UNITED STATES v. GOLDESBERRY (2025)

Appellant Raymond Lee Goldesberry appeals his conviction after a jury trial for aggravated sexual abuse of a minor under 12 in Indian Country in violation of 18 U.S.C. §§ 1151, 1153, and 2241(c). He urges reversal on two grounds. First, Mr. Goldesberry contends the evidence was legally insufficient to sustain the verdict. Second, he argues the government committed prosecutorial misconduct in its closing argument. Exercising jurisdiction under 28 U.S.C. § 1291, we conclude the government's evidence was not sufficient to establish beyond a reasonable doubt the knowledge element of the charged offense, so we must vacate Mr. Goldesberry's conviction.¹

I 2

This case began one night in 2017, when Mr. Goldesberry's daughter K.G.—then almost 12 years old—entered her parents’ bedroom after having a nightmare. The room was completely dark, and Mr. Goldesberry was asleep in bed. Mr. Goldesberry and K.G.’s mother Michelle Goldesberry usually slept in the same bed, but that night, Mrs. Goldesberry was not there. K.G. climbed into bed next to her father and fell asleep. In the night, K.G. awoke to Mr. Goldesberry touching her “[u]nder [her] underwear.” App. II at 138. About four years later, the government prosecuted Mr. Goldesberry based on this conduct.

In October 2021, Mr. Goldesberry was charged in the United States District Court for the Northern District of Oklahoma with one count of aggravated sexual abuse of a minor in Indian Country, in violation of 18 U.S.C. §§ 1151, 1153, and 2241(c). The indictment alleged the offense occurred “[b]etween in or about May 2017 and on or about September 31, 2017,” and involved K.G., “a child ․ who had not attained the age of 12 years.” App. I at 13.³ Mr. Goldesberry entered a not-guilty plea. In March 2022, the case proceeded to jury trial.⁴

Mr. Goldesberry maintained his innocence. He never denied a sexual act occurred but insisted what happened was an accident. According to Mr. Goldesberry, the evidence would show “during the course of the time that he was sleeping,” he did not know “his wife ․ had gotten out of bed and left the house, and in her place, his daughter, [K.G.], was l[y]ing where [his] wife” usually would be sleeping. App. II at 107.

The trial lasted three days. We describe the trial evidence in detail, discussing the witnesses in the order in which the jury heard them.

A

The government presented six witnesses. K.G. testified first. She did not want to be at the trial because, in her view, “this whole situation [was] absolutely ridiculous.” App. II at 120. At the time of trial in March 2022, K.G. was 16 years old. When she was younger, K.G. experienced nightmares, particularly in “the later portion of 2017.” App. II at 136. K.G. said she sometimes “would crawl into [her] mom and dad's bed because [she] was scared” after waking up from a nightmare. App. II at 136. She did that “[s]everal times collectively over the years,” but did not regularly sleep in her parents’ bed. App. II at 173.

K.G. agreed “something happen[ed] with” Mr. Goldesberry in 2017 “that was inappropriate.” App. II at 136. She described having a nightmare and then getting into her parents’ bed, but that night, only Mr. Goldesberry was there. K.G. thought Mr. Goldesberry was asleep “[b]ecause he was snoring.” App. II at 137. At some point in the night, K.G. woke up when she “felt his hand where it shouldn't have been[:] ․ [u]nder [her] underwear.” App. II at 137–38. She said Mr. Goldesberry's hand was “[b]arely” moving because “he was still very much out of it.” App. II at 138. K.G. explained she was facing “[t]he wall” with her back to Mr. Goldesberry. App. II at 139. K.G. could not see Mr. Goldesberry when he touched her, but she heard “[h]im snoring, like, going in and out of sleep.” App. II at 140. It was “[d]eep snoring.” App. II at 140.

K.G. “freaked out” as soon as she felt Mr. Goldesberry's hand and “woke him up.” App. II at 141. She knew her father was awake at that point “[b]ecause he stopped snoring and he opened his eyes and he breathed regularly.” App. II at 141. Mr. Goldesberry “was very confused,” K.G. explained, and “he apologized profusely,” stating, “I thought you were your mother.” App. II at 141. K.G. testified that at the time of the incident, she and her mother were “close to the same height, close to the same build, and had similar hair.” App. II at 142. K.G. also confirmed that, by 2017, she had developed pubic hair.

That night, K.G. asked Mr. Goldesberry not to “tell my mom [what happened] because I [ was] embarrassed.” App. II at 142. K.G. said, “I feel like it should have been my responsibility to wake him up.” App. II at 144. For the same reason, K.G. did not tell her older sister, Faith, about the incident. K.G. thought the whole thing “was a mistake and an accident.” App. II at 143.

The first person K.G. told about the touching in 2017 was her friend Meagan Bartlett. This happened in May 2020, when the two were spending time together at Meagan's grandparents’ house in Miami, Oklahoma. “I just kind of felt like I just wanted to unload all my stuff,” K.G. testified. App. II at 145. “And I was basically just listing all of the things that I was embarrassed about or ha[d] made me sad and that have just been low points in my life․ [The touching incident in 2017] got grouped into all of that.” App. II at 145. K.G. recalled Meagan “seemed very concerned, even though I stated it was an accident.” App. II at 146.

K.G. next testified about the investigation conducted by Child Protective Services (CPS). According to K.G., Mr. Goldesberry “was very shocked” when he learned CPS was investigating, and he was surprised CPS required them “to separate” while the investigation was ongoing. App. II at 148–49. K.G. explained she “had to leave” home during the investigation. App. II at 158.

K.G. recalled a time, soon after the CPS investigation began, when Meagan and her father, Nathan Bartlett, stopped by the Goldesberry residence. K.G. confronted Meagan for reporting the incident to CPS, “yelling about the whole situation.” App. II at 149. K.G. also observed Mr. Bartlett and Mr. Goldesberry standing in the garage and talking; Mr. Goldesberry was crying. On cross-examination, K.G. again testified that, “from the beginning,” she said the touching incident “was a big misunderstanding, a big mistake.” App. II at 167–68. “I know that it was a mistake,” K.G. explained, “[b]ecause nothing had ever happened beforehand, and nothing happened afterward. And we moved on. I wanted to move on.” App. II at 188.

Meagan Bartlett testified next. At the time of trial, Meagan was twenty-two years old. She knew Mr. Goldesberry because “[h]e was my dad's best friend my entire life.” App. II at 190. Meagan said Mr. and Mrs. Goldesberry “were like my second parents.” App. II at 191. She saw the Goldesberrys a few times per month and spent time with K.G. and Faith.

Meagan did not specify what K.G. disclosed to her in May 2020. When asked if K.G. “share[d] things with you that made you worried,” Meagan answered “[y]es.” App. II at 192. According to Meagan, K.G. “did not disclose everything. I felt like there was more.” App. II at 193. Meagan encouraged K.G. to tell Mrs. Goldesberry about what K.G. had shared with her.

About a day or two later, Meagan and K.G. returned from Miami to K.G.’s house in Tulsa. When they arrived, K.G. and Mrs. Goldesberry had a private conversation, without Meagan in the room. Meagan recalled that K.G. and Mrs. Goldesberry also went for a drive together. When they got back, Mrs. Goldesberry asked Mr. Goldesberry to “go back to their bedroom with her and [K.G.].” App. II at 194. Meagan was not privy to that conversation either. Meagan testified that, later that night, she discussed with Faith and Mrs. Goldesberry “what [Mrs. Goldesberry] had been talking to [K.G.] about.” App. II at 195. When asked, “Did you learn other things that night that worried you?”, Meagan answered, “Yes․ I encouraged [Faith and Mrs. Goldesberry] to go to the police.” App. II at 195. Meagan testified she “tried to get [K.G.] to come back to Miami with me that evening along with Faith. [K.G.] declined. Faith accepted. And we left the house.” App. II at 195. Meagan said Faith “opened up ․ and gave me the same details [K.G.] did.”⁵ App. II at 210. Meagan did not further specify what Faith told her.

The next day, Meagan reported to CPS. She made the call after consulting with her father, Mr. Bartlett, about the “best course of action to take.” App. II. at 196. When K.G. found out about this call, Meagan recounted, K.G. reacted “[a]ngrily.” App. II at 196.

Then Mr. Bartlett took the stand. He had two decades of experience in law enforcement and previously worked as a campus police officer. Mr. Bartlett and Mr. Goldesberry met in high school and had known each other for 26 years. Mr. Bartlett described Mr. Goldesberry as his “best friend” and someone who had “always been there for me.” App. II at 221. Mr. Bartlett testified he first learned of the allegations against Mr. Goldesberry from Meagan when she told him about her conversation with K.G. in or around May 2020. Mr. Bartlett “thought that this is something that happened due to the amount of details that I heard [from Meagan],” and he knew the incident “needed to be reported.” App. II at 222. Mr. Bartlett testified he “sat [his] daughter down, and we called” CPS together, so Meagan could make the report. App. II at 222.

A few days later, Mr. Bartlett learned CPS contacted the Goldesberrys. Mr. Bartlett described visiting Mr. Goldesberry at his home shortly thereafter. According to Mr. Bartlett, Mr. Goldesberry was crying and “said that he messed up and messed up real bad.” App. II at 225. During the conversation, Mr. Goldesberry asked Mr. Bartlett “if he was speaking to Officer Nathan or brother Nathan.” App. II at 226. Mr. Bartlett said “Brother Nathan.” App. II at 226. Mr. Goldesberry told Mr. Bartlett he was “talking to another man of God about the issue,” but “wouldn't necessarily talk [to Mr. Bartlett] about what he had done.” App. II at 225–26. Mr. Bartlett said he tried to record the conversation with Mr. Goldesberry. “[T]here was a possibility,” Mr. Bartlett thought, that Mr. Goldesberry might say something about the 2017 incident “that needed to be recorded” “[b]ecause of the details that Meagan had given me and ․ the way he was acting.” App. II at 225–26. But Mr. Goldesberry did not “say anything else about the allegations.” App. II at 226.

Ms. Kelsey Blevins was the government's next witness. Ms. Blevins worked as the manager of forensic services for the Child Abuse Network. She testified “as an expert in child and adolescent forensic interviews, sexual abuse disclosures, and victim behavior in response to child abuse.” App. II at 259. According to Ms. Blevins, a child discloses abuse through a process involving several phases, including “denial[,] ․ tentative disclosure, active disclosure, recantation, and reaffirmation.” App. II at 261. Ms. Blevins explained “tentative disclosure” involves a child “giving vague information or pieces of information to test and see what the reaction is going to be to their disclosure, to see if they're going to be supported or believed.” App. II at 262. A child engaging in “tentative disclosure” “may” recant her admissions, Ms. Blevins testified, if the person she tells is not supportive. App. II at 263. Ms. Blevins described characteristics of “tentative disclosure” to include “minimizing,” such as saying “ ‘[i]t only happened one time,’ or ‘[i]t's not that big of a deal.’ ” App. II at 263–64. She explained a child could also minimize by taking responsibility for the abuse or deeming the abuse an accident.

Ms. Blevins described her forensic interview of K.G. during the CPS investigation. K.G. “ma[d]e a disclosure of sexual abuse,” Ms. Blevins testified, and “disclosed that her father touched her on her vagina on her skin outside and inside her body.” App. II at 278. K.G. was “protective of [Mr. Goldesberry] and a bit defensive,” Ms. Blevins said. App. II at 278. She thought K.G. exhibited “some of those characteristics of tentative disclosure, like minimizing and discounting.” App. II at 278–79. On cross-examination, Ms. Blevins agreed K.G. consistently maintained Mr. Goldesberry was “half asleep” during the incident. App. II at 302. And she clarified, “[I]t wouldn't be my job to determine if it was an accident or not an accident. It would just be my job to listen to the child's statement and whatever they're disclosing ․” App. II at 297. Ms. Blevins said she had no reason to think K.G. “was not being a hundred percent truthful” about her description of the events. App. II at 322.

Deniece Ishem, a CPS employee, testified next about a “child safety meeting” with the Goldesberrys in June 2020. App. II at 328. During that meeting, Mr. Goldesberry confirmed he touched K.G. inappropriately in 2017, but “thought [K.G.] was his wife” when it happened. App. II at 329–32. Ms. Ishem recalled Mr. Goldesberry appeared “remorseful of the incident.” App. II at 332.

Brenna Gusman, a behavioral health case manager and wellness coach for Oklahoma's Family & Children's Services, was the government's last witness. Ms. Gusman also had experience “as a child abuse/neglect investigator” for CPS. App. II at 350. In her role as an investigator, Ms. Gusman “determine[d] whether a child was safe or not in the home and if the allegations had occurred or not.” App. II at 353. In May 2020, she was assigned to investigate the incident involving Mr. Goldesberry and K.G. Ms. Gusman discovered Mr. Goldesberry had no prior child welfare history or criminal history. She ordered no contact between Mr. Goldesberry and K.G. during the CPS investigation. According to Ms. Gusman, Mrs. Goldesberry was most concerned with “the timeline of the investigation and when [CPS] would be complete with [it].” App. II at 357.

As part of the investigation, Ms. Gusman interviewed Mr. Goldesberry. Mr. Goldesberry was forthcoming and gave what Ms. Gusman believed were “specific details based on his own memory.” App. II at 362. He “admit[ted] to putting his fingers inside [K.G.’s] vagina.” App. II at 361–62. He asked Ms. Gusman whether “it was technically considered penetration if he only made it past [K.G.’s] labia majora.” App. II at 362.

Ms. Gusman concluded Mr. Goldesberry engaged in a sexual act with K.G. because “there was no disputing that it had happened.” App. II at 367. Ms. Gusman “recommended trauma-informed therapy for” K.G. App. II at 364. She also recommended a “parent education group” class for Mrs. Goldesberry, “where parents of victims of sexual abuse learn about how to help care for the child[ and] lead them through recovering from whatever it is that they may have experienced.” App. II at 365.

At the close of the government's case-in-chief, Mr. Goldesberry moved for a judgment of acquittal under Federal Rule of Criminal Procedure 29. He argued the prosecution had not established beyond a reasonable doubt “that [the sexual act] was knowingly done.” App. II at 382. The district court reserved ruling on Mr. Goldesberry's Rule 29 motion, and trial resumed.

B

Mr. Goldesberry did not testify, but the defense called two witnesses: Carla Barboza, a CPS employee, and Mrs. Goldesberry, the defendant's wife of over 20 years.

Ms. Barboza assumed the Goldesberry case for CPS after Ms. Gusman completed the investigation. Ms. Barboza's role was to “work with the families when children [were] not placed in custody” and “to make sure that the behaviors that led [CPS] to be involved with them w[ould] be corrected.” App. II at 391. The Goldesberrys cooperated with CPS, Ms. Barboza explained. She had no concerns about reuniting K.G. with her family, including Mr. Goldesberry.

Mrs. Goldesberry testified last. She described how the primary bedroom in the Goldesberry home looked in 2017. The windows had “blackout curtains” because Mrs. Goldesberry “work[ed] night shift[s] and wanted to make sure [her] room was blacked out” so she could sleep. App. II at 403. With the blackout curtains drawn and the lights off, the bedroom was “completely dark.” App. II at 403. Mrs. Goldesberry worked as a labor-and-delivery nurse, and she sometimes had to go into the hospital with little notice. When unexpectedly called into work overnight, Mrs. Goldesberry would usually “give [Mr. Goldesberry] a quick kiss and say, you know, ‘Bye. I love you. Have a good day.’ ” App. II at 412. But she did not do so every time, and “there[ would be] times that he[ would] forget” she had gone. App. II at 412. Mrs. Goldesberry said her husband was “a heavy sleeper” who suffers from “sleep apnea.” App. II at 404.

Mrs. Goldesberry described the night in May 2020 when Meagan Bartlett and K.G. came back home to Tulsa after visiting Meagan's grandparents in Miami. “Meagan was really insistent that [K.G.] and I needed to talk,” Mrs. Goldesberry recalled, “and I wasn't sure why. But she just said we needed to talk alone.” App. II at 406–07. Mrs. Goldesberry and K.G. had a “private conversation” that night. App. II at 406. When asked about “the subject matter of the conversation that you and [K.G.] had in private that night,” Mrs. Goldesberry answered, “It was about [K.G.’s] self-harm.” App. II at 407. Up until that point, Mrs. Goldesberry was not aware K.G. “had started cutting [herself],” but she knew K.G. had experienced bullying at school and “was already struggling” with mental health issues and “grief from the loss of both of her grandparents.” App. II at 407–08. Mrs. Goldesberry said K.G. was seeing a counselor.

Mrs. Goldesberry testified that after she and K.G. talked privately, they went out together to pick up a pizza. When they returned, Mrs. Goldesberry “ask[ed] [K.G.] if she could talk with ․ her dad and I. And we needed to have a conversation about what we had talked about earlier.” App. II at 409. According to Mrs. Goldesberry, she and K.G. then had a private conversation with Mr. Goldesberry about K.G.’s self-harm. Mrs. Goldesberry confirmed that, at that point, she was not yet aware of the touching incident involving K.G. and Mr. Goldesberry.

Mrs. Goldesberry first learned about the “bed incident” in 2017 from Ms. Gusman. Mrs. Goldesberry believed what happened was a mistake and not intentional. K.G. rarely slept in her parents’ bed. Mr. Goldesberry often initiated sex with Mrs. Goldesberry at night by touching her while they were in bed together. Mrs. Goldesberry testified that, in 2017, she and K.G. were “about the same size,” and “by the time [K.G.] was 12,” they were “around the same height[ ].” App. II at 423.

Before the defense rested, Mr. Goldesberry renewed his Rule 29 motion outside the presence of the jury. The district court again reserved ruling on the motion. The jury then returned to the courtroom for closing arguments and instructions. The government argued testimony about the incident itself and the events thereafter established Mr. Goldesberry knowingly committed the sexual act with K.G. Mr. Goldesberry argued “[a]ll of the evidence ․ is that this was an unfortunate incident in the parents’ bed that happened because of Mr. Goldesberry's mistake or accident in believing his wife, who he'd gone to sleep with, was still there with him when he woke up.” App. II at 508.

The jury found Mr. Goldesberry guilty of the offense charged. The district court denied Mr. Goldesberry's Rule 29 motion. Mr. Goldesberry was sentenced to 30 years’ imprisonment, the mandatory minimum under 18 U.S.C. § 2241(c). This timely appeal followed.

II

Mr. Goldesberry claims the government failed to present sufficient evidence to support his conviction. Sufficiency of the evidence is a question of law reviewed de novo. United States v. Rufai, 732 F.3d 1175, 1188 (10th Cir. 2013); see also United States v. Garcia, 74 F.4th 1073, 1117 (10th Cir. 2023) (“We review the denial of a judgment of acquittal for insufficient evidence de novo.”). We begin by describing the background principles that inform sufficiency review and then distill the mechanics of our approach.

A

To determine whether evidence is sufficient to uphold a conviction, “we examine, in the light most favorable to the government, all of the evidence together with the reasonable inferences to be drawn therefrom and ask whether any rational juror could have found the essential elements of the crime beyond a reasonable doubt.” United States v. Arutunoff, 1 F.3d 1112, 1116 (10th Cir. 1993). That standard is well settled and oft stated, but what does it mean, really?

Most fundamentally, sufficiency review strikes a delicate balance. Sufficiency review “is ‘highly deferential.’ ” United States v. Burtrum, 21 F.4th 680, 685 (10th Cir. 2021) (quoting United States v. Pickel, 863 F.3d 1240, 1251 (10th Cir. 2017)). “[I]t is the responsibility of the jury—not the court—to decide what conclusions should be drawn from evidence admitted at trial.” Cavazos v. Smith, 565 U.S. 1, 2, 132 S.Ct. 2, 181 L.Ed.2d 311 (2011). When we review the sufficiency of evidence, “[w]e will not weigh conflicting evidence or second-guess the fact-finding decisions of the jury.” United States v. Summers, 414 F.3d 1287, 1293 (10th Cir. 2005). This approach preserves the factfinder's role “to resolve conflicts in the testimony, to weigh the evidence, and to draw reasonable inferences from basic facts to ultimate facts.” Jackson v. Virginia, 443 U.S. 307, 319, 99 S.Ct. 2781, 61 L.Ed.2d 560 (1979). “[T]he inevitable consequence of this settled law is that judges will sometimes encounter convictions that they believe to be mistaken, but that they must nonetheless uphold.” Cavazos, 565 U.S. at 2, 132 S.Ct. 2.

At the same time, proper deference to the jury's verdict does not mean abdication of the judicial role. We must be vigilant in enforcing “[t]he constitutional standard ․ that protects an accused against a conviction” “except upon proof beyond a reasonable doubt of every fact necessary to constitute the crime with which he is charged.” Jackson, 443 U.S. at 315, 99 S.Ct. 2781 (quoting In re Winship, 397 U.S. 358, 364, 90 S.Ct. 1068, 25 L.Ed.2d 368 (1970)). The Supreme Court has long acknowledged “the traditional understanding in our system that the application of the beyond-a-reasonable-doubt standard to the evidence is not irretrievably committed to jury discretion.” Id. at 317 n.10, 99 S.Ct. 2781. And the jury's function in a criminal trial “has never been thought to include a power to enter an unreasonable verdict of guilty.” Id. We may “impinge[ ] upon ‘jury’ discretion only to the extent necessary to guarantee the fundamental protection of due process of law.” Id. at 319, 99 S.Ct. 2781. Sufficiency review thus safeguards a defendant's right to proof beyond a reasonable doubt.

B

Guided by these foundational principles, we now summarize, in plain terms, our approach to the sufficiency issue before us.

First, we “must consider the burden of proof in [our] sufficiency-of-the-evidence analysis.” Rufai, 732 F.3d at 1188. Our “critical inquiry” is “to determine whether the record evidence could reasonably support a finding of guilt beyond a reasonable doubt,” Jackson, 443 U.S. at 318, 99 S.Ct. 2781—i.e., whether a factfinder could reasonably conclude the defendant's guilt has been established “with utmost certainty,” Winship, 397 U.S. at 364, 90 S.Ct. 1068. And we hold the government to its burden “on each element of the offense.” Thomas v. United States, 409 F.2d 730, 731 (10th Cir. 1969).

Second, we do not “examin[e] the evidence in bits and pieces,” United States v. Rodriguez-Flores, 907 F.3d 1309, 1312 (10th Cir. 2018) (quoting United States v. Brooks, 438 F.3d 1231, 1236 (10th Cir. 2006)), but instead “consider[ ] the collective inferences to be drawn from the evidence as a whole,” United States v. Tennison, 13 F.4th 1049, 1059 (10th Cir. 2021) (quoting United States v. Nelson, 383 F.3d 1227, 1229 (10th Cir. 2004)). In doing so, we must “consider the evidence and all reasonable inferences in the light most favorable to the government.” United States v. Reddeck, 22 F.3d 1504, 1507 (10th Cir. 1994). “[W]e do not give the government the benefit of every potential inference but rather, only those inferences reasonably and logically flowing from the other evidence adduced at trial.” United States v. Santistevan, 39 F.3d 250, 258 (10th Cir. 1994). An inference is unreasonable if it requires the jury “to engage in a degree of speculation and conjecture that renders its findings a guess or mere possibility.” United States v. Jones, 44 F.3d 860, 865 (10th Cir. 1995) (quoting United States v. Jones, 49 F.3d 628, 632 (10th Cir. 1995)). Any inference must “be made beyond a reasonable doubt” if it “goes to an ultimate conclusion underpinning criminal liability, e.g., satisfying an element of a crime necessary for conviction.” Summers, 414 F.3d at 1295 n.4 (quoting United States v. Rahseparian, 231 F.3d 1257, 1264 (10th Cir. 2000)).

Third, we may ask, after viewing the evidence and inferences from it in the government's favor, whether the evidence “gives equal or nearly equal circumstantial support to a theory of guilt and a theory of innocence”—that is, whether the evidence is in equipoise. United States v. Lovern, 590 F.3d 1095, 1107 (10th Cir. 2009) (Gorsuch, J.) (quoting United States v. Caseer, 399 F.3d 828, 840 (6th Cir. 2005)); see United States v. Simon, 12 F.4th 1, 32 (1st Cir. 2021) (explaining the equipoise principle “takes hold only after [the inquiring court] ha[s] drawn all reasonable inferences in favor of the verdict” (emphasis and alterations in original) (quoting Magraw v. Roden, 743 F.3d 1, 5 (1st Cir. 2014))). When the trial record—after it is properly construed for the prosecution—is in equipoise, we must reverse the conviction because “under these circumstances a reasonable jury must necessarily entertain a reasonable doubt.” Lovern, 590 F.3d at 1107 (quoting Caseer, 399 F.3d at 840).⁶

To be clear, the equipoise principle does not require a court to ask “whether a reasonable jury could possibly conceive of an alternative interpretation of the evidence at trial.” Burtrum, 21 F.4th at 686 (quoting United States v. White Bull, 646 F.3d 1082, 1089 (8th Cir. 2011)); see also United States v. Wood, 207 F.3d 1222, 1228 (10th Cir. 2000) (“[T]he evidence necessary to support a verdict ․ ‘need not negate all possibilities except guilt.’ ” (quoting United States v. Wilson, 182 F.3d 737, 742 (10th Cir. 1999))); United States v. Hooks, 780 F.2d 1526, 1530 (10th Cir. 1986) (explaining we do not “reverse criminal convictions on the ground that the evidence was consistent with a reasonable hypothesis of innocence”). Rather, the rule of equipoise—correctly applied after all reasonable inferences are drawn for the government—is a helpful tool that ensures the principled application of the sufficiency standard.

III

With the parameters of our review thus established, we move to the merits. The issue is whether a reasonable juror could find the evidence proved beyond a reasonable doubt that Mr. Goldesberry touched K.G. knowingly— rather than by mistake or accident—and we view that evidence in the government's favor. As we explain, reversal is required because the evidence of Mr. Goldesberry's knowledge is insufficient to support a finding of guilt beyond a reasonable doubt.

A

Section 2241(c), the charging statute here, provides a person is guilty if he “knowingly engages in a sexual act with another person who has not attained the age of 12 years ․ or attempts to do so.” 18 U.S.C. § 2241(c). Recall, “our inquiry is directed to ascertaining whether ․ there is sufficient direct or circumstantial evidence ․ on each element of the offense.” Thomas, 409 F.2d at 731 (emphasis added). As the trial judge instructed, the charged offense required proof beyond a reasonable doubt of the following elements:

First: the defendant knowingly engaged in or attempted to engage in a sexual act with [K.G.]; Second: at the time, [K.G] had not attained the age of 12 years; Third: the defendant is an Indian; and Fourth: the act occurred in Indian country within the Northern District of Oklahoma.

App. I at 136 (italics omitted); App. II at 527.

Mr. Goldesberry does not challenge the actus reus element—that he “engaged in ․ a sexual act with [K.G.].” Nor could he. As Mr. Goldesberry concedes, the trial evidence could readily establish this element beyond a reasonable doubt. See Op. Br. at 10 n.4 (“The precise character of the touching was established and is undisputed: K.G. testified that she awoke and called out because she felt a hand inside her underwear, moving slightly and ‘touching my vagina’—on ‘the outside’ and ‘barely the inside,’ just past the labia majora.” (citations omitted) (quoting App. II at 138–41)).

On appeal, Mr. Goldesberry challenges only the mens rea element— knowledge. The charging statute does not define “knowingly.” We have held, as a general matter, “the term ‘knowingly’ means that an act was done ․ voluntarily and intentionally, and not because of mistake or accident.” United States v. Dobbs, 629 F.3d 1199, 1204 (10th Cir. 2011) (internal quotation marks omitted); Dixon v. United States, 548 U.S. 1, 5, 126 S.Ct. 2437, 165 L.Ed.2d 299 (2006) (“[U]nless the text of the statute dictates a different result, the term ‘knowingly’ merely requires proof of knowledge of the facts that constitute the offense.” (quoting Bryan v. United States, 524 U.S. 184, 193, 118 S.Ct. 1939, 141 L.Ed.2d 197 (1998))). Mr. Goldesberry has consistently maintained the 2017 incident was an accident: he did not know K.G. was in his wife's spot in their bed when the touching occurred.

As all agree, the government offered no direct evidence of Mr. Goldesberry's mens rea. There is nothing wrong with that. “The government can, and ordinarily does, prove knowledge and intent through circumstantial evidence.” United States v. Banks, 884 F.3d 998, 1018 (10th Cir. 2018). We have never “require[d] the government to produce a ‘smoking gun’ that explicitly reveals the contents of [a] defendant's mind.” United States v. Camick, 796 F.3d 1206, 1220 (10th Cir. 2015) (quoting United States v. Ashley, 606 F.3d 135, 140–41 (4th Cir. 2010)). Rather, “[i]nferences are necessary and indeed proper in a criminal trial.” Summers, 414 F.3d at 1295; see Shoppin’ Bag of Pueblo, Inc. v. Dillon Cos., Inc., 783 F.2d 159, 163 (10th Cir. 1986) (“Often no direct evidence of specific intent exists and inferences from conduct are necessary.”); United States v. Norman T., 129 F.3d 1099, 1104 (10th Cir. 1997) (“A rational trier of fact” can infer “the required intent based on the circumstances surrounding the incident.”). “Short of incriminating statements from a defendant, proof of knowledge and intent must be based on drawing inferences from the defendant's actions, the testimony and actions of others, and documentary evidence.” Rufai, 732 F.3d at 1191–92. Of course, the government's burden does not ease where its evidence is circumstantial. Recall, every inference the government seeks must be reasonable—that is, there must be a “reasonable probability that the conclusion flows from the proven facts.” Jones, 44 F.3d at 865 (quoting Jones, 49 F.3d at 632). And an inference must “be made beyond a reasonable doubt” if—as here—it goes to an element of the crime needed to convict. Summers, 414 F.3d at 1295 n.4 (quoting Rahseparian, 231 F.3d at 1264).

Accordingly, “the key question” before us concerns “the adequacy of the evidence to sustain an inference” beyond a reasonable doubt that Mr. Goldesberry knowingly touched K.G. Jones, 44 F.3d at 866.

B

Mr. Goldesberry argues “a factfinder could not reasonably conclude that the government proved beyond a reasonable doubt that [he] touched K.G. knowingly.” Op. Br. at 19. He also invokes the equipoise principle, insisting the evidence at most “gives equal or nearly equal circumstantial support to a theory of guilt and a theory of innocence.” Op. Br. at 31 (quoting Lovern, 590 F.3d at 1107).⁷

With the evidence and all reasonable inferences viewed in the government's favor, the record establishes that Mr. Goldesberry committed a sexual act by touching K.G. But, as to the mens rea element, the evidence is insufficient to convict.⁸ We cannot affirm Mr. Goldesberry's conviction because it does not rest on proof of every element beyond a reasonable doubt. See Jackson, 443 U.S. at 318, 99 S.Ct. 2781. We explain our conclusion in two steps. We first discuss why the trial evidence did not support an inference of knowledge beyond a reasonable doubt. We then address the government's contrary arguments.

1

Before the incident. On the night in question, Mr. and Mrs. Goldesberry's bedroom was completely dark. Mrs. Goldesberry had gone to sleep with Mr. Goldesberry but was called into the hospital sometime in the middle of the night. When called into work like this, Mrs. Goldesberry often gave Mr. Goldesberry a kiss goodbye—but sometimes she did not, and sometimes he forgot even when she did. At some point after Mrs. Goldesberry left for work, K.G. entered her parents’ bedroom because she had experienced a nightmare. K.G. did this on occasion, but not frequently. K.G. then crawled into bed next to her father. Mr. Goldesberry is a heavy sleeper who suffers from sleep apnea, and K.G. assumed he was asleep because he was snoring. K.G. fell asleep too.

This evidence cannot support an inference beyond a reasonable doubt that Mr. Goldesberry was aware of his daughter's presence in bed. Undisputed testimony shows Mr. Goldesberry fell back asleep—if he was ever awake— before K.G. entered the room. Undisputed testimony further shows the room was dark and K.G. did not frequently enter her parents’ bed. At most, the evidence about events before the touching occurred supports an inference that Mrs. Goldesberry may have kissed her husband goodbye before leaving for the hospital on the night of the incident. But anything beyond that is not grounded in proven facts.

The incident. K.G. testified about what happened that night, and her testimony was undisputed. She awoke to Mr. Goldesberry touching her “under [her] underwear.” App. II at 138. Mr. Goldesberry was still snoring, but K.G. instantly “freaked out,” “jumped up,” “turned around,” and said “what are you doing?” App. II at 140–41, 168, 513. Mr. Goldesberry immediately removed his hand. App. II at 174. He also “stopped snoring,” “was very confused,” “apologized profusely,” and “said ‘I thought you were your mother; I'm sorry.’ ” App. II at 141, 170. As background, Mr. Goldesberry often initiated sex with his wife by touching her vagina in the middle of the night. Also, K.G. and her mother were similar (though not identical) sizes at the time of the offense, and K.G. had developed pubic hair.

Everyone agrees, this evidence obviously supported an inference of actus reus. But that does not mean it supported an inference of mens rea. See Lovern, 590 F.3d at 1104 (“[T]he government's proof, while it showed many things, failed to include direct or circumstantial evidence that [the defendant] possessed this particular mens rea.”). Viewed in the light most favorable to the government, this evidence shows Mr. Goldesberry expressed no indicia of awareness when he touched K.G., immediately stopped touching her when she woke him (which was as soon as she felt his hand), showed “profuse[ ]” remorse, had an explanation for why he might sub-consciously touch his bed-mate's vagina in the middle of the night, and had no obvious way to know the genitals he touched belonged to a child. As to the mens rea element, therefore, the evidence about the incident itself does not support an inference of guilt. This evidence supports a theory Mr. Goldesberry acted by mistake at least as much as—if not more than—a theory Mr. Goldesberry acted with knowledge.⁹ Cf. Nevils, 598 F.3d at 1167 (reversal may be required when evidence is sufficiently “supportive of innocence”).

After the incident. Immediately after the touching incident, K.G. was embarrassed and asked Mr. Goldesberry not to tell Mrs. Goldesberry what happened. Neither spoke about it for several years. Then in May 2020, K.G. confided in her friend Meagan about the 2017 incident and other life difficulties, telling Meagan “it was an accident.” App. II at 146. Meagan thought K.G.’s sister and mother should report it to the police. Meagan then told her father about the incident, and he also believed the incident needed to be reported. Meagan reported the touching to CPS the next day. Sometime later, Mr. Goldesberry expressed remorse about the touching in a conversation with Mr. Bartlett.

CPS opened an investigation. In a forensic interview, K.G. said she felt safe with her father and that what happened was a mistake. She believed Mr. Goldesberry had mistaken her for Mrs. Goldesberry while half asleep. Mr. Goldesberry cooperated with the investigators and provided them with “specific details” about the incident. App. II at 362, 394. Following the investigation, CPS employee Ms. Barboza had no concerns about reuniting K.G. and Mr. Goldesberry.

The evidence of what happened after the touching incident supports actus reus, not mens rea. Viewed in the government's favor, the evidence presented to the jury primarily shows K.G. and Mr. Goldesberry have both acknowledged the touching since it occurred. Specifically, K.G. told Meagan about the touching (and that she believed it was an accident), and Mr. Goldesberry expressed remorse to Mr. Bartlett and told investigators he had inappropriately touched K.G. But these facts do not support an inference of knowledge beyond a reasonable doubt. If anything, K.G.’s consistent position that Mr. Goldesberry acted by mistake supports a theory of mistake, as do Mr. Goldesberry's cooperation with CPS and CPS's conclusion K.G. was safe to return home.

After reviewing the pre-incident evidence, incident evidence, and post-incident evidence as a whole and in the government's favor, we conclude the evidence simply could not support a theory of knowledge over a theory of mistake. We emphasize it is not our role to weigh the evidence. Even if the jury discounted evidence favorable to Mr. Goldesberry, however, the government would still have no affirmative evidence of mens rea. See United States v. Mills, 29 F.3d 545, 550 (10th Cir. 1994) (“Even if the jury disbelieved the entire defense testimony, that disbelief cannot constitute evidence of the crimes charged ․”); Wessel v. Buhler, 437 F.2d 279, 282 (9th Cir. 1971) (“Of course, the jury could have disbelieved the denials, but disbelief does not become a substitute for affirmative evidence.”).¹⁰ At best, the evidence of knowledge “gives equal or nearly equal circumstantial support to a theory of guilt and a theory of innocence,” and “under th[o]se circumstances a reasonable jury must necessarily [have] entertain[ed] a reasonable doubt.” Lovern, 590 F.3d at 1107 (quoting Caseer, 399 F.3d at 840).

2

The government insists there was “ample circumstantial evidence” establishing knowledge. Ans. Br. at 30. It points mostly to post-incident events, focusing on the following four categories of circumstantial evidence:

1. Mr. Goldesberry recalled during his interview with Ms. Gusman that, on the night of the incident, he “rolled over on his stomach,” “reached past the shorts and panties,” and “began to stimulate after his fingers were past the labia majora,” App. II at 362; Ans. Br. at 38;

2. Mr. Goldesberry told Mr. Bartlett he “messed up real bad,” asked Mr. Bartlett whether “he was speaking to Officer Nathan or brother Nathan,” and ultimately told Mr. Bartlett nothing “else about the allegations,” App. II at 225–26; Ans. Br. at 32–33 11;

3. K.G.’s testimony—which downplayed the abuse—was consistent with what Ms. Blevins called the “tentative disclosure” signs of “minimizing” and “empowerment,” and K.G. gave apparently conflicting statements about her age at the time of the incident, App. II at 264, 278–79, 152–54; Ans. Br. at 35–36; and

4. K.G. told Meagan Bartlett unspecified “things ․ that made [Meagan] worried,” K.G.’s sister Faith “gave [Meagan] the same details,” Meagan and Mr. Bartlett each thought this information needed to be reported to police, and, shortly after K.G. told Meagan about the incident, Meagan saw K.G. speak privately with her parents, App. II at 192–95, 210, 223; Ans. Br. at 31.

None of this evidence disturbs our analysis or changes our conclusion.

The interview with Ms. Gusman. We begin with Mr. Goldesberry's interview with Ms. Gusman, which is likely the government's strongest evidence of knowledge. The government argues “[a] reasonable jury could have inferred that Goldesberry, when taking the actions he was able to remember precisely and describe three to four years later, would have been able to distinguish between his 37-year-old wife and his 10 or 11-year-old daughter.” Ans. Br. at 39. We disagree.

To be sure, the jury could reasonably infer Mr. Goldesberry was half-awake during the incident, even that he knew he was touching a female body next to him. But it requires mere guess work to also infer Mr. Goldesberry knew this person was K.G. Consider, K.G. was occupying her mother's usual place in bed, the room was completely dark, Mr. Goldesberry had been asleep when K.G. walked in, K.G. and Mrs. Goldesberry had similar-sized bodies at that time, K.G. had developed pubic hair, Mr. Goldesberry was still snoring when K.G. noticed the touching, and Mr. Goldesberry apologized profusely once K.G. woke him up. The government identifies no evidence a juror could have used to find it was more likely than not—let alone established beyond a reasonable doubt— that Mr. Goldesberry knew the woman next to him was K.G. See Webb v. United States, 347 F.2d 363, 364 (10th Cir. 1965) (“[T]he jury's function [in a criminal case] is broad enough to allow it to make common sense inferences from proven facts ․” (emphasis added)). The gap between what the jury heard—that Mr. Goldesberry later remembered specific details of the incident—and what the prosecution insists can be inferred from this evidence to convict—that he touched K.G. knowingly, not by mistake—is bridged only by speculation. Rosenberg v. United States, 120 F.2d 935, 937 (10th Cir. 1941) (“Conviction cannot be superimposed upon presumption and thus reach the ultimate conclusion of guilt.”).

As discussed, we agree with Mr. Goldesberry that “what the prosecution was asking the jury to do was to infer mens rea from the actus reus itself.” Op. Br. at 33. But the government cannot prove mens rea simply by repeatedly proving actus reus—unless, as Mr. Goldesberry acknowledges, the separate mens rea “inference may be made beyond a reasonable doubt.” Op. Br. at 33 (quoting Rahseparian, 231 F.3d at 1264). Here it could not. See Stallings v. Tansy, 28 F.3d 1018, 1024 (10th Cir. 1994) (“When ‘the evidence must be buttressed by surmise and conjecture, rather than logical inference in order to support a conviction, ․ such conviction [cannot be allowed] to stand.’ ” (alterations in original) (quoting State v. Romero, 67 N.M. 82, 83, 352 P.2d 781 (1960))).

Viewed properly in the government's favor, reasonable inferences from Ms. Gusman's testimony lend no more support to guilt than to innocence. Mr. Goldesberry's statements to Ms. Gusman seem to describe the very theory of defense he supported with evidence at trial: that he committed the actus reus because he mistook K.G. for his wife. App. II at 508.

The government's other categories of evidence provide even less support for affirmance. We must be mindful in our sufficiency review “that at some point along a rational continuum, inferences may become so attenuated from underlying evidence as to cast doubt on the trier of fact's ultimate conclusion.” Summers, 414 F.3d at 1295; Sunward Corp. v. Dun & Bradstreet, Inc., 811 F.2d 511, 521 (10th Cir. 1987) (“Such an inference is infirm because it is not based on the evidence.” (quoting Daniels v. Twin Oaks Nursing Home, 692 F.2d 1321, 1326 (11th Cir. 1982))). That is the case here.

Mr. Goldesberry's discussion with Mr. Bartlett. From Mr. Goldesberry's statements to Mr. Bartlett, the government argues “the jury could have inferred that Goldesberry knew of the wrongfulness of his conduct and did not want to discuss it with a law enforcement officer.” Ans. Br. at 33. The government thus relies on Mr. Goldesberry's remorse and refusal to speak with law enforcement as proven facts to support an inference of knowledge. Here, too, we must reject the government's argument because the evidence of knowledge—properly viewed in favor of the prosecution—does not support an inference of knowledge beyond a reasonable doubt.

First, the government's evidence readily supports an inference that Mr. Goldesberry expressed remorse for engaging in a sexual act with his daughter. But a reasonable juror would expect a father who sexually touched his daughter to feel remorseful whether he acted knowingly or by mistake—that is, whether Mr. Goldesberry is guilty or innocent of the charged offense. The evidence of remorse presented to the jury therefore cannot support a theory of knowledge any more than a theory of mistake.

Second, on Mr. Goldesberry's refusal to speak with Mr. Bartlett, this refusal may not even support an inference Mr. Goldesberry was avoiding law enforcement, rather than an inference he did not want to discuss a traumatizing topic. But even assuming a jury could infer Mr. Goldesberry was concerned about speaking with a law enforcement officer, this still would not support the further inference that Mr. Goldesberry believed he was guilty, as opposed to an inference that he believed he was innocent but accused of a crime. Mr. Goldesberry would have been right to be concerned, after all, because Mr. Bartlett did try to record his conversation with Mr. Goldesberry as evidence for a “criminal trial.” App. II at 225. Mr. Bartlett's testimony thus provided the jury “no non-speculative reason to favor” the government's case. Lovern, 590 F.3d at 1107.

K.G.’s post-incident statements. Concerning K.G.’s post-incident statements, the government first emphasizes Ms. Blevins’ testimony about abuse disclosures. The government argues, “in light of Ms. Blevins's expert testimony, a reasonable jury could have concluded that K.G.’s 2020 disclosure to Meagan Bartlett was a delayed tentative disclosure consistent with abuse by a close family member,” after which K.G. “recanted.” Ans. Br. at 35–36. It goes on, “the jury could reasonably have understood K.G.’s” testimony to contain signs of “minimization,” “discounting,” “and empowerment” that “corroborat[ed], rather than contradict[ed], her original disclosure.” Ans. Br. at 36.

Separately, the government points to an apparent conflict in K.G.’s statements concerning the age element of the offense. On one hand, K.G. said during her forensic interview that she did not recall her age at the time of the incident, and Mr. Goldesberry told CPS investigators the incident occurred “three or four years” before 2020 (which would mean the incident occurred when K.G. was ten or eleven); but on the other hand, at trial, K.G. testified she was twelve at the time of the incident. Ans. Br. at 36–37. Compare App. II at 154, 117, 329, 380, with App. II at 153. The government tells us a “jury could have inferred” K.G. “had been influenced by her concern for her father and her family situation, and that the same concern motivated K.G.’s insistence that her father had been asleep and had ‘made a mistake’ ” during the incident. Ans. Br. at 37. We are not persuaded.

First, Ms. Blevins’ testimony, viewed in the government's favor, is no boon for the prosecution. There is simply no evidence K.G. ever said anything to anyone that would cast doubt on her testimony the incident was an accident. As the jury heard it, K.G.’s account never varied from the time she disclosed the incident in May 2020 to the moment of trial in March 2022. Given the evidence presented, the jury could not reasonably infer K.G. recanted her disclosure to Meagan, as the government seems to assume.

Likewise, the jury could not reasonably infer K.G. believed Mr. Goldesberry acted knowingly—when she consistently said otherwise— because sex abuse victims often downplay their abuse or say they would not tolerate abuse. Ms. Blevins acknowledged she had no reason to doubt K.G.’s statements. Indeed, the government's argument would imply any statement K.G. could have made about the incident would have supported Mr. Goldesberry's guilt. Of course, K.G. would have made the same statements she did if Mr. Goldesberry acted by mistake.¹²

Second, K.G.’s age-related testimony does not support an inference of knowledge. For one, K.G. never contradicted herself. While K.G. did not recall the date of the incident during her forensic interview, the government omits an important detail. Before the interview, in 2020, K.G. told Meagan the incident happened “two to three years” earlier—that is, when K.G. was about twelve. App. II at 146. Nor is it surprising K.G. did not recall a specific date in 2020—when “[the touching incident] wasn't something that was discussed as regular[ly] as it has been in this case”—before recalling the timeline when prosecutors asked her to focus on dates. App. II at 154–56. And though the evidence shows Mr. Goldesberry told CPS investigators the incident occurred “three or four years” before 2020, it does not show K.G. ever said that. Regardless, even if the jury could infer K.G. contradicted herself out of “concern for her father,” it would require vast speculation to infer beyond a reasonable doubt this also meant Mr. Goldesberry acted knowingly.

At best, the jury could have used Ms. Blevins’ testimony and K.G.’s age-related statements to discount K.G.’s exculpatory testimony. But as we have already discussed, that disbelief would not provide the government with affirmative evidence to meet its burden of proving knowledge beyond a reasonable doubt. Mills, 29 F.3d at 550 (explaining “disbelief” of defense's evidence “cannot constitute evidence of the crimes charged ․”); Wessel, 437 F.2d at 282 (similar).

What Meagan Bartlett heard about the incident. Meagan and Mr. Bartlett's further testimony is also insufficient. According to the government, the jury could “infer that more concerning details about Goldesberry's abuse had been disclosed to Meagan” than what Meagan actually testified about at trial. Ans. Br. at 32. The government tells us this is true because, recall, K.G. shared “things” that made Meagan worried, K.G.’s sister Faith later told Meagan the “same details” K.G. did, Meagan and Mr. Bartlett thought the incident needed to be reported, and Meagan saw K.G. speak with her parents shortly after K.G.’s disclosure.

The government's argument requires the jury to engage in speculation. From Meagan and Mr. Bartlett's testimony, a jury could reasonably infer K.G. told Meagan that Mr. Goldesberry touched her—a fact about actus reus not in dispute. A jury could not reasonably infer the unspecified “things” K.G. shared with Meagan included the completely novel information that Mr. Goldesberry acted knowingly. The jury heard nothing that would support such an inference.

The fact Faith discussed concerning things with Meagan also cannot secure an inference of knowledge beyond a reasonable doubt. The jury never knew what “details” Faith told Meagan. That Meagan and Mr. Bartlett thought these unspecified “things” needed to be reported is similarly inapposite. Learning about the undisputed actus reus—whether committed knowingly or by mistake—would make any reasonable person concerned. As for Meagan seeing K.G. and her parents meet in private, the jury only heard evidence about what happened at that meeting from Mrs. Goldesberry: “It was about [K.G.’s] self-harm.” App. II at 407. Even if the jury found Mrs. Goldesberry less than credible, the jury had no evidence from which to infer that private meeting concerned the touching incident, much less that it revealed Mr. Goldesberry's knowledge at the time of the incident. So again, this testimony is not helpful to the government.

3

The evidence presented to the jury could establish actus reus beyond a reasonable doubt. Here, that is insufficient to prove the contested essential element: mens rea. See Lovern, 590 F.3d at 1104 (“[T]he government's proof, while it showed many things, failed to include direct or circumstantial evidence that [the defendant] possessed this particular mens rea.”); Dobbs, 629 F.3d at 1204 (“[A]lthough there is no question that a rational jury could have found that Mr. Dobbs ‘received’ the two images, we conclude that it could not have found that Mr. Dobbs did so knowingly.”); Rahseparian, 231 F.3d at 1263 (“None of this evidence, even when considered in a light most favorable to the government, supports beyond a reasonable doubt an inference [of knowledge].”). Importantly, because sufficiency review requires us to view the evidence and all inferences in the government's favor, our conclusion may have been different if the government presented any evidence that suggested the touching did not occur by accident. No such evidence was before the jury. Under these circumstances, we cannot say that Mr. Goldesberry was convicted “upon proof beyond a reasonable doubt of every fact necessary to constitute the crime with which he is charged.” Jackson, 443 U.S. at 315, 99 S.Ct. 2781 (emphasis added) (quoting Winship, 397 U.S. at 364, 90 S.Ct. 1068). The equipoise principle further assures us of our conclusion. Viewed in the government's favor, at best, the evidence on mens rea is in equipoise: it equally supports the theory that Mr. Goldesberry touched K.G. knowingly and the theory that he committed the act by mistake.

***

The government must both “convince the jury with its proof” and then “also satisfy the courts” that the evidence permits a reasonable conclusion that the defendant is guilty beyond a reasonable doubt. United States v. Powell, 469 U.S. 57, 67, 105 S.Ct. 471, 83 L.Ed.2d 461 (1984). We cannot say the government has carried its burden in this case.

IV

Mr. Goldesberry's conviction is VACATED and his case REMANDED for further proceedings consistent with this opinion.

Our review for sufficiency of the evidence is quite limited: “[W]e will not overturn a jury's verdict unless no reasonable juror could have concluded, on the basis of the evidence presented, that the defendant was guilty of the crime charged.” United States v. Dowlin, 408 F.3d 647, 657 (10th Cir. 2005) (quotation omitted). This leaves us with “very little leeway” in evaluating sufficiency-of-the-evidence claims. United States v. Flechs, 98 F.4th 1235, 1243 (10th Cir. 2024) (quotation omitted). Although the majority acknowledges this stringent standard, it improperly substitutes its preferred view of the evidence to conclude that no reasonable juror could find Goldesberry acted knowingly when he digitally penetrated K.G.’s vagina. I respectfully dissent.

I.

Goldesberry concedes he engaged in a sexual act with his daughter. Aplt. Br. at 13. The only question is whether he did so knowingly. He says it was all a mistake—that he mistook his eleven-year-old daughter for his thirty-seven-year-old wife—because he was half-asleep and the bedroom was dark. Id. at 19–20. The majority agrees and concludes no reasonable juror could find otherwise. Maj. Op. at –––– – ––––.

Cases like Goldesberry's often present difficulties because direct evidence of a defendant's intent is seldom available. United States v. Johnson, 971 F.2d 562, 566 (10th Cir. 1992). Only Goldesberry knows what he was thinking when he digitally penetrated his daughter's vagina. But this does not mean the jury was required to accept blindly his version of the facts. Our judicial system enables—and often requires—jurors to make mens rea determinations based on the plausibility of a defendant's testimony under the circumstances:

Because motivation is difficult to discern, the jury is permitted to draw inferences of subjective intent from a defendant's objective acts and statements. Thus, even when a defendant, as here, denies having the requisite intent, a jury may disbelieve the defendant if [his] words and acts in the light of all the circumstances make [his] explanation seem improbable.

United States v. Porter, 928 F.3d 947, 957 (10th Cir. 2019) (quotations omitted); United States v. Wyatt, 964 F.3d 947, 953 (10th Cir. 2020) (same); Flechs, 98 F.4th at 1248 (same); 1 Wayne R. LaFave and Austin W. Scott, Jr., Substantive Criminal Law, § 3.5, at 318 (1986) (same); see Webb v. United States, 347 F.2d 363, 364 (10th Cir. 1965) (noting the jury's authority to apply “common sense inferences” to the case presented).

This case is no exception to the rule. Like defendants in previous cases, Goldesberry “denie[d] having the requisite intent” to support his conviction. Porter, 928 F.3d at 957 (quotation omitted). But the jury “disbelieve[d]” Goldesberry because his “words and acts in the light of all the circumstances” rendered his theory of the case implausible. Id. And importantly, this determination was not without foundation: “[C]ommon sense inferences from proven facts” strongly support the jury's disbelief of Goldesberry's explanation. Webb, 347 F.2d at 364. For example, the jury learned at trial that Goldesberry had been married for twenty years and that the union had produced two children. From this proven fact, the jury could reasonably infer Goldesberry was familiar with his wife's body and could tell the difference between her developed body and his eleven-year-old daughter's body. Similarly, the jury heard evidence that K.G. had climbed into her parents’ bed on several previous occasions after experiencing a nightmare. The jury could reasonably infer from this that Goldesberry was aware he was sleeping next to his daughter, and not his wife.

Taken together, these proven facts—along with the reasonable inferences drawn therefrom—cast doubt on Goldesberry's claim that the touching occurred by mistake. But perhaps most telling that Goldesberry acted with knowledge is that trial testimony contradicted his earlier statements to Child Protective Services. At trial, the defense repeatedly claimed Goldesberry was “half-asleep” and the room was dark when he digitally penetrated K.G.’s vagina. R. Vol. II at 168. But when an investigator from Child Protective Services interviewed Goldesberry in 2020—years after the incident occurred—Goldesberry described his clear and specific recollection of the sexual touching. He told the investigator “he rolled over on his stomach to stimulate,” “reached past the shorts and panties,” and “began to stimulate after his fingers were past the labia majora.” Id. at 362. He specifically asked the investigator “if it was technically considered penetration if he only made it past the labia majora.” Id. And he claimed that only after his daughter exclaimed, “Dad, what are you doing?” did he realize he was touching the genitalia of an eleven-year-old child, and not of his wife. Id. at 141.

The jury saw through that. Based on Goldesberry's detailed recollection of his actions, it rejected Goldesberry's defense that he was half-asleep at the time, leading him to mistake K.G. for her mother. It rejected Goldesberry's claim that a dark room prevented him from understanding that he was sexually abusing an eleven-year-old. And (as noted above) it concluded that Goldesberry—an adult male who had been married for nearly twenty years at the time of the incident—must have known the difference between his wife's body and a child's.¹

The majority claims that, even after all of this, “the government ․ still [has] no affirmative evidence of mens rea.” Maj. Op. at ––––. It contends the prosecution's evidence speaks only to the actus reus—and not the mens rea—of the crime. Id. at ––––. This argument falls short. All agree the government presented no direct evidence suggesting Goldesberry knowingly abused his daughter. But as the majority acknowledges, “[t]he government can, and ordinarily does, prove knowledge and intent through circumstantial evidence” alone. Id. at –––– (quoting United States v. Banks, 884 F.3d 998, 1018 (10th Cir. 2018)). I see no reason why the jury could not have examined the circumstantial evidence surrounding the uncontested actus reus— including the proven facts and associated inferences that cast doubt on Goldesberry's theory of the case—to assess the contested mens rea. This practice is as essential as it is intuitive: Absent a direct confession, “proof of knowledge and intent must be based on drawing inferences from the defendant's actions, the testimony and actions of others, and documentary evidence.” United States v. Rufai, 732 F.3d 1175, 1191–92 (10th Cir. 2013) (emphasis added). And in my view, that is exactly what happened in this case.

To the extent the majority relies on the equipoise theory to support its conclusion, I disagree. The equipoise theory requires reversal of a conviction where “the evidence ․ gives equal or nearly equal circumstantial support to a theory of guilt and a theory of innocence.” United States v. Lovern, 590 F.3d 1095, 1107 (10th Cir. 2009) (quotation omitted). Several circuits have rejected this doctrine entirely, and for good reason. See, e.g., United States v. Caraballo-Rodriguez, 726 F.3d 418, 432 (3d Cir. 2013) (en banc); United States v. Vargas-Ocampo, 747 F.3d 299, 301 (5th Cir. 2014) (en banc); United States v. Nevils, 598 F.3d 1158, 1167 (9th Cir. 2010) (en banc). The equipoise theory sits in tension with our standard of deference to the jury: It allows appellate courts “to act as the jury in deciding which inference was the most plausible, rather than asking the proper question ․ whether the jury's inference was merely rational.” Caraballo-Rodriguez, 726 F.3d at 434. Our legal system entrusts jurors, who have the benefit of watching the trial and gaining insights unavailable to us, with decision-making power. And by eliminating the deference typically afforded jury verdicts, the equipoise theory inescapably allows courts to usurp the jury's function in a way that undermines the judicial process. Vargas-Ocampo, 747 F.3d at 301. Perhaps that is why our Circuit has neither applied nor discussed the equipoise theory in the sixteen years since (or at any time before) Lovern.

Even accepting the equipoise theory as doctrinally correct (which I do not), I disagree with its application here because the evidence does not “equally support[ ] the defense's theory (the touching occurred by mistake) and the prosecution's theory (the touching occurred knowingly).” Maj. Op. at –––– n.8. In my view, the government offered ample evidence to support Goldesberry's conviction. And “[a]lthough a different jury might have drawn different inferences, the jury's verdict as to [Goldesberry] represents a permissible view of the evidence.” Lovern, 590 F.3d at 1112 (O'Brien, J., dissenting).

Recall our limited standard of review for sufficiency of the evidence: “[W]e will not overturn a jury's verdict unless no reasonable juror could have concluded, on the basis of the evidence presented, that the defendant was guilty of the crime charged.” Dowlin, 408 F.3d at 657. Under the facts of this case, I am not prepared to say that no reasonable juror could have concluded Goldesberry knowingly abused his daughter. Nor can I join the majority's alternative assertion that “[a]t best, the evidence of knowledge ‘gives equal or nearly equal circumstantial support to a theory of guilt and a theory of innocence.’ ” Maj. Op. at –––– (quoting Lovern, 590 F.3d at 1107). The majority plainly disagrees with how the jury evaluated the evidence. But it cannot substitute its preferred view of the evidence for that of the jury. “Because rational people can sometimes disagree, the inevitable consequence of this settled law [governing sufficiency of the evidence] is that judges will sometimes encounter convictions that they believe to be mistaken, but that they must nonetheless uphold.” Cavazos v. Smith, 565 U.S. 1, 2, 132 S.Ct. 2, 181 L.Ed.2d 311 (2011). I would leave the jury's verdict undisturbed.

II.

Because I would conclude the evidence was sufficient to support Goldesberry's conviction, I now address (and would ultimately reject) his alternative argument for a new trial based on statements that he contends constitute prosecutorial misconduct.

“We analyze whether a statement constitutes prosecutorial misconduct using a two-step process.” United States v. Rodella, 804 F.3d 1317, 1335 (10th Cir. 2015) (quotation omitted). “First, we determine whether the prosecutor's statements were improper.” Id. (quotation and internal quotation marks omitted). “Second, we determine whether the prosecutor's improper statements were harmless beyond a reasonable doubt.” Id. (quotation omitted). “The [g]overnment generally bears the burden of proving that an improper statement is harmless beyond a reasonable doubt.” Id. (quotation omitted). “But when a defendant fails to object to an allegedly improper statement during trial, we review only for plain error and it is the defendant rather than the [g]overnment who bears the burden of persuasion with respect to prejudice.” Id. (quotation omitted). “[U]nless there is reason to believe [the statements] influenced the jury's verdict,” we consider them harmless. United States v. Green, 435 F.3d 1265, 1268 (10th Cir. 2006) (quotation omitted).

Goldesberry contends the prosecution made four plainly improper statements during closing arguments that affected his substantial rights. First, he claims the prosecution wrongly stated that K.G. did not testify she (1) heard Goldesberry snoring when he touched her or (2) saw Goldesberry's eyes were closed when she turned to call out. Goldesberry argues these comments increased the likelihood that the jury would believe he was fully awake when he initiated contact with K.G.

The government concedes these statements inaccurately recounted K.G.’s testimony. Aple. Br. at 45. But it is unlikely they affected the outcome of the proceedings. We have previously held that any prejudice resulting from minor factual errors is cured when the district court instructs the jury to rely on its own recollection of witness testimony above the parties’ summaries. See United States v. Sorensen, 801 F.3d 1217, 1242 (10th Cir. 2015). And here, both the prosecution and the court instructed the jury to do just that. R. Vol. II at 512 (“Your memory of the testimony controls. So rely on your memory.”); id. at 96 (“If any reference by the Court or counsel to the witness’ testimony conflicts with your own recollection, it is your own recollection that should control during your deliberations and not the statement of Court or counsel.”). Further, as in Sorensen, the district court instructed the jury that counsel's statements and arguments are not evidence. 801 F.3d at 1242–43; R. Vol. II at 522. The jury also heard defense counsel's statements that “[Goldesberry] was snoring. [K.G.] has never said anything different than that,” and that “she turned around and looked at her dad and saw that his eyes were still closed.” R. Vol. II at 503. Under these circumstances, the prosecution's misstatements were not flagrant enough to influence the jury's verdict.

Second, the prosecution characterized K.G.’s crawling into bed with her parents as a “routine[ ]” occurrence. R. Vol. II at 487. Goldesberry contends this was a misstatement because K.G. testified she did not “regularly” climb into her parents’ bed. Id. at 173–74, 404. But both parties acknowledge K.G. had gone to her parents’ bed “[s]everal times collectively” after nightmares. Id. at 173. And we have previously held that parties are entitled “to a reasonable amount of latitude in drawing inferences from the evidence during closing arguments.” United States v. Kepler, 74 F.4th 1292, 1316 (10th Cir. 2023) (quotation omitted). Further, even if the prosecution's statement mischaracterized the evidence, it was “not so egregious as to influence the jury to convict.” United States v. Hammers, 942 F.3d 1001, 1016 (10th Cir. 2019) (quotation omitted). It is unlikely that the jury's verdict hinged on whether K.G. “routinely” or “occasionally” entered her parents’ bed. What matters is that the jury heard K.G. had joined her parents in bed multiple times, and that conclusion was supported by witness testimony.

Third, Goldesberry challenges the prosecution's characterization of the family's living arrangement after the abuse. The prosecution argued the family “chose” Goldesberry over K.G. by sending her to live with her aunt and uncle during a difficult time. R. Vol. II at 494. Goldesberry contends that, in making this argument, the prosecution wrongly implied Goldesberry did not care about protecting his daughter despite earlier testimony that the family could not afford for Goldesberry to continue to stay in a hotel. This, he claims, invited the jury to convict him on the theory that he was a bad person.

This argument also fails. The prosecution's statements are not plain mischaracterizations; “[a]rguments may be forceful, colorful, or dramatic, without constituting reversible error.” United States v. Gregory, 54 F.4th 1183, 1210–11 (10th Cir. 2022). Though the prosecution's statement was dramatic, it fell within the latitude afforded parties during closing arguments. Further, the statement speaks more pertinently to the family's bias towards Goldesberry than to Goldesberry's character. This claim aligns with the government's broader trial theme that the family closed ranks around Goldesberry and punished K.G. to prevent her from further disclosures. Accordingly, this statement also does not constitute plain error.

Fourth, Goldesberry argues the prosecution “baselessly insinuated that K.G.’s mental health problems were [ ] Goldesberry's fault.” Aplt. Br. at 44. On several occasions, the prosecution alluded to Goldesberry watching K.G. “deteriorate” after the abuse. R. Vol. II at 512, 516. Goldesberry contends that, in context, it is “clear that the prosecution was (1) attributing K.G.’s mental health problems to the 2017 incident and (2) blaming [ ] Goldesberry's supposed insistence on secrecy for K.G. not getting better.” Aplt. Br. at 44. This, he argues, “distorted the record in multiple ways and [was] thus plainly improper.” Id.

The prosecution's argument has at least some support in the record. Although K.G. and her parents attributed her struggles to other causes, no testimony suggested that she expressed concern with or sought assistance for her mental health difficulties before the incident. And in context, the government contended only that Goldesberry should not have kept quiet about the incident. R. Vol. II at 512 (“[T]he least that he could do, if this truly was a mistake, was not focus on keeping it quiet. The least he could do is not watch for years his child deteriorate in front of his eyes. Watch his child keep this a secret for him.”). Even if the prosecution's statement was improper, there is no reason to believe it influenced the jury's verdict. Accordingly, there is no plain error.

Goldesberry failed to object to any of these alleged misstatements at trial, and he has not satisfied his burden to show prejudice as a result of those statements. Accordingly, I would deny his request for a new trial based on prosecutorial misconduct.

III.

For these reasons, I would AFFIRM Goldesberry's conviction.

FOOTNOTES

1.   Because we conclude the evidence was legally insufficient to support Mr. Goldesberry's conviction, we need not consider his prosecutorial misconduct argument.

3.   The grand jury later returned a superseding indictment adding an alternative charge, Count 2, for sexual abuse of a minor 12 or older. The government added Count 2 after Mr. Goldesberry made clear he would challenge the age element of his offense at trial. The parties disputed whether the incident occurred before or after K.G.’s twelfth birthday in September 2017. The jury convicted Mr. Goldesberry on Count 1. On appeal, Mr. Goldesberry no longer contests the age element. See Reply Br. at 18 (“Mr. Goldesberry can't complain that the evidence was legally insufficient on the age element.”).

4.   District Judge John Antoon II, from the United States District Court for the Middle District of Florida, sat by designation to conduct the trial.

5.   Faith did not testify at trial.

6.   We note our sister circuits are divided on how the equipoise principle informs sufficiency review.Like our circuit, the First, Second, Sixth, Seventh, Eighth, and Eleventh Circuits agree the equipoise principle is a helpful tool in sufficiency review and retain use of the principle in some form. See, e.g., United States v. Simon, 12 F.4th 1, 32 (1st Cir. 2021) (“[T]he equipoise principle is entrenched in this circuit's jurisprudence.”); United States v. Glenn, 312 F.3d 58, 70 (2d Cir. 2002) (“Jackson requires that a rational juror be able to find the defendant guilty beyond a reasonable doubt, and if the evidence viewed in the light most favorable to the prosecution gives equal or nearly equal circumstantial support to a theory of guilt and a theory of innocence, then a reasonable jury must necessarily entertain a reasonable doubt.” (internal quotation marks and citations omitted)); United States v. Caseer, 399 F.3d 828, 840 (6th Cir. 2005) (similar); Cosby v. Jones, 682 F.2d 1373, 1376–83 (11th Cir. 1982) (similar); United States v. Johnson, 592 F.3d 749, 755 (7th Cir. 2010) (“In this situation, the evidence is essentially in equipoise; the plausibility of each inference is about the same, so the jury necessarily would have to entertain a reasonable doubt on the conspiracy charge.”); United States v. Boesen, 491 F.3d 852, 857 (8th Cir. 2007) (stating the equipoise rule requires courts to examine whether “the government's evidence is equally strong to infer innocence as to infer guilt” (quoting United States v. Davis, 103 F.3d 660, 667 (8th Cir. 1996))).The Fifth and Ninth Circuits have decided the equipoise principle is unhelpful in reviewing sufficiency challenges. United States v. Vargas-Ocampo, 747 F.3d 299, 301 (5th Cir. 2014) (en banc) (concluding the equipoise principle “is not helpful in applying the Supreme Court's standard” for sufficiency challenges); United States v. Nevils, 598 F.3d 1158, 1167 (9th Cir. 2010) (en banc) (holding a principle that “required reversal” when innocence “was not any less likely than the incriminating explanation advanced by the government ․ strayed from the” Supreme Court's standard (internal quotation marks and citations omitted)). Of course, these courts still engage in robust sufficiency review. See Vargas-Ocampo, 747 F.3d at 302 (“In abandoning use of the ‘equipoise rule’ in this circuit, we do not render the Jackson standard toothless. On the contrary, courts remain empowered to consider, for instance, whether the inferences drawn by a jury were rational, as opposed to being speculative or insupportable, and whether the evidence is sufficient to establish every element of the crime.”); Nevils, 598 F.3d at 1167 (“In reaching this conclusion, ․ we acknowledge our obligation under Jackson to identify those rare occasions in which ‘a properly instructed jury may ․ convict even when it can be said that no rational trier of fact could find guilt beyond a reasonable doubt[.]’ ” (second and third alterations in original) (quoting Jackson, 443 U.S. at 317, 99 S.Ct. 2781)).The Fourth and D.C. Circuits do not appear to have directly addressed the viability of the equipoise principle. See United States v. Stavrakis, Nos. 20-4149, 20-4184, 2022 WL 563242, at *7 (4th Cir. Feb. 24, 2022) (“We have no occasion to wade into [the vitality of the equipoise principle].”); United States v. Shi, 991 F.3d 198, 213 (D.C. Cir. 2021) (Silberman, J., concurring) (suggesting, but only in a concurrence, the court's decision in Curley v. United States, 160 F.2d 229 (D.C. Cir. 1947) “establishes the rule of equipoise”); Curley, 160 F.2d at 233 (“[I]f, upon the whole of the evidence, a reasonable mind must be in balance as between guilt and innocence, a verdict of guilt cannot be sustained.”).

7.   In United States v. Lovern, we considered a sufficiency challenge to the defendant's conviction for violating the Controlled Substances Act, 21 U.S.C. §§ 801 et seq. See 590 F.3d 1095, 1097–1100 (10th Cir. 2009). Like Mr. Goldesberry, the defendant in Lovern challenged the sufficiency of the evidence on the mens rea element of the statute, which required the government to prove the defendant acted knowingly. See id. at 1104. Reviewing de novo, we concluded the government's evidence amounted to “speculation and conjecture” because it left “a reasonable fact-finder with a number of equally reasonable inferences about” the defendant's mens rea. Id. at 1107, 1109 (quoting United States v. Truong, 425 F.3d 1282, 1288 (10th Cir. 2005)). Put differently, we concluded the evidence was in equipoise. Under those circumstances, reversal was required, we held, because “a reasonable jury must necessarily entertain a reasonable doubt” where “the evidence ․ gives equal or nearly equal circumstantial support to a theory of guilt and a theory of innocence.” Id. at 1107 (alteration in original) (italics omitted) (quoting Caseer, 399 F.3d at 840).The government's answer brief does not discuss the equipoise principle or cite Lovern. Since briefing completed, the government has attempted to seek distance from the equipoise principle and its application here. For the first time at oral argument, the government maintained Lovern should be cabined to its facts. And shortly after oral argument, the government contended in a letter of supplemental authority that a 2013 Third Circuit decision abrogated a case cited in Lovern. We must conclude the government has waived these late-blooming arguments. United States v. Gaines, 918 F.3d 793, 800–01 (10th Cir. 2019) (holding we generally “decline to consider an appellee's contentions raised for the first time in oral argument”); Standing Akimbo, LLC v. United States ex rel. Internal Revenue Serv., 955 F.3d 1146, 1158 n.9 (10th Cir. 2020) (finding arguments “waived because they were improperly raised for the first time in Rule 28(j) letters”). In any event, as Mr. Goldesberry persuasively points out, the government does not dispute Lovern is a precedential decision of this court. Accordingly, we evaluate the parties’ arguments by applying the familiar sufficiency standard, including the equipoise principle expressed in Lovern.

8.   We agree with Mr. Goldesberry that the evidence—at best—equally supports the defense's theory (the touching occurred by mistake) and the prosecution's theory (the touching occurred knowingly). With or without the rule of equipoise, every sufficiency of the evidence challenge presents the same question: “whether the record evidence could reasonably support a finding of guilt beyond a reasonable doubt.” Jackson, 443 U.S. at 318, 99 S.Ct. 2781. And here, reversal is required even without regard to equipoise. As we will explain, the record permits nothing more than “speculative or insupportable” inferences of mens rea and thus cannot support the conclusion that the government “establish[ed] every element of the crime.” Vargas-Ocampo, 747 F.3d at 302; see id. (concluding the equipoise principle is not a helpful tool in administering the Jackson standard but acknowledging “courts remain empowered to consider, for instance, whether the inferences drawn by a jury were rational, as opposed to being speculative or insupportable, and whether the evidence is sufficient to establish every element of the crime.”); Nevils, 598 F.3d at 1167 (concluding the equipoise principle is not a helpful tool in administering the Jackson standard but acknowledging the evidence construed in favor of the government “may still be so supportive of innocence that no rational juror could conclude that the government proved its case beyond a reasonable doubt”).

9.   The jury could have reasonably inferred Mrs. Goldesberry kissed Mr. Goldesberry goodbye, which would have given him a reason to know she was gone. But inferring knowledge beyond a reasonable doubt from this predicate inference would require impermissible guessing. It would require further inferring (1) Mrs. Goldesberry's good-bye kiss woke up Mr. Goldesberry in the first place; (2) Mr. Goldesberry was awake enough during the incident to remember Mrs. Goldesberry's kiss; and (3) Mr. Goldesberry had reason to believe it was more likely that his daughter entered his bed than that his wife had returned or that he misremembered her leaving. Nothing supports this string of inferences—as opposed to, say, the inference Mrs. Goldesberry's kiss never woke up Mr. Goldesberry; or the inference Mr. Goldesberry was half-asleep and did not recall Mrs. Goldesberry had left, or the inference Mr. Goldesberry felt a woman next to him and assumed his wife had returned. See Lovern, 590 F.3d at 1107 (“At best, the instant message leaves a reasonable fact-finder with a number of equally reasonable inferences ․”). An inference of knowledge based on this string of individual inferences would also violate the rule that “we may not uphold a conviction obtained by piling inference upon inference.” United States v. Anderson, 189 F.3d 1201, 1205 (10th Cir. 1999) (quoting United States v. Valadez-Gallegos, 162 F.3d 1256, 1262 (10th Cir. 1998)); see Rosenberg v. United States, 120 F.2d 935, 937 (10th Cir. 1941) (“[A]n inference of fact which is essential to the establishment of the offense cannot be rested upon another inference.”).

10.   We are also mindful not to conduct our sufficiency review “in bits and pieces.” United States v. Rodriguez-Flores, 907 F.3d 1309, 1312 (10th Cir. 2018) (quoting United States v. Brooks, 438 F.3d 1231, 1236 (10th Cir. 2006)). We do not question that individual pieces of evidence, alone insufficient to support an inference of knowledge, might together support such an inference. That is not the case here. No pieces of evidence begin to carry the government's burden, so they do not carry that burden together.

11.   At oral argument, the government stated Mr. Goldesberry's statements to Ms. Gusman and Mr. Bartlett were its best evidence of mens rea.

12.   The government seems to suggest K.G.’s self-harm is further evidence of Mr. Goldesberry's mens rea. Ans. Br. at 34. Again, the government relies on impermissible speculation to support the inference of knowledge. Ms. Blevins testified “in my training and experience, I often speak with kids who are experiencing self-harming and suicidal ideation.” App. II at 272. But Ms. Blevins did not say self-harm is distinctive to victims of sexual abuse nor did any evidence suggest that self-harm is more likely when an inappropriate touching occurs knowingly than by mistake.

1.   K.G.’s testimony that her father acted by mistake—which, the majority argues, supports Goldesberry's theory of the case, Maj. Op. at ––––—does nothing to undermine the jury's conclusion. The jury heard that Child Protective Services recommended K.G. undergo “trauma-informed therapy” as a result of her father's sexual touching. R. Vol. II at 364. It heard expert testimony that child victims often “minimize” abuse, delay disclosure, or deny the abuse altogether out of fear, shame, or guilt, id. at 263–64—and it saw K.G. demonstrate some of these characteristics, id. at 144 (“I feel like it should have been my responsibility to wake him up.”). It heard that K.G. engaged in self-harm after the incident, that she eventually confided in Ms. Bartlett, and that her parents sent her to live somewhere else after Ms. Bartlett reported the abuse. And it heard that Goldesberry confessed “he had messed up[,] and messed up real bad.” Id. at 225. Based on these facts, the jury reasonably determined that the circumstantial evidence of Goldesberry's mens rea proved guilt beyond a reasonable doubt despite K.G.’s contrary testimony.

ROSSMAN, Circuit Judge.