The interview room had two doors and windows facing the street. It was unlocked, and no security code was needed to enter or leave. It contained a small round table that the officers and defendant sat around; defendant sat in the chair closest to the door to the lobby, Steigleder sat to his right, and Stowe sat to the right of Steigleder. Neither officer blocked defendant's access to the exit door. The room did not have any cameras or police equipment in it, and the temperature was comfortable.
The interview began with Stowe asking defendant to tell his side of the story.2 Defendant was not advised of his Miranda rights *438then or at any point during the interview. The officers never raised their voices, and there was no show of physical force or indication that the officers might engage in physical force.
Defendant first said that he was installing some cable devices in the complainant's home and, while he was in a crouched position, his belt broke and his pants fell down. The complainant came into the room while his pants were still down. He said that his penis was not exposed. He showed the officers the belt, which he had brought with him. It looked like it had been cut rather than torn from use.3 Defendant stood up at some point to show the officers how his pants were fitting; the officers did not tell him to sit down. Otherwise, the officers and defendant remained seated during the interview.
*176Stowe pointed out to defendant that his version of the incident was inconsistent with the complainant's version. The second or third time that defendant told his story, he said that his penis may have been exposed a little through a hole in his underwear. When Stowe told defendant that that was also different from the complainant's account, defendant said that the complainant was lying. Each time that defendant told the story, he added details that he had not included before. Stowe was "unconvinced" by defendant's account of events.
Steigleder, who did the majority of the questioning, told defendant that she did not think his story made sense and communicated that she did not believe he was telling the truth. At the suppression hearing, she testified:
"I would point out the inconsistency [with the complainant's statement] and give him another chance to tell me the story, and he told the story and he changed it a little, and then I'd point out the next inconsistency and he would change it again. And we just kind of-it was kind of I let him tell the story and then I would point out why it didn't make sense and then said, you know, I don't think that's quite the way it happened, you know, is there something else. And then you know, he would change the story."
According to Steigleder, defendant changed his version of events several times to try to make it line up with the complainant's: First, as noted, defendant said that his belt broke and his pants fell down, but that his penis was not exposed; next, he said that maybe the button on his boxers had been undone and slightly exposed his penis; then, he said that he had intentionally dropped his pants to ventilate his testicles.
At that point, Steigleder "changed tactics" and asked defendant whether he watched pornography and how often. Defendant admitted that he watched pornography a couple of times a week with his wife and he also watched by himself. Steigleder then told defendant that she believed he had been aroused by the fact that the complainant was in the house and he might get caught:
"I said what I think happens, this is-you know, you get these urges or you have these fantasies and you act on it and I-you know, I think I even told him, you know, I'm *177sure maybe you didn't really mean for her to come in, it's just the fact that she was there and that you could get caught is where that arousal is coming from."
She also said that she believed he had a sexual addiction problem and that he should get counseling. Steigleder then said, "tell me what really happened," and, at that point, defendant admitted that he had gotten aroused, exposed his penis, and started to play with himself when the complainant walked into the room. The interview concluded without defendant being cited and defendant left on his own.
The interview lasted "at the most" one and one-half hours. Steigleder testified that he had told defendant "[p]robably right about * * * three" times that defendant was free to leave at any time, that he had made those statements throughout the interview, and that defendant appeared to have understood what he had been told. Defendant never asked to stop the interview, leave, talk with anyone else, or take a break. The officers did *439not tell defendant that they knew he had committed a crime or that they could find out the truth through other means. They also did not tell defendant that they planned to arrest him and, in fact, they testified that they did not intend to do that.
After he was charged with private indecency, defendant moved to suppress the statements he had made during the interview, arguing that suppression was required under Article I, section 12, of the Oregon Constitution because they were obtained while he was under compelling circumstances and he was not given Miranda warnings before being questioned.4 The trial court denied suppression, concluding that the circumstances of the interview were not compelling and, therefore, Miranda warnings were not required.
The court reasoned that defendant drove to the police station voluntarily and "clearly was eager to go to present his side of the story," and "the officers were very clear that they told him that he was free to leave." Although the court found that there was "some pressure" on defendant, particularly at the end of the interview, in the totality *178of the circumstances, that pressure did not, in the court's view, "outweigh[ ] the fact that he was there voluntarily, that he was there making up stories voluntarily."
The case was then tried to the court, and the court found defendant guilty of private indecency. Defendant appeals the judgment of conviction, assigning error to the trial court's denial of his suppression motion.
Article I, section 12, which protects a person's right against compelled self-incrimination,5 requires Miranda warnings before the police interrogate a person "who is in full custody or in circumstances that create a setting which judges would and officers should recognize to be compelling." State v. Roble-Baker , 340 Or. 631, 638, 136 P.3d 22 (2006) (internal quotation marks omitted). If an officer fails to give the requisite Miranda warnings in either of those situations, we must suppress the statements made in direct response to the unwarned questioning. State v. D. P. , 259 Or. App. 252, 262, 313 P.3d 306 (2013) (citing State v. Jarnagin , 351 Or. 703, 713, 277 P.3d 535 (2012) ). We must also suppress "evidence that derives from or is a product of that constitutional violation." Jarnagin , 351 Or. at 713, 277 P.3d 535.
There is no dispute in this case that defendant was not in full custody; the issue therefore is whether his encounter with the police evolved into a setting that was compelling. In assessing that question, we consider the totality of the circumstances, including "(1) the location of the encounter; (2) the length of the encounter; (3) the amount of pressure exerted on the defendant; and (4) the defendant's ability to terminate the encounter." Roble-Baker , 340 Or. at 640-41, 136 P.3d 22 (internal citations omitted). Those factors are non-exclusive, and we do not apply them "mechanically"; rather, our "overarching inquiry is whether the officers created the sort of police-dominated atmosphere that Miranda warnings were intended to counteract." Id. at 641, 136 P.3d 22. The answer to that question "turns on how a reasonable person in the suspect's position would have understood his or her situation." Shaff , 343 Or. at 645, 175 P.3d 454 ; see also State v. Courville , 276 Or. App. 672, 677, 368 P.3d 838 (2016) (" 'Compelling circumstances'
*179exist when the circumstances are such that an objectively reasonable person would feel compelled to answer the officer's questions.").
Once a defendant has raised the legal issue, it is "the state's burden to show that defendant's unwarned statements were made before the circumstances became compelling," Roble-Baker , 340 Or. at 639, 136 P.3d 22 ; in other words, the state "must show that defendant's unwarned statements were not made while she was in compelling circumstances, and the reviewing court must determine the point at which, if ever, the circumstances became compelling," id . at 640, 136 P.3d 22 (emphasis added). Whether *440the totality of the circumstances, as found by the trial court, were compelling, such that Miranda warnings were required, is a legal question that we review for legal error. State v. Northcutt , 246 Or. App. 239, 245, 268 P.3d 154 (2011).
We begin by observing that defendant does not identify the specific statements he seeks to have suppressed. However, we understand him to argue that he was in compelling circumstances at least by the time that he admitted to the conduct that formed the basis for the charge-that is, when he stated that he had become aroused and pulled his pants down and exposed his penis in the complainant's home-and therefore those and subsequent statements must be suppressed because they were unwarned.6 As explained below, we conclude that, when Steigleder asked defendant about his use of pornography, the circumstances had evolved to the point where a reasonable person would have felt compelled to answer the officers' questions and Miranda warnings were therefore required. Consequently, defendant's subsequent statements, including his admission, should have been suppressed, and the trial court erred in concluding otherwise.
*180We acknowledge that this is a close case. With regard to the first factor, the location of the encounter, the unfamiliar, police-station setting of the interview tended-necessarily-toward a "police-dominated atmosphere." Cf. Shaff , 343 Or. at 646, 175 P.3d 454 ("[T]he fact that the interview occurs in familiar surroundings diminishes the police-dominated atmosphere that Miranda warnings were intended to counteract."); State v. Schwerbel , 233 Or. App. 391, 397, 226 P.3d 100, rev. den. , 349 Or. 172, 243 P.3d 70 (2010) (public location of police encounter made circumstances less compelling). However, that effect was lessened somewhat by the fact that defendant came to the station voluntarily, on his own power and at a time of his own choosing. See, e.g. , State v. Barber , 179 Or. App. 674, 679, 41 P.3d 455, rev. den. , 334 Or. 632, 54 P.3d 1042 (2002) (no compelling circumstances where the defendant agreed to voluntarily come to the police station for questioning, made an appointment and then arrived an hour late, and never indicated that he wanted to leave).
Likewise, the second factor, the length of the interview, would not, considered in isolation, compel a conclusion that the circumstances here were compelling. The trial court found that the encounter lasted, at most, one and one-half hours. That is not an insignificant amount of time to undergo questioning by police officers; however, neither is it determinative of compelling circumstances. See Roble-Baker , 340 Or. at 643, 136 P.3d 22 (implicitly concluding that over five hours of questioning in police station not dispositive); Northcutt , 246 Or. App. at 249-50, 268 P.3d 154 (one and one-half hours of questioning, although longer than many police-citizen encounters, not a dispositive factor given Roble-Baker ); see also State v. Saunders , 221 Or. App. 116, 119-20, 188 P.3d 449, rev. den. , 345 Or. 416, 197 P.3d 1105 (2008) (one and one-half hour interview in defendant's home did not constitute compelling circumstances). Rather, as we explained in Northcutt , "any consideration of the durational factor is necessarily dependant on the character or quality of the interaction"; thus, "the principle emphasis is properly on the qualitative dynamics addressed in the third and fourth of the Roble-Baker factors." 246 Or. App. at 250, 268 P.3d 154.
Thus, we turn to those factors and consider the length and location of the interview in light of the "qualitative *181dynamics" addressed there. The third factor-the amount of pressure exerted on the defendant during the encounter-is fairly amorphous. Often, the pressure comes from the police confronting the defendant with evidence of his or her *441guilt. However, "what matters is not whether evidence of guilt was apparent to the suspect; rather, it is whether the officers used that evidence in a coercive manner." Shaff , 343 Or. at 650, 175 P.3d 454 (emphasis added). In Shaff , the court found that the officer's questions and comments "were not coercive, aggressive, or repetitive," although they included asking the defendant whether his argument with a woman in the house had become physical, commenting on the woman's evident injuries, and giving the defendant false information that the woman had reported an assault.7 Id. (emphasis added). Thus, we have understood Shaff to stand for the proposition that "mere reference to evidence of a suspect's guilt in the course of nonaggressive questioning that does not assume the suspect's guilt is not sufficiently coercive to give rise to compelling circumstances." Northcutt , 246 Or. App. at 249, 268 P.3d 154 (internal quotation marks omitted). Rather, "[a]s illuminated in Shaff , the gravamen of the third Roble-Baker factor is the use of aggressive and coercive police interrogation practices, especially included, but not limited to, those explicitly predicated on assumptions of a suspect's guilt or calculated to contradict a suspect's assertions of innocence." Northcutt , 246 Or. App. at 250, 268 P.3d 154.
We applied Shaff in Saunders , which the state invokes as support for its position that the pressure exerted on defendant in this case was "minimal" and "does not control the analysis." In Saunders , we had originally concluded that confronting the defendant with the child victim's allegations of sexual abuse, as well as the "highly incriminating evidence" of her hand-drawn pictures, contributed to a finding of compelling circumstances. State v. Saunders , 211 Or. App. 73, 82, 153 P.3d 144 (2007), vacated and remanded , 344 Or. 277, 179 P.3d 671 (2008). However, given the Supreme *182Court's "clarification" in Shaff regarding the effect of confronting a defendant with evidence of his or her guilt, we reversed course on remand, stating simply, "There is no evidence in the record of the suppression hearing that [the officers] used the victim's allegations or the drawing in a coercive manner ." Saunders , 221 Or. App. at 120, 188 P.3d 449 (emphasis added); see also Courville , 276 Or. App. at 678-79, 368 P.3d 838 (concluding that, although the officer's manner of questioning-encouraging the defendant to talk when he indicated a desire not to answer and making statements that assumed the defendant's guilt-exerted pressure on the defendant to speak, the circumstances were not meaningfully distinguishable from Saunders ).
In contrast to Saunders , however, as defendant points out, in State v. Machain , 233 Or. App. 65, 225 P.3d 75 (2009), we found that the circumstances were compelling where the pressure exerted on the defendant escalated during the approximately two and one-half hour interview at a sheriff's office. We explained that, instead of the "brief[ ] and noncoercive [ ]" use of evidence of guilt evident in Shaff and Saunders , in Machain , the detectives
"repeatedly told defendant that they would be able to disprove any falsehood, that they knew that she was lying, and that she needed to tell the truth and 'be honest now.' The detectives, who had told defendant that they would be able to figure out 'everything' and that 'we know what happened now,' repeatedly asked questions that assumed defendant's guilt."
233 Or. App. at 75-76, 225 P.3d 75. Under those circumstances, we concluded, "a reasonable person in defendant's position would have understood herself to be compelled to answer the detectives' questions during the interview." Id. at 76, 225 P.3d 75. Similarly, in State v. Ford , 244 Or. App. 289, 293, 296, 260 P.3d 637, adh'd to as modified on recons , 245 Or. App. 500, 263 P.3d 1110 (2011), we held that the circumstances were compelling where, among other circumstances, the officer told the defendant five or six times over the course of an hour that they knew what was going on because they had talked to the alleged victim and the defendant needed to *442" 'tell me exactly *183what happened.' "8 See also State v. Heise-Fay , 274 Or. App. 196, 207, 360 P.3d 615 (2015) (coercive practices contributed to compelling circumstances where reasonable person in defendant's position would have understood that she would be subject to arrest for hindering prosecution if she did not cooperate); State v. Mattheisen , 273 Or. App. 641, 651, 359 P.3d 1218 (2015) (where officers persisted in confronting the defendant with evidence of his guilt, despite his repeated denials of wrongdoing, and "repeatedly leveraged inadmissible polygraph results as dispositive evidence of his guilt," the interrogation practices were coercive).
While perhaps not as obviously intimidating as the police practices in Machain , Heise-Fay , and Mattheisen -where the officers also intimated that they had other, corroborating evidence of the defendants' guilt in addition to the victims' allegations, or made implicit threats-we agree with defendant that here, too, the interrogation practices used by the police ultimately became coercive.
At the outset of the interview, the officers simply confronted defendant with incriminating evidence-the complainant's version of events-in a noncoercive and nonaggressive manner and asked defendant for his side of the story. However, after he answered, they persisted, escalating the pressure on defendant to continue talking at every stage by telling him that his explanations did not make sense in light of the complainant's statement, indicating that they did not believe him, and repeatedly asking him to explain the inconsistencies. After about the third time that the officers asked defendant to explain the discrepancies between his story and the complainant's, the questioning became more intrusive. At that point, Steigleder switched strategies and began talking to defendant about his sexual habits. She asked defendant if he watched pornography and when he said that he did, Steigleder expressed her belief that defendant was aroused by the prospect of being in proximity to the complainant and getting caught, that he had a sexual addiction problem, and that he needed counseling.
*184Steigleder then said, "tell me what really happened." Those questions and remarks implicitly assumed that defendant was guilty of the charged behavior.
At that juncture, the police tactics went beyond the mere brief and noncoercive use of incriminating evidence at issue in Shaff and instead were calculated to contradict defendant's repeated assertions of innocence and pressure him to continue talking. As in Ford , "the encounter was prolonged and the officer persistently pressured defendant for more information in ways that assumed defendant's guilt." 244 Or. App. at 297, 260 P.3d 637. And, at that point, defendant had been questioned by the officers for nearly one and one-half hours, at the police station.9 Thus, considered together, the first three Roble-Baker factors tend toward a conclusion that the circumstances had evolved into a setting in which Miranda warnings were required when the officer began talking to defendant about his sexual habits-that is, right before he confessed.
The question remains what to make of the fourth Roble-Baker factor-defendant's ability to terminate the encounter. As the state emphasizes, defendant was told that he was free to leave. Moreover, there were no practical barriers to defendant terminating the interview. Defendant had arrived at the police station on his own volition, and he was not dependent on the officers for leaving. The interview was in an unsecured meeting room, near the front lobby, and defendant had access to the exit. The officers did not try to control defendant's movements by their words, tone, or body positioning. Nor did defendant make any requests to stop the interview or take a break that the officers disregarded. Cf. Roble-Baker , 340 Or. at 642-43, 136 P.3d 22 (although the defendant was told that " 'she'd always been free to leave' " the police head-quarters, her requests to suspend the interview were twice disregarded, *443and she was, for all practical purposes, required to remain because she was dependent on the officers for transportation and she was told that her son was being interviewed at his school and would be brought to the station).
However, the fact that a defendant is free to leave does not necessarily control the outcome.
*185Mattheisen , 273 Or. App. at 652, 359 P.3d 1218. Similar to this case, in Mattheisen , the defendant had traveled voluntarily to the interview (the office of a private polygraph examiner), the door remained partially open during part of the encounter, the officer had initially advised the defendant that he was free to leave, and the defendant had agreed, in response to an inquiry just before his arrest, that he felt free to leave. Thus, the record supported the trial court's findings that the defendant " 'was not prevented from leaving' " and "would have been able to end the interview." Id. Nonetheless, we concluded that "this one of four factors does not overcome the otherwise compelling circumstances" presented by the other factors. Id. (footnote omitted). Instead, "[i]n the totality of the circumstances, the exertion of this pressure on the defendant through several interviews in unfamiliar surroundings with the seemingly objective confirmation of the polygraph" constituted compelling circumstances, notwithstanding that defendant was free to end the encounter. Id .
We reach a similar conclusion here. Although the record supports the trial court's finding that defendant was told that he was free to leave, the record does not disclose when, in the context of the escalating nature of the interview, defendant was so instructed. That is, we do not know whether defendant received that information anywhere close in time to the point where the questioning intensified and became coercive. In her testimony, in response to a question from the prosecutor, Steigleder indicated only that the instructions were given "throughout" the interview, and the prosecutor did not attempt to pinpoint it further. Given the relatively long length of the interview, that omission is significant; consequently, in our view, while the fourth factor favors the state, it does so only slightly. And, as in Mattheisen , we conclude that it does not overcome the otherwise compelling circumstances presented by the combination of the location of the interview, its length, and, especially, the amount of pressure exerted on defendant during the interview.
In sum, considering all of the circumstances, we conclude that the state failed to carry its burden that defendant was not in compelling circumstances at the point at which defendant admitted to the behavior that formed the basis for *186the charge-that is, when he admitted that he had become aroused and pulled his pants down and exposed his penis in the complainant's home. Thus, the court erred in failing to suppress those unwarned statements and the statements that followed. Moreover, that error was not harmless. See State v. Davis , 336 Or. 19, 32, 77 P.3d 1111 (2003) (we will affirm a judgment of conviction notwithstanding evidentiary error if the error is harmless, that is, if "there was little likelihood that the error affected the jury's verdict"). Here, we cannot conclude that the error in admitting defendant's statements, which corroborated the complainant's account and provided the state's strongest evidence of defendant's guilt, had "little likelihood" of affecting the court's verdict. The state does not contend otherwise. Accordingly, we reverse and remand.
Reversed and remanded.