the force with which the *392error plausibly might have swayed the jury toward guilt. See Glasser v. United States, 315 U.S. 60, 67-68, 62 S. Ct. 457, 86 L. Ed. 680 (1942). In sum, the overwhelming evidence test is appropriate only when an error is plausibly relevant to the entire verdict and might have exerted substantial force in swaying the jury toward guilt. It is probably for this reason that the United States Supreme Court originally cautioned against “overemphasis” on “ £over-whelming evidence’ ” in harmless error review. Chapman, 386 U.S. at 23.
¶47 In light of the overwhelming evidence test’s clear failings, it is not surprising that this court has not consistently applied it. Instead, we have frequently determined harmlessness not by weighing the error against the evidence but by considering the limited relevance of an error to the challenged verdict. In some cases, errors have been relevant only to discrete elements or issues, and thus, in determining harmlessness we have considered only evidence related to those elements or issues. See, e.g., State v. Watt, 160 Wn.2d 626, 639-40, 160 P.3d 640 (2007); State v. Thomas, 150 Wn.2d 821, 844-45, 83 P.3d 970 (2004). In many cases, errors simply have not been plausibly relevant to the challenged verdict, and we have found those errors harmless for that reason alone, regardless of the evidence of guilt. See, e.g., Bobenhouse, 166 Wn.2d at 894-95; State v. Berube, 150 Wn.2d 498, 509, 79 P.3d 1144 (2003).6 Thus, although we often speak of the overwhelming evidence test *393as if it were our universal standard for harmlessness, see, e.g., Frost, 160 Wn.2d at 782; Watt, 160 Wn.2d at 635-36; Guloy, 104 Wn.2d at 426, we also often ignore it. This inconsistency will continue until we identify a clear, workable, and effective framework to govern review of constitutional errors at criminal trials.
3. Synthesis
¶48 The three-step framework I propose reflects existing federal standards and much of our own precedent. If adopted explicitly, this framework would provide a clear, workable, and effective way of promoting consistent and just adjudications in this difficult area of the law.
¶49 First, the reviewing court must determine whether the error warrants automatic reversal or is instead subject to constitutional harmless error analysis. See, e.g., Gonzalez-Lopez, 548 U.S. at 148-50 & n.4; Frost, 160 Wn.2d at 779-82. If an error is structural, the defendant’s conviction must be reversed.
¶50 Second, if the error is subject to harmless error analysis, the State must show that the error was not plausibly relevant to the verdict. See, e.g., Kroll, 87 Wn.2d at 840-41. Properly determining whether and how a given type of error had the potential to sway an honest, fair-minded, and reasonable jury away from innocence and toward guilt must depend upon logic, common sense, experience, and scientific inquiry. Mere speculation is insufficient in light of the fact that “academic” and “technical” errors are the very reason the harmless error doctrine was created in the first place. Guloy, 104 Wn.2d at 426. Whether a given error was plausibly relevant to a particular verdict will depend upon the context of the error and often will require a searching review of the record. See, e.g., State v. Bonds, 98 Wn.2d 1, 17-18, 653 P.2d 1024 (1982) (jury’s *394related finding rendered instructional error logically irrelevant); State v. Maupin, 128 Wn.2d 918, 929, 913 P.2d 808 (1996) (noting that “a witness’s proffered testimony [may be] so incredible that its exclusion is harmless error,” but such a determination must be made “from the record”). If the court concludes, beyond a reasonable doubt, that the error was not plausibly relevant to the verdict, it must be found harmless.
¶51 Third, if the error was plausibly relevant to the verdict, the State must establish that the error could not plausibly have been the cause of a guilty verdict from an honest, fair-minded, and reasonable jury. See, e.g., Rice, 120 Wn.2d at 569; Easter, 130 Wn.2d at 242; Watt, 160 Wn.2d at 639-40. This step requires consideration of all relevant untainted evidence and a determination of whether eliminating the error plausibly could have tipped the scales from guilt to innocence. A searching review of the record often will be necessary. See, e.g., State v. Anderson, 171 Wn.2d 764, 770, 254 P.3d 815 (2011). The stronger the plausible sway of the error, the more compelling the relevant untainted evidence must be. See Glasser, 315 U.S. at 67-68. If the error plausibly could have exerted substantial force on the jury in some way, overwhelming evidence to the same effect will be necessary to find the error harmless beyond a reasonable doubt. Some errors are egregious enough that the State will bear a significant burden persuading the court that the error did not contribute to the verdict. See, e.g., State v. Monday, 171 Wn.2d 667, 680-81, 257 P.3d 551 (2011). Once again, these determinations must be informed by logic, common sense, experience, and scientific inquiry. In the end, to find an error harmless, the reviewing court must be convinced, beyond a reasonable doubt, that no honest, fair-minded, and reasonable jury plausibly would have been swayed from innocence to guilt because of the error. If an error was harmless, the conviction should not be disturbed.
¶52 Although it is unlikely that any framework will ever render harmless error analysis entirely objective and con*395sistent, faithful adherence to this three-step process will allow for the development of useful precedent and proper adjudication of harmlessness. This is true especially because at all steps of this process, a reviewing court must provide substantive analysis and reasoning in support of its conclusions. See Clemons, 494 U.S. at 753-54.
4. Application
¶53 Reversal is improper in this case. As the majority properly notes, imposing this instruction was not structural error. Majority at 380 n.l. It was also not plausibly relevant to the guilty verdict and clearly did not tip the scales from innocence to guilt. I am persuaded, beyond a reasonable doubt, that the error was harmless and Coristine’s conviction should be upheld.
¶54 Coristine does not challenge the legal accuracy of the instruction itself. Such a challenge would not succeed; it is drawn from the pattern jury instructions and is based on the statutory language. Clerk’s Papers (CP) at 20; 11 Washington Practice: Washington Pattern Jury Instructions: Criminal 19.03, at 296 (3d ed. 2008). Instructional error is generally subjected to constitutional harmless error review. See Hedgpeth v. Pulido, 555 U.S. 57, 60-61, 129 S. Ct. 530, 172 L. Ed. 2d 388 (2008); State v. Jones, 99 Wn.2d 735, 748-49, 664 P.2d 1216 (1983). Giving an affirmative defense instruction over defendant’s objection does implicate the defendant’s dignity and autonomy. See majority at 375-76. But standing alone, it does not warrant reversal unless, at least, it plausibly might have affected the outcome of his trial.
¶55 In my view, the affirmative defense instruction simply was not plausibly relevant to the jury’s verdict. The decision to give the gratuitous instruction was rendered only after the parties rested, and neither party so much as mentioned the affirmative defense in closing arguments. It could have had no effect on trial strategy or presentation. The instruction was a proper statement of the law. See RCW *3969A.44.030(1). It was not misleading. It cast no doubt on the State’s burden of proving Coristine’s guilt beyond a reasonable doubt.
¶56 I also cannot see how this instruction might have confused the jury. Contra majority at 381. One can always speculate about fanciful ways that an error might have affected the final verdict. In Kroll, for example, we could have speculated about the jury becoming confused because of the erroneous instruction on second degree murder and then also speculated about the possible effects of that theoretical confusion on the first degree murder conviction. But such confusion would not have been plausible. We instruct jurors to consider the greater charges first before reaching the lesser degree. E.g., 11A Washington Practice: Washington Pattern Jury Instructions: Criminal 155.00, at 626 (3d ed. 2008). Since the jurors convicted on the greater charge, they would not have reached the lesser, and any chance of confusion would have been so remote as to be dismissed beyond a reasonable doubt. In contrast, as a result of the forced insanity defense, there was an actual conflict between Jones’s defense and the defense forced upon him, and potentially damaging psychiatric evidence was admitted. Jones, 99 Wn.2d at 748-50.
¶57 Nothing like what happened in Jones happened here. There simply was no conflict between the statutory affirmative defense that “the defendant reasonably believed that [the victim] was not mentally incapacitated,” CP at 20, and the defendant’s theory of the case that the victim was not mentally incapacitated. If the jurors had found that “the defendant reasonably believed that [the victim] was not mentally incapacitated or physically helpless,” “it [was their] duty to return a verdict of not guilty.” CP at 20. If the jurors concluded the state had not “proved beyond a reasonable doubt” that the victim “was incapable of consent by reason of being physically helpless or mentally incapacitated,” it was their “duty to return a verdict of not guilty.” CP at 16. If anything, the affirmative defense instruction *397helped Coristine by giving the jury another way to acquit him. We can say — we should say — that we are confident, beyond a reasonable doubt, that the improper instruction did not contribute to the verdict.
¶58 The actual evidence presented in this case confirms this. The rape took place at Coristine’s residence. His victim had just moved in and was one of many roommates. L.F. testified that she drank heavily at the party and went to bed when her head began to “spin[ ].” 2 Verbatim Report of Proceedings (VRP) at 92. The following day, L.F. went to the hospital and reported to a nurse and police officer that she had been raped while “in and out of a conscious state” while intoxicated in her own bed. 2 VRP at 182. L.F. testified that Coristine subsequently called her at the hospital and told her, “I think it was me, I’m sorry.” Id. at 108. Coristine allegedly explained that he had been “really drunk” and asked L.F. not to tell his wife. Id. L.F. moved out of the house the next day.
¶59 When a detective interviewed him, Coristine admitted he had talked to L.F. on the phone while she was at the hospital, but he denied making self-incriminating statements. At the same time, he told the detective that he was very intoxicated at the party and that he remained unsure of what he had done. Coristine’s DNA (deoxyribonucleic acid) was retrieved from L.F.’s body.
¶60 At trial, Coristine testified that he initially did not remember having sex with L.F. but that “after a couple days [he] did remember everything.” 3 VRP at 358, 380. He testified that when he came upstairs after the party in response to a noise, L.F. pulled him into her bedroom and goaded him into having sex with her. Id. at 354-55 (“I was kind of intoxicated but I still had a level head, but I kind of lost my balance and fell into the room.”). Coristine testified that he did not resist for fear of hurting L.F. and that he did not call out for help because he assumed they were making enough noise in the doorway for someone to hear. Coristine’s sister-in-law testified that she was awake at the time *398and in her bedroom across the hall, but she did not hear any noise from L.F.’s room.
¶61 The record as a whole clearly demonstrates Coristine’s guilt. His version of events was implausible and self-serving. Numerous inconsistencies and farfetched allegations riddled Coristine’s testimony. He acknowledged that L.F. had gone to sleep for the night but then alleged that she later grabbed him and physically forced him into her room. Coristine’s assertion that he could not physically resist L.F.’s advances without injuring her defies common sense. Coristine also alleged he did not call for help because he and L.F. were making a lot of noise — an explanation which is dubious on its own and flatly was contradicted by the testimony of a third party. Beyond a reasonable doubt, the error did not contribute to the verdict.
¶62 There will be real consequences to the court’s decision today. L.F. already had to endure the trial and testify; she should not have to do so again, and she may not be able to do so again. See 2 VHP at 88 (“ [I] t’s something I’ve tried really, really hard to block out as best I can. . . . Last night things came back to me that I didn’t want to come back.”). There is no guarantee that other relevant witnesses will be available or remain able to recall relevant details. In either case, the resources expended at the first trial will have been needlessly wasted. Our constitution does not demand this result.
¶63 I respectfully dissent.
J.M. Johnson, J., concurs with González, J.Myriad examples exist. See, e.g., State v. Clark, 143 Wn.2d 731, 776, 24 P.3d 1006 (2001) (“Clark’s shackling on the first day of voir dire was more than logically offset by over two weeks of observing Clark in the courtroom without shackles.”); State v. Bourgeois, 133 Wn.2d 389, 407-08, 945 P.2d 1120 (1997) (improper communication with juror held harmless because no information was “ ‘communicated... to the jury that was in any manner harmful to the appellant’ ” (quoting State v. Johnson, 56 Wn.2d 700, 709, 355 P.2d 13 (1960))); State v. Handran, 113 Wn.2d 11, 16, 775 P.2d 453 (1989); State v. Caliguri, 99 Wn.2d 501, 509, 664 P.2d 466 (1983) (tape improperly played to jury outside defendant’s presence could not plausibly have been harmful notwithstanding its containing a “colorful expression of [an] agent’s desire to place Caliguri under arrest, a desire which Caliguri’s presence at trial must already have made the jury aware of”); State v. Bonds, 98 Wn.2d 1, 17-18, 653 P.2d 1024 (1982); State v. Hall, 95 Wn.2d 536, 539-40, 627 P.2d 101 (1981) (failure of instruction to discriminate between *393victims irrelevant where only evidence was of indiscriminate firing at all three alleged victims); Kroll, 87 Wn.2d at 840-41 (erroneous instruction related to charge upon which defendant was found not guilty).