In this criminal case, defendant appeals a judgment convicting him of one count of fourth-degree assault, ORS 163.160, of his wife, arguing that the trial court erred in admitting evidence that defendant had previously threatened to kill her. The trial court admitted the evidence on the theory that it was relevant to show that, contrary to defendant’s assertions, he had caused the complainant’s injuries intentionally, not accidentally. On appeal, defendant contends, among other things, that the court erred in admitting the evidence because the state failed to carry its burden of demonstrating, as required by OEC 404(3), that the evidence was logically relevant to show something other than defendant’s bad character. We agree that the court erred in admitting the evidence of the threat, and we conclude that that error was not harmless. Accordingly, we reverse defendant’s conviction and remand for further proceedings.
“‘We evaluate the denial of a defendant’s motion to exclude evidence of other acts in light of the record made before the trial court when it [made its decision].”’ State v. Johnson, 281 Or App 51, 53, 380 P3d 1023 (2016) (quoting State v. Brumbach, 273 Or App 552, 553, 359 P3d 490 (2015), rev den, 359 Or 525 (2016) (citing State v. Pitt, 352 Or 566, 575, 293 P3d 1002 (2012))). Because the state’s request to present evidence of defendant’s prior threat against the complainant, and defendant’s objection to that request, took place during trial, we recount the evidence presented before that request and during the hearing on the request.
One afternoon in December 2011, shortly after the complainant arrived home from work, defendant and the complainant had an altercation at their home. The complainant’s boss, who had dropped her off, heard the altercation and called 9-1-1. A Klamath County Sheriffs Deputy responded, and the complainant told him that defendant had punched her twice in the face. The complainant had an abrasion on her face, and defendant had recent injuries on his knuckles.
In an interview with the deputy, defendant said that he had not punched the complainant. Instead, he asserted that, while he had been angrily pulling his coat out of a closet *163during the argument, the closet door had gotten caught on the coat and the door had bounced off of his own face and then hit the complainant’s face as well. He also said that the complainant had tried to push him out the door, and that, while she was doing that, he put his arm out in front of him and “stiff-armed” her, accidentally hitting her face with an open hand. At trial, the complainant testified that she had tried to push defendant out the door and that her injuries were caused accidentally by the closet door, as defendant had asserted to the deputy. She denied that defendant had punched her.
During the complainant’s testimony, the state sought to present evidence of a threat that defendant had made against the complainant at some point in the past. At a hearing outside the presence of the jury, the complainant testified, without elaboration, that, “a few years ago,” defendant had threatened to take her out into the woods and kill her. The court decided that, in light of defendant’s assertion that he had injured the complainant accidentally, not intentionally, and the suggestion that he might have been acting in self-defense (by “stiff-arming” the complainant while she tried to push him out the door) the evidence was admissible under State v. Harris, 81 Or App 574, 726 P2d 943 (1986), rev den, 302 Or 476 (1987), which we discuss in more detail below. Here, it suffices to note that, in Harris, we applied the test that the Supreme Court established in State v. Johns, 301 Or 535, 549-56, 725 P2d 312 (1986), for admission of certain evidence to prove “intent” under OEC 404(3). In making its ruling, the trial court in this case applied the Johns test and, following our application of the test in Harris, decided that the evidence satisfied its requirements.
After the ruling, the complainant testified before the jury that, sometime in the past, defendant had threatened to take her out into the woods to kill her. The jury found defendant guilty, and he appeals the resulting conviction.
Before setting out the parties’ arguments, we explain the relevant law in some detail. In this case, our analysis is governed by OEC 404(3), which provides:
“Evidence of other crimes, wrongs or acts is not admissible to prove the character of a person in order to show that *164the person acted in conformity 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.”1
In State v. Hampton, 317 Or 251, 855 P2d 621 (1993), the Supreme Court articulated a general test for admissibility of “evidence of other crimes, wrongs or acts” under OEC 404(3):
“‘(1) The evidence must be independently relevant for a noncharacter purpose [such as, in this case, proof of motive]; (2) the proponent of the evidence must offer sufficient proof that the uncharged misconduct was committed and that defendant committed it; and (3) the probative value of the uncharged misconduct evidence must not be substantially outweighed by the dangers or considerations set forth in OEC 403. Each of these requirements must be satisfied before uncharged misconduct evidence is admissible under OEC 404(3).’”
Id. at 254 (quoting State v. Johnson, 313 Or 189, 195, 832 P2d 443 (1992) (alteration in Hampton)); see also State v. Turnidge (S059155), 359 Or 364, 450, 374 P3d 853 (2016) (Turnidge) (citing Hampton in discussion of relevance of motive evidence).
*165In Johns, 301 Or at 549-56, the Supreme Court addressed the relevance and admissibility of evidence of other crimes, wrongs or acts—also known as “prior bad acts”—to prove intent under a doctrine-of-chances theory. The doctrine of chances is premised on the view that “‘an unusual and abnormal element [—accident—] might perhaps be present in one instance, but the oftener similar instances occur with similar results, the less likely is the abnormal element likely to be the true explanation of them.’” Id. at 553 (quoting John Henry Wigmore, 2 Evidence § 302, 246 (Chadbourne Rev. 1979)). The doctrine is “‘the instinctive recognition of that logical process which eliminates the element of innocent intent by multiplying instances of the same result until it is perceived that this element cannot explain them all.’” Id. (quoting Wigmore, 2 Evidence § 302 at 246).
Because the idea underlying the doctrine of chances is that “‘similar results do not usually occur through abnormal causes,”’ id. (quoting Wigmore, 2 Evidence § 302 at 246), “[i]n order for other acts evidence to be logically relevant under the doctrine of chances, the other act must be similar to the charged act.” State v. Tena, 281 Or App 57, 64, 384 P3d 521 (2016).2 Thus, “in Johns, the court formulated a six-factor test for determining whether another act is sufficiently similar to a charged act to be admitted on a doctrine-of-chances theory.”3 Tena, 281 Or App at 64-65.
*166After Johns was decided, both we and the Supreme Court applied the Johns test to a wide variety of evidence offered to show intent under OEC 404(3), sometimes without considering whether the evidence was offered on a doctrine-of-chances theory. See, e.g., State v. Moen, 309 Or 45, 67, 68, 786 P2d 111 (1990) (applying Johns test to evidence of recent threats against the victim that showed the defendant’s “hostile motive, which in turn is probative of [the defendant’s] intent”); Harris, 81 Or App at 578 (in an attempted murder case, applying Johns test to evidence that defendant disliked “whites” and acted violently when he had been drinking).
However, in Turnidge, the Supreme Court recently explained that the Johns test applies only to evidence admitted under a doctrine-of-chances theory. 359 Or at 434-37. The court explained, “Prior bad acts evidence can be relevant to a defendant’s intent on theories other than the doctrine of chances.” Id. at 436. If the theory of relevance on which the evidence is offered does not rely on the doctrine of chances, the evidence is not subject to the Johns test. Id. The court explained that evidence that shows a defendant’s motive—for example, in a homicide case, a recent threat to kill the victim if she dated another man—might be admitted to show that the defendant killed the victim intentionally after he saw her on a date with another man. Id. That type of evidence is not admitted under a doctrine-of-chances theory and, consequently, is not subject to the Johns test. *167Id. at 436-37; see also Tena, 281 Or App at 70 (“Evidence offered to prove intent by showing the defendant’s motive is not subject to the Johns analysis.”); Johnson, 281 Or App at 56 (“ [Fallowing Turnidge, evidence of‘hostile motive’ need not meet the Johns test.”).
Instead of applying the Johns test, the court in Turnidge considered whether the evidence at issue in that part of the opinion—evidence that the defendant had previously conducted a “trial run” of the charged bank bombing by calling in a bomb threat to a nearby bank and watching the law enforcement response—was logically relevant to show that the defendant had formed and executed a plan that included both the trial run and the charged bank bombing. 359 Or at 440-42. The court concluded that the evidence of the “trial run” was logically relevant for that purpose; accordingly, it was admissible under OEC 404(3) to show, by way of the defendant’s plan, that the defendant’s commission of the charged acts was intentional. Id. at 442.4
With that legal context in mind, we turn to the parties’ arguments in this case. In his opening brief, defendant argues that his threat against the complainant was not relevant to his intent in injuring the complainant because it failed the Johns test.5 In response, the state contends that *168the relevance of the evidence does not depend on the Johns test and that, despite its lack of similarity to the charged conduct, the evidence of the threat was logically relevant and admissible under OEC 401 and OEC 404(3) “for a non-propensity purpose to prove the requisite mental state or the absence of mistake or accident.” Defendant responds that the threat evidence was not relevant for any nonpropensity purpose because the threat was too remote in time, and the evidence of it was too undeveloped, to establish a logical link with the charged assault.
As explained below, in light of the Supreme Court’s opinion in Turnidge, the state is correct that OEC 404(3) does not require the evidence at issue here to satisfy the Johns test in order to be admissible. On the record here, however, contrary to the state’s argument, the prior-threat evidence is not relevant to show defendant’s motive for the charged assault. Accordingly, the trial court erred in admitting the evidence under OEC 404(3).
“We review for legal error [a] trial court’s decision to admit evidence of uncharged misconduct under OEC 404(3).” State v. Stubblefield, 279 Or App 483, 490, 380 P3d 1126 (2016) (citing Turnidge, 359 Or at 429). We begin with the trial court’s reasoning. As noted, in making its decision, the trial court relied on Harris. In that case, during a disagreement in a tavern, the defendant “pulled [his] pistol and fired three shots, killing * * * two other men. [The victim, who had been standing with the defendant and the two other men,] ran out of the tavern. Defendant fired at him, hitting the side of the tavern.” Harris, 81 Or App at 576. For his conduct toward the victim, the defendant was charged with attempted murder, and the trial court admitted evidence of the defendant’s prior racially motivated threats and incidents in which he had pointed guns at people and shot a dog “to prove that defendant intended to kill [the victim] when he shot at him.” Harris, id. at 576. We held that the Johns test applied “in determining whether the evidence is relevant to the defendant’s intent.” Id. at 577. After citing *169similarities between the prior incidents and the charged crime—the defendant acted with malice, he was antagonistic toward “whites” (the defendant was Native American and the victim was apparently Caucasian), he had been drinking when he became aggressive, and he had held a gun and acted threateningly in all the incidents—we concluded that “[u]nder Johns, [the disputed] evidence tends to show that, when defendant shot at [the victim], he intended to kill him.” Id. at 578.
The facts of Harris did not allow an inference that the defendant shot at the victim by accident and, accordingly, we did not mention that possibility in our analysis. See Johns, 301 Or at 553 (doctrine of chances draws its persuasive force from the principle that “‘the oftener similar instances occur with similar results, the less likely is the abnormal element likely to be the true explanation of them’” (quoting Wigmore, 2 Evidence § 302 at 246)). Thus, in Harris, we were not evaluating the evidence under a doctrine-of-chances theory of relevancy.
As the Supreme Court explained in Turnidge, the Johns test does not apply to evidence offered to prove intent on a theory other than the doctrine of chances. Turnidge, 359 Or at 436. Accordingly, after Turnidge, our holding in Harris—that the evidence of the defendant’s prior acts was admissible because it satisfied the Johns test—is no longer good law.
In this case, defendant did assert that he had caused the complainant’s injuries by accident: he contended that he had injured the complainant’s face by accidentally hitting it with the closet door and, perhaps, by “stiff-arming” her while she tried to push him out the door. Nevertheless, the court did not admit the evidence of the threat under a doctrine-of-chances theory. As explained above, the doctrine of chances relies on the understanding that, the more often a particular result has occurred because of an abnormal cause, the less likely it is that the abnormal cause is the explanation for another occurrence of the result. Johns, 301 Or at 549-56. Here, the record does not indicate that the result of the charged incident—injury to the complainant or anyone else—had ever occurred before; thus, the court’s *170admission of the evidence did not rest on the understanding that, because defendant had accidentally injured the complainant or someone else before, accident was less likely to be the true explanation for his injury of the complainant on the day in question. Instead, the court relied on our analysis in Harris, which, as we have explained, is incorrect in light of Turnidge. Accordingly, the state is correct that the Johns test does not apply.
Thus, we turn to the state’s argument that the evidence was relevant for a nonpropensity purpose under OEC 401 and OEC 404(3). The state, as the proponent of the evidence, bears the burden of demonstrating relevance and, under OEC 404(3), that the evidence is “probative of something other than disposition to do evil.” State v. Garrett, 350 Or 1, 6, 248 P3d 965 (2011). As explained above, in Hampton, 317 Or at 254, the Supreme Court articulated a general test for admissibility of prior bad acts evidence. The first requirement is that “[t]he evidence must be independently relevant for a noncharacter purpose such as, in this case, proof of motive.” Id. (internal quotation marks and brackets omitted).
The parties dispute whether that requirement is satisfied here. As noted above, the evidence of the prior threat was that, “a few years ago,” defendant had told the complainant that he would take her out into the woods to kill her. The record does not reveal anything else about the context in which the threat was made. It does reveal that defendant and the complainant had known each other since they “were kids going to school.” At the time of the charged assault, they had been married for not quite three years, lived together, and had a three- or four-year-old child. The complainant testified that the argument that led to the charged assault began when she got home from work, when she and defendant “started arguing around, yelling at each other back and forth.” She did not remember what the argument was about; she explained, “It was probably something little and then it just got blown up.” There is no evidence about any previous arguments or that either party had been injured in an argument before.
*171The state contends that “a threat to kill someone gives rise to the following reasonable inference: the person making such a serious threat must have deep-seated hostile feelings for the target of the threat, which feelings are unlikely to dissipate to an insignificant level even over a period of years.” Defendant responds that, in the absence of further development of the record, it is not reasonable “to infer that defendant’s hostility toward [the complainant] persisted over a period of years * * * and that that hostility resurfaced on the day of the alleged assault.”
Thus, the state’s proffered theory of relevance is that the threat evidence is probative of defendant’s intent to injure the complainant by way of a theory of motive: The threat evidence demonstrates that defendant felt hostility toward the complainant a few years before the charged assault, and, in the state’s view, the jury could infer that that hostility continued sufficiently until the day of the charged assault to provide a reason for defendant to have intentionally injured the complainant on the day in question.
“‘Motive is *** a cause or reason that moves the will and induces action, an inducement which leads to or tempts the mind to commit an act.’ State v. Walker, 244 Or 404, 412, 417 P2d 1004 (1966) (citations omitted). Motive is a relevant circumstantial fact that refers to why a defendant did what he did; it generally need not be established by the prosecution to prove guilt. State v. Rose, 311 Or 274, 283, 810 P2d 839 (1991). Evidence of motive is often pertinent as the basis to infer that the act was committed, or to prove the requisite mental state, or to prove the identity of the actor. Wright & Graham, 22 Federal Practice and Procedure 479, & 5240. Of course, ‘courts must be on guard to prevent the motive label from being used to smuggle forbidden evidence of propensity to the jury.’ Id.”
Hampton, 317 Or at 257 n 12 (omission in Hampton; emphasis added).
To determine whether evidence of a prior act is relevant to show the defendant’s motive to commit the charged act, we look to whether the state has showed “‘some substantial connecting link between the two acts.’” Turnidge, 359 Or at 451 quoting State v. Flett, 234 Or 124, 128, 380 P2d 634 *172(1963)).6 “[T]o establish relevancy, the state need not affirmatively prove ‘why’ [the evidence showing a defendant’s motive] made it more likely that he committed the crimes. Instead, the required connection can be inferred when the nature of the evidence at issue, evaluated in light of the circumstances of the crime, makes the inference a logical one.” Id. at 450 (citing Hampton, 317 Or at 258). Evaluating whether the inference is logical requires detailed consideration of the evidence in the record: The court reaches its conclusion “[o]n review of the record, and evaluating the contested evidence in light of the circumstances of the crimes.” Id. at 451.
In Turnidge, one of the questions before the court was whether evidence that the defendant vehemently disliked police, corrections officers, and the government made it more likely that he intended to kill law enforcement officers when he planted a bomb at a bank and made a bomb threat to that bank and another bank. 359 Or at 448-49; id. at 450 (citing OEC 401). To illustrate the proposition that a logical inference must link evidence of motive to the charged crime, the court discussed, among other cases, Flett, in which, the Turnidge court explained, “no connection could be logically inferred between the evidence and the defendant’s motive or intent”:
“In Flett, the defendant was accused of killing her husband, and she argued at trial that his death had occurred during a violent fight. The state introduced evidence that the defendant had told a neighbor several months before that she had had a one-time affair. [The Supreme C]ourt assumed that the state permissibly could use evidence of marital infidelity (particularly, recently occurring *173infidelity) ‘in a proper case/ because it might have ‘some slight probative value/ such as proving motive, particularly where ill will toward the deceased spouse might be an issue. [Flett, 234 Or] at 126-27. The court stated, however, that ‘the connection between isolated acts of marital infidelity and the purposeful slaying of a spouse is extremely tenuous in any case.’ Id. at 127. Turning to the record, the court observed that the defendant and the victim regularly had quarreled and drunk excessively, and ‘no evidence [tended to show] that the hostility, if any, of one spouse toward the other had anything to do with marital fidelity.’ Id. at 127-28. Given those circumstances, the court concluded that the trial court abused its discretion in admitting the evidence, which ‘had nothing to do with the issues that the jury was called upon to decide.’ Id. at 128. In noting further that the state had attempted to ‘blacken the defendant’s character by proof of collateral misconduct having so little to do with the crime as to be virtually irrelevant/ the court commented that the evidence was improper ‘unless the state is prepared to show some substantial connecting link between the two acts.’ Id.”
Turnidge, 359 Or at 451 (third brackets in Turnidge).
In State v. Garrett, as in Flett, the Supreme Court concluded that the record did not support the inference that the state sought to have the jury make. In that case, the defendant was charged with the murder of Williams, and the state sought to admit evidence relating to the defendant’s earlier convictions for assault of Moore, which had been motivated by the defendant’s desire to steal drugs from Moore. 350 Or at 3, 5. In seeking the admission of the evidence of the assault, the state asserted that, “although no large amounts of drugs were found in Williams’s residence, she was known in her neighborhood as a marijuana dealer. *** [I]t is therefore permissible to infer that drugs were present in the Williams residence and that those drugs were stolen when Williams was murdered.” 350 Or at 8-9.
The Supreme Court disagreed. It held that the evidence of the defendant’s assault of Moore was irrelevant because the circumstances of the Williams homicide did not support an inference that burglary or robbery had taken place. 350 Or at 11. The court explained:
*174“There is nothing in the record to establish that a robbery or burglary was committed during the Williams homicide. Instead, the record shows that nothing was missing from Williams’s residence and that Williams’s checkbook, cash, and jewelry were undisturbed. The state’s assertion that Williams’s marijuana was stolen is based on a detective’s testimony that it is common for drug dealers to possess large quantities of marijuana and that detectives were surprised that only a small amount of marijuana was found in Williams’s residence. The detective’s experience that drug dealers usually have large quantities of drugs on hand is insufficient in this context to give rise to a permissible inference that the person or persons who murdered Williams did so in the course of burglarizing and robbing Williams of a large quantity of marijuana. In light of the evidence of sexual assault [of Williams] and the absence of any evidence of robbery or burglary, the state’s claim that a robbery or burglary took place in furtherance of the Williams homicide is mere conjecture.”
Garrett, 350 Or at 9.
Flett and Garrett demonstrate that the mere possibility that the same motive that caused an earlier crime or act also caused the charged crime is not enough to make evidence of the prior act relevant. Instead, something in the circumstances of the charged crime must suggest that the identified motive is at work. Said another way, there must be a logical match between the motive and the specific circumstances of the charged crime, or, as the Supreme Court put it in Flett, a “substantial connecting link” between the two acts. 234 Or at 128; see also Turnidge, 359 Or at 450 (state must establish “some sort of logical connection” between purported motive and the charged crime). Evidence of a defendant’s motive for committing a prior act is not admissible to prove the defendant’s motive in committing a later act when the possibility that the same motive caused both acts is, as in Garrett, “mere conjecture.” 350 Or at 9.
In Flett, as described above, the state contended that the defendant’s one-time affair demonstrated ill will toward the victim (the defendant’s husband) and that ill will could have caused her to kill him during an argument. 234 Or at 127. The court did not consider that proposition in the abstract. Instead, it turned to the circumstances of *175the killing for confirmation of the possible motive, but it found none: “[T]he defendant and the victim regularly had quarreled and drunk excessively, and ‘no evidence [tended to show] that the hostility, if any, of one spouse toward the other had anything to do with marital fidelity.’ [Flett, 234 Or] at 127-28.” Turnidge, 359 Or at 451 (second brackets in Turnidge). Although it was possible, in the abstract, that ill will arising from recent infidelity might provide a motive for killing a spouse, see Flett, 234 Or at 126-27, the record in Flett did not allow a reasonable inference that it had done so in that case.
Similarly, in Garrett, the record did not support the inference that robbery or burglary might be the motive for the charged murder. Although a detective had testified that the victim was known as a marijuana dealer and that investigators were surprised that she did not have more marijuana in her house, the circumstances of the crime did not support an inference that she had been robbed. 350 Or at 9. On the contrary, the victim had been sexually assaulted, and there was no evidence of robbery or burglary. Id. Because “the state’s claim that a robbery or burglary took place * * * [was] mere conjecture,” id., there was no substantial link between the proffered motive and the crime. Thus, evidence that, on another occasion, the defendant had assaulted someone in order to steal drugs was irrelevant. Id. at 11.
In State v. Clarke, 279 Or App 373, 379 P3d 674 (2016), we applied that principle in a case where the circumstances amply indicated that the same motive caused the prior act and the charged crime. There, the state sought to present evidence that the defendant had contemplated killing his ex-girlfriend, Sisson, to show his motive for killing the victim, who was the defendant’s roommate. The victim and Sisson had recently begun a romantic relationship, and the state presented evidence that the defendant was upset about his breakup with Sisson, that he was angry with the victim for bringing Sisson to their apartment, and that the victim had told the defendant that he (the victim) had been intimate with Sisson. Id. at 375. That evidence supported the inference that the “defendant spent his time contemplating whether to kill Sisson because he was angry about the end of their relationship.” Id. at 389. The same evidence, *176as well as the timing—defendant’s anger with the victim, defendant’s contemplation of killing Sisson, other threats by the defendant against the victim, and the murder all took place within a few weeks, id. at 377—allowed the inference that the defendant “would have been moved by that same anger to kill the victim due to his jealously of the victim’s relationship with Sisson.” Id. at 389. Thus, in Clarke, the circumstances demonstrated a “substantial connecting link” between the prior act and the charged crime.
Finally, we turn to the arguments and facts before us. As noted above, the state contends that, once one person has threatened to kill another, it is reasonable to infer that, at any time in the future, the hostility demonstrated by the threat persists and provides a motive for injuring the other. That assertion is similar to the state’s assertion in Flett that an act of infidelity is always relevant in a case involving the murder of a spouse. It is possible that the ill will demonstrated by a prior infidelity—or, in this case, the hostility demonstrated by a prior threat—could motivate a later assault or murder. As Flett and Garrett demonstrate, however, that possibility is not enough to make the purported motive evidence relevant. Instead, courts must evaluate a past act to determine, first, what motive the prior threat demonstrates, and, second, whether the circumstances of the charged crime support an inference that the same motive is at work.
Here, regarding the prior act, the record reveals only that, a few years before the charged conduct, defendant threatened to kill the complainant. We agree with the state that the threat shows that, when it was made, defendant was angry with, and hostile toward, the victim. However, on this record, we have no way of evaluating whether a substantial link exists between that hostility and the charged assault. That is, the record lacks the information necessary to determine whether the same hostility might have been at work in both incidents.
The record contains no evidence about the context in which the threat was made. It lacks information about what, if anything, preceded or led to the threat and whether the threat was the type of threat that, in another context, *177has been described as “hyperbole,” “rhetorical excess!],” or an “impotent expression!] of anger or frustration.” State v. Rangel, 328 Or 294, 303, 977 P2d 379 (1999) (internal quotation marks omitted). In short, it does not contain evidence from which a nonspeculative inference could be drawn regarding whether the motive for the threat—hostility—was likely to persist until, or recur on, the date of the charged crime and motivate the commission of that crime.
The threat does not suggest any motive more specific than generalized hostility. Cf., e.g., Turnidge, 359 Or at 436 (using, as an example, the jealousy of victim dating another man as providing the necessary link between a recent threat and a charged crime by the jealous actor). Moreover, nothing in the circumstances of the charged assault suggests that a particular type of hostility was at work. Cf., e.g., id. at 452-53 (defendant’s acts of planting a bomb and calling in a threat that would bring law enforcement to the scene allowed inference that motive was to injure or kill law enforcement officers; thus, evidence of defendant’s hatred of law enforcement was relevant).
Furthermore, to the extent that generalized hostility can provide the necessary link between two incidents, given the lapse of “a few years” between the threat and the charged assault, it is not reasonable to infer, on this record, that defendant held a general hostility toward the complainant that could have persisted unabated and caused the charged assault. Cf., e.g., State v. Salas-Juarez, 349 Or 419, 427, 245 P3d 113 (2010) (evidence that Russell had wanted to “slash” someone at a bar earlier on the evening of the charged stabbing, combined with evidence that he was still angry later in the evening, was relevant to Russell’s motive for stabbing the victim); Clarke, 279 Or App at 389 (contemplation of killing ex-girlfriend relevant to motive for killing roommate a month later based on roommate’s ongoing romantic relationship with ex-girlfriend); Moen, 309 Or at 65, 69 (threats against victim and family member made three weeks before murder relevant to show defendant’s “hostile motive” toward victim).
Given the lack of information about the circumstances surrounding the threat, the lack of any suggestion *178of a particular motive in the circumstances of the assault, and the lack of support for an inference that defendant’s general hostility continued over the intervening “few years,” the record does not support an inference that the hostility that caused the threat also motivated the charged assault. Accordingly, the state did not carry its burden of showing that the evidence of the threat was relevant, and the court erred in admitting the evidence.
We cannot say that the error was harmless. See State v. Davis, 336 Or 19, 32, 77 P3d 1111 (2003) (“Oregon’s constitutional test for affirmance despite error consists of a single inquiry: Is there little likelihood that the particular error affected the verdict?”). The threat—in particular, its wording—was highly inflammatory, and, as such, created a risk that the jury would convict defendant based on his past acts or his character. Moreover, it was susceptible to the chain of inferences, forbidden under OEC 404(3), that defendant was likely to have assaulted the victim on the day in question because he had a propensity toward domestic abuse. Accordingly, we cannot say that the improper admission of the threat evidence had little likelihood of affecting the verdict. See id.
Reversed and remanded.
In State v. Williams, 357 Or 1, 15, 346 P3d 455 (2015), the Supreme Court held that OEC 404(4) “supersedes” OEC 404(3). In several recent cases, however, we have held that “‘the types of relevant evidence set out in OEC 404(3) *** remain viable theories for admission of prior acts evidence.’” State v. Tena, 281 Or App 57, 63, 384 P3d 521 (2016) (quoting State v. Clarke, 279 Or App 373, 382, 379 P3d 674 (2016)). “Thus, in evaluating whether evidence of‘other crimes, wrongs or acts’ is admissible for nonpropensity purposes, we may draw on the ‘settled principles’ of relevance embodied in OEC 404(3) and case law construing that provision.” Clarke, 279 Or App at 382 (quoting State v. Turnidge (S059155), 359 Or 364, 434, 374 P3d 853 (2016)).
Here, because the state has not argued, below or on appeal, that the evidence is relevant and admissible for propensity purposes under OEC 404(4), we need not decide that question. See State v. Hudman, 279 Or App 180, 187, 379 P3d 659 (2016) (where trial court admitted prior bad acts evidence under OEC 404(3) and state did not advance theory of admissibility under OEC 404(4) as an alternative basis for affirmance, “our task on appeal is limited to determining whether the trial court erred in admitting [the disputed evidence] under OEC 404(3)”); State v. Davis, 279 Or App 223, 233, 381 P3d 888 (2016) (where state advanced no propensity argument either before the trial court or on appeal, “we confine our analysis to the questions presented in defendant’s opening brief: whether the court erred under OEC 404(3) in admitting the [disputed] evidence and whether any error harmed defendant”).
As we observed in Tena, 281 Or App at 64 n 4:
“Because the doctrine of chances is premised on the idea that it is unlikely that multiple instances of similar conduct will be the result of an innocent intent, evidence is logically relevant under the doctrine only when the other acts involve innocent intents. In other words, because the doctrine is based on the idea that recurring similar mistakes or accidents are increasingly unlikely, the doctrine supports the admission of other acts evidence only when the other acts were, or are claimed to have been, the product of a mistake or accident.”
Specifically, in Johns, the Supreme Court stated:
“[I]n evaluating prior crime evidence on the issue of intent or absence of mistake, the trial judge should make these determinations:
“(1) Does the present charged act require proof of intent?
“(2) Did the prior act require intent?
“(3) Was the victim in the prior act the same victim or in the same class as the victim in the present case?
“(4) Was the type of prior act the same or similar to the acts involved in the charged crime?
*166“(5) Were the physical elements of the prior act and the present act similar?
“(6) If these criteria are met, is the probative value of the prior act evidence substantially outweighed by the danger of unfair prejudice, confusion of issues or misleading the jury, undue delay or presentation of cumulative evidence?”
301 Or at 555-56.
Following the enactment of OEC 404(4) in 1997, we held that evidence of other acts by a criminal defendant could be admitted without applying the sixth Johns factor; that is, it could be admitted without balancing the probative value of the evidence against the danger of unfair prejudice. See, e.g., State v. McIntyre, 252 Or App 16, 27 n 10, 284 P3d 1284 (2012) (stating that the enactment of OEC 404(4) “effectively removed [the sixth] factor from the Johns analysis”). However, in the Supreme Court’s recent decision in Turnidge, 359 Or at 442, the court ruled, “if a trial court determines that prior bad acts evidence is relevant to a nonpropensity purpose under OEC 404(3), the court, on a proper motion, must weigh the probative value of the evidence against its potential to unduly prejudice the defendant.”
Although the court did not cite Hampton in that section of its analysis, it evaluated the logical relevance of the disputed “plan” evidence; in doing so, it was applying the first requirement of the Hampton test—that the evidence be independently relevant for a noncharacter purpose. See Turnidge, 359 Or at 442 (discussing logical relevance of the plan evidence); Hampton, 317 Or at 254 (to be admissible under OEC 404(3), “evidence must be independently relevant for a noncharacter purpose”); see also Turnidge, 359 Or at 450 (in evaluating a question of relevance under OEC 401, citing Hampton’s discussion of relevance of motive evidence under OEC 404(3)).
We reject the state’s argument that that argument was not preserved because defendant did not raise it to the trial court until after the court admitted the evidence. The trial court made its ruling, which, as we have explained, was based on Harris, without giving defendant an opportunity to object. Defendant objected when the victim testified to the threat, and he raised the Johns issue at the next available opportunity, noting that the court had not given him an opportunity to raise it earlier. The court agreed and addressed defendant’s argument, adhering to its determination that the evidence was admissible under Harris. Under those circumstances, defendant was not required to raise his argument before the court ruled. See Peeples v. Lampert, 345 Or 209, 220, 191 P3d 637 (2008) (preservation not required when “a party has no practical ability to raise an issue”). Moreover, the purposes of preservation were served by defendant’s later objection. See id. at 219-20 (rule of preservation allows trial courts “to *168consider and rule on a contention, thereby possibly avoiding an error altogether or correcting one already made,” ensures fairness to an opposing party, and fosters full development of the record).
In Turnidge, the court explained how to determine the relevance of motive evidence to prove intent under OEC 401. The same principles apply to determine, under OEC 404(3), whether prior bad acts evidence is relevant to prove motive and thereby to prove intent, with additional focus on the requirement that the “chain of logical relevance” connecting the disputed evidence to the charged crime must “not ultimately rely on an inference relating to [the] defendant’s character or propensity.” Garrett, 350 Or at 9 (describing holding of State v. Johnson, 340 Or 319, 338, 131 P3d 173 (2006), regarding OEC 404(3)); see also State v. Pratt, 309 Or 205, 210, 785 P2d 350 (1990) (“[O]ther crime evidence is admissible * * * upon ‘any theory of logical relevance * * * that does not run afoul of the “propensity to commit crimes or other acts” prohibition.’” (Quoting Johns, 301 Or at 548 (second omission in Pratt))).