Full opinion text
Brennan, Circuit Judge. Darryl Christensen, a Polk County, Wisconsin Jail corrections officer, sexually assaulted plaintiffs J.K.J. and M.J.J. over three years during their incarcerations. Plaintiffs sued Christensen and the county under 42 U.S.C. § 1983, alleging Eighth and Fourteenth Amendment claims, in addition to a state law negligence claim against the county. After trial, the jury found Christensen and the county liable for J.K.J. and M.J.J.'s injuries and awarded each $ 2 million in compensatory damages. The jury also levied punitive damages against Christensen, awarding $ 3,750,000 to each plaintiff. Both defendants moved for new trials, and the county also moved for judgment as a matter of law. The district court denied those requests and defendants now appeal the judgments entered against them. We see no reason to disturb the jury's verdict against Christensen and so affirm the denial of his request for a new trial. His assaults were predatory and knowingly criminal. But to impose liability against the county for Christensen's crimes, there must be evidence of an offending county policy, culpability, and causation. These are demanding standards. Christensen's acts were reprehensible, but the evidence shows no connection between the assaults and any county policy. We therefore reverse and remand for entry of judgment in favor of the county. I. BACKGROUND A. Christensen's Sexual Assaults M.J.J. and J.K.J. were inmates at Polk County Jail at various times between 2011 and 2014. Christensen admits he engaged in sexual acts with the women individually. To hide his offenses, Christensen planned his encounters to occur when no one was present and in locations where he controlled access. He also urged plaintiffs not to discuss or report his sexual advances because he would lose his job and family if caught. Plaintiffs complied with Christensen's secrecy directive and his assaults were kept hidden from jail officials. Polk County authorities discovered Christensen's assaults against M.J.J. and J.K.J. after a former inmate reported her own sexual encounters with Christensen to an investigator in a neighboring county. When notified of the former inmate's allegations, county authorities initiated an internal investigation and confronted Christensen, who immediately resigned. The investigation continued, which led to the discovery of Christensen's abuse of plaintiffs, and ultimately to his prosecution. He eventually pleaded guilty to several counts of sexual assault and is serving a 30-year prison sentence. B. Trial Evidence Plaintiffs sued the county and Christensen in separate actions and the cases were consolidated for jury trial. Plaintiffs alleged that defendants were deliberately indifferent to a serious risk of sexual assault in violation of their Eighth and Fourteenth Amendment rights, and that the county violated state law by negligently supervising Christensen. At trial, Christensen admitted his offenses but challenged the harms plaintiffs suffered. He argued plaintiffs consented to his overtures and that their encounters were the product of "voluntary attraction." Although not stated directly, his position implied that any award of damages should correspond to plaintiffs' level of consent. Plaintiffs denied consenting to Christensen's advances and offered expert testimony showing their mental trauma from his assaults. Against the county, plaintiffs made four principal allegations: (1) the jail's sexual assault policies and training were inadequate; (2) the jail customarily tolerated sexually offensive comments by guards; (3) the investigation of a former guard revealed the jail's sexual assault policy was inadequate and that the jail minimized sexual abuse; and (4) the jail failed to widely implement recommendations under the Prison Rape Elimination Act (PREA), 34 U.S.C. §§ 30301 -09. The sum of these allegations, plaintiffs argued, prove the county was deliberately indifferent to a known risk of sexual assault by jail staff. The county disagreed, arguing that the trial evidence did not support the jury's liability finding and damages awards. These claims were heavily contested, and we recount the evidence noting those facts the county disputed. Although we summarize the trial evidence, on appeal we view the facts in the light most favorable to the jury's verdict. See Lindsey v. Macias , 907 F.3d 517, 518 n.1 (7th Cir. 2018). 1. Policies and training Plaintiffs alleged the jail had no policy either to prevent or detect sexual assaults, and that its policies on sexual misconduct were "practically nonexistent." At trial, the county produced several policies prohibiting sexual contact between guards and inmates, and two stand out. Policy I-100 forbids any mistreatment or harassment of inmates, explaining inmates' rights and informing them that it is never acceptable for "any inmate [to] be the object of verbal, physical, emotional, psychological, or sexual harassment by facility staff." The policy continues, "[a]ny officer engaged in such actions is subject to disciplinary charges and/or termination." Inmates are also provided a handbook when booked into the jail that says: Every inmate has the right to be safe from sexual abuse and harassment. No one has the right to pressure you to engage in sexual acts. If you are being pressured threatened, or extorted for sex, you should report this to staff immediately. Likewise, Policy C-202 prohibits any "intimate social or physical relationship with a prisoner." It also informs jail staff that sexual contact with any inmate is a criminal offense under Wisconsin law, and any officer that suspects such conduct has a duty to report it. See Wis. Stat. § 940.225(2)(h) (categorizing sexual contact and sexual intercourse by a correctional staff member with an inmate as a Class C felony). Plaintiffs also claimed the county never trained officers to avoid sexual assaults. But the jail's onboarding and continuing education programs instruct employees that sexual contact with prisoners is a crime and never permitted. The Wisconsin Department of Corrections (DOC) approved these programs annually, requiring: (1) eight to ten weeks of "field training," during which a new corrections officer shadows an experienced officer to learn jail policies and procedures; (2) completion of a 160-hour jail training program to become a certified corrections officer; (3) 24 hours of continuing education each year to be recertified; and (4) daily training, which includes specific training on the jail's prohibition against fraternizing with inmates. At trial, Christensen acknowledged the jail trained him that sexual contact with inmates is a felony and against jail policies. Specifically, Christensen testified: • He knew his conduct violated jail policy; • He was trained his conduct was a crime; • He knew he was putting plaintiffs at risk; • He never forgot that sex with inmates was a crime; and • He agreed he did not require more training to know his conduct was a crime. Plaintiffs' expert witness on prison training standards, Jeffrey Eiser, testified that the jail's policies prohibited sexual contact between inmates and guards. Eiser also corroborated that the county trained Christensen that sexual contact with inmates was a felony and against jail policy. To support their contention that the jail never trained its staff, plaintiffs relied on two witnesses. The first, Lynelle Manning, was a jailer with the county for about 20 months. Manning testified that although she was never officially certified as a correctional officer, she received "formal training" by the jail and shadowed a senior officer for weeks. She also received and read the jail's policy and inmate booking manuals, which contain the jail's prohibition of sexual contact between guards and inmates. Manning also testified that during her employment she never heard sexually charged conversations between jail staff and inmates. Plaintiffs' second witness, Sergeant Steven Schaefer of the county's sheriff's department, worked at the jail from 2002 until 2015. Schaefer testified "we were all required to attend" countywide training on sexual harassment. He provided the training to new employees from time to time. According to Schaefer, that training instructed on the jail's numerous prohibitions between staff and inmates, including improper comments, becoming too close or too familiar, sharing personal information, and sexual relationships. He also agreed that improper relationships between inmates and guards were "something that the jail as a whole took very seriously." Notwithstanding Schaefer's testimony, plaintiffs' counsel told the jury during closing argument: "You heard Sergeant Schaefer say, 'We never trained on it. We never trained on it. We never trained on it.' " 2. Inappropriate speech Next, plaintiffs alleged that jail staff routinely made sexually inappropriate comments about female inmates without repercussions. According to plaintiffs, Captain Scott Nargis, who oversaw daily operations of the jail, was the reason that sexually offensive speech was accepted at the jail. During adverse examination, plaintiffs' counsel asked Nargis if he ever "engaged in tier talk which is not necessarily flattering talk amongst co-workers"; Nargis answered "yes." Nargis also agreed that he participated in tier talk "on occasion" to establish trust among subordinate officers. Plaintiffs never asked Nargis on the witness stand if he himself made sexual comments. Nor did plaintiffs present evidence that tier talk connoted "sexual talk," that Nargis's "tier talk" was sexually explicit, or that Nargis made comments sexual in nature with, about, or around inmates or guards. Evidence to suggest Nargis knew about offensive comments by jail staff was scarce and unclear as to timing. Nargis testified that during Christensen's twelve-year employment, he once heard Christensen comment on a female's "rear end." He did not recall whether that comment was made about an inmate. Nargis also recalled being told that Christensen once remarked about an inmate's breasts. Evidence of inappropriate sexual comments by other jail staff was also sparse and unspecific. J.K.J. testified she believed two other corrections officers once overheard Christensen making flirtatious comments to inmates. Christensen also testified to overhearing a jail guard, Allen Jorgenson, and two other guards make suggestive comments to inmates. But J.K.J. and Christensen offered no specifics on the alleged comments, and there was no evidence these incidents were reported to the county or any jail supervisor. 3. Investigation of former guard At trial, plaintiffs introduced one other allegation of sexual contact between a jail guard, Jorgenson, and an inmate, N.S.: another inmate saw Jorgenson put his arm around N.S.'s waist and "pat her on the butt." This occurred in 2012, two years before the discovery of Christensen's violations. Sergeant Steven Schaefer reported these allegations to Nargis, who in turn questioned Jorgenson and N.S. individually. Each denied any improper relationship or contact. Despite these denials, Nargis requested the assistance of chief deputy sheriff Steven Moe to further investigate Jorgenson. To plaintiffs, the Jorgenson investigation proves the county "minimized" and ignored allegations of a guard assaulting an inmate. At trial, the jury considered the findings of the Jorgenson investigation, including Jorgenson's interactions with N.S. Another inmate believed Jorgenson and N.S. had an "inappropriate relationship" but "no physical relationship." It was also reported that Jorgenson misused a jail camera to focus on inmates longer than necessary. In addition to an internal investigation, Nargis and Moe reached out to former inmates as part of their review. Because of inconsistent witness accounts, Nargis and Moe could not confirm that Jorgenson engaged in any sexual contact with N.S. Still, Nargis and Moe concluded that Jorgenson's affiliation with N.S. violated jail policy. As a result, Jorgenson was issued a written reprimand for "foster[ing] a friendship relationship" by giving "undue, unfair, or simply too much attention" to N.S., who continued to deny any improper actions or relationship up to the point of Jorgenson's reprimand. After Jorgenson was written up, N.S. recanted her denials in a letter to Nargis. In response, Nargis and Moe reopened the investigation "to take a whole fresh look at the situation." N.S.'s letter detailed that Jorgenson made sexually harassing gestures and crude and indecent remarks, and asserted allegations of Jorgenson putting his arm around N.S.'s waist and touching her "back and butt." After this second review, Nargis and Moe could not confirm these allegations and decided the reprimand remained the appropriate level of discipline. At trial, no evidence was submitted that Nargis or Moe erred in the Jorgenson investigation or performed their inquiries in bad faith. In closing, plaintiffs' counsel argued to the jury that the "jail knew that one of their trusted friends was committing sexual assault against at least one inmate, N.S." but considered it "no big deal." Jorgenson also made inappropriate remarks, of which inmates and staff were aware. But there was no evidence Jorgenson's improper comments were reported to Nargis, Moe, or any county policymaker before the N.S. investigations. On that point, the county argued the N.S. allegations prompted complaints by various female coworkers, who claimed Jorgenson made inappropriate comments to them as well. Those coworker complaints led to a human resources investigation that resulted in Jorgenson resigning. 4. Prison Rape Elimination Act (PREA) The county's sexual assault policies were inadequate to prevent and detect assaults, and the county deliberately avoided opportunities to reduce sexual assault risks, according to plaintiffs. Both arguments were based on the county's purported underutilization of policy recommendations from PREA. Again, plaintiffs zeroed in on Nargis. They claimed the jail intentionally shunned PREA because Nargis openly "denigrated ... PREA standards," citing a 2014 email from Nargis to jail staff about PREA training: Seems to be that everyone is in a tizzy to train their staff on PREA. There is no requirement for use [sic] to be compliant with everything that the law calls for, but nevertheless it is federal law. So we'll hit the basics of PREA training. At trial, plaintiffs termed this "the tizzy email." To plaintiffs, Nargis's choice of the word "tizzy" was "mocking" PREA and "indicat[ed] that he disliked PREA." They also claimed the email never discussed any specific PREA measures. Rather, it merely restated the jail's current anti-sexual assault policies. Plaintiffs argued "the tizzy email" proves that Nargis and the jail "consciously disregarded" PREA standards, and by extension, disregarded the risk of sexual assaults at the jail. Plaintiffs' expert Eiser opined that the jail's sexual assault policies and training were inadequate because they did not fully adopt certain components of PREA. Eiser conceded compliance with PREA is not mandatory for county jails in Wisconsin, and that PREA standards are better viewed as optional "best practices." Eiser also testified there is no empirical data that compliance with the proposed best practices would yield a better result. Plaintiffs agree that state law, not PREA, governs county jails in Wisconsin, but did not offer evidence that the jail's sexual assault policies or training fell below state legal or administrative standards. As for compliance with state law, the county argued the DOC annually reviews the jail's policies, including its policy prohibiting fraternization with inmates. In each year of plaintiffs' incarcerations, the DOC found the jail to be in full compliance with all applicable Wisconsin statutes and regulations. Language addressing PREA was added to the jail's anti-fraternization policy in 2012, with an accompanying PREA training in 2014. The county also noted that in the past nine years, during which the jail housed 14,100 inmates, Jorgenson's circumstance was the only known improper relationship between a guard and an inmate. C. Verdict and Post-Verdict Motions The district court bifurcated the trial into liability and damages phases. At the close of the liability phase, during the jury instruction conference, the court found the evidence failed to show a pattern of constitutional violations known to county policymakers. As a result, the court excluded this basis of liability from the jury instructions, leaving plaintiffs to argue that the "risk of the inadequacy of the training, supervision, and/or adoption of policies [was] plainly obvious." The court also rejected a jury instruction as to whether plaintiffs consented to Christensen's sexual contact and thus reduced plaintiffs' harm. After a five-day trial, the jury found for plaintiffs on all claims and awarded each plaintiff $ 2 million in compensatory damages against the county and Christensen. The jury also awarded $ 3,750,000 to each plaintiff in punitive damages against Christensen. After the verdict, Christensen moved for a new trial under FED. R. CIV. P. 59. Christensen argued there was insufficient proof that he harmed plaintiffs or was aware of the substantial risk of harm his actions imposed. The district court rejected these arguments, relying on plaintiffs' testimony that they never consented to Christensen's advances. The county also moved for judgment as a matter of law under Rule 50(b) and for a new trial under Rule 59. This resulted in partial yet hollow success for the county-the district court granted judgment to the county on plaintiffs' state law negligence claims, but denied the county judgment on plaintiffs' § 1983 claims, as well as a new trial. The district court rejected the county's contention that plaintiffs failed to prove it was culpable for and the cause of Christensen's violations, as required for liability under Monell v. New York City Dep't of Soc. Servs , 436 U.S. 658, 98 S.Ct. 2018, 56 L.Ed.2d 611 (1978). Although the court noted the evidence against the county was "not overwhelming," it concluded three subjects supported the verdict: (1) jail supervisor Nargis "generally acknowledged his awareness of tier talk" at the jail; (2) Nargis was aware of sexual comments by correctional officers to inmates and female employees through the Jorgenson investigation, in addition to two comments made by Christensen; and (3) the county held only one PREA training session. For the district court, this was sufficient evidence for the jury to find "that Nargis and others within the [county] ... acted with deliberate indifference to the need for better training, supervision and policies." So the verdict against the county remained intact. II. DISCUSSION The county and Christensen both argue the district court improperly denied them judgment as a matter of law or a new trial under Rules 50 and 59. At the outset, we note Christensen never filed a post-verdict motion for judgment as a matter of law under Rule 50. Without such a motion, he forfeited his request for judgment under that rule, and our review is limited to his request for a new trial under Rule 59. See Unitherm Food Sys., Inc. v. Swift-Eckrich, Inc. , 546 U.S. 394, 400-01, 404-05, 126 S.Ct. 980, 163 L.Ed.2d 974 (2006) ; accord Collins v. Lochard , 792 F.3d 828, 831 (7th Cir. 2015). A district court may enter judgment as a matter of law under Rule 50 when it "finds that a reasonable jury would not have a legally sufficient evidentiary basis" to support its verdict. FED. R .CIV. P. 50(a)(1) ; see also Rule 50(b). We review the denial of a Rule 50 motion de novo and proceed "on the basis of the evidence the jury actually had before it." Houskins v. Sheahan , 549 F.3d 480, 493 (7th Cir. 2008) (internal citation omitted) (denying Monell claim). We will overturn a jury verdict if it is clear plaintiffs failed to present enough evidence to support their claims. Id . (citing Filipovich v. K & R Express Sys., Inc. , 391 F.3d 859, 863 (7th Cir. 2004) ). "Our job is to assure that the jury had a legally sufficient evidentiary basis for its verdict." Filipovich , 391 F.3d at 863. Under Rule 59, a district court may order a new trial "for any reason for which a new trial has heretofore been granted in an action at law in federal court." FED. R. CIV. P. 59(a)(1)(A). A new trial is appropriate if the jury's verdict is against the manifest weight of the evidence or if the trial was in some way unfair to the moving party. Martinez v. City of Chicago , 900 F.3d 838, 844 (7th Cir. 2018) (citation and quotation marks omitted). We will not disturb a district court's Rule 59 decision except under exceptional circumstances showing a clear abuse of discretion. Id . First, we consider whether the district court improperly refused to grant the county's motion for judgment as a matter of law. Later, we turn to Christensen's claim that he is entitled to a new trial. A. The County The county argues Monell precludes the jury's finding of § 1983 liability against it. Under Monell , "a municipality can be found liable under § 1983 only where the municipality itself causes the constitutional violation at issue." City of Canton, Ohio v. Harris, 489 U.S. 378, 385, 109 S.Ct. 1197, 103 L.Ed.2d 412 (1989) (emphasis in original); see also Connick v. Thompson , 563 U.S. 51, 60, 131 S.Ct. 1350, 179 L.Ed.2d 417 (2011) ("[U]nder § 1983, local governments are responsible only for their own illegal acts." (emphasis in original)). Thus, a municipality cannot be held liable under § 1983 solely because one of its agents or employees may have violated an individual's constitutional right. Monell , 436 U.S. at 691, 694, 98 S.Ct. 2018 (rejecting § 1983 liability predicated on theory of respondeat superior ). Instead, a municipality's own policy or custom must have caused the constitutional violation. Id. ; Glisson v. Indiana Dep't of Corr. , 849 F.3d 372, 379 (7th Cir. 2017) (en banc) ("The central question is always whether an official policy ... caused the constitutional violation."). To establish municipal liability under Monell , a plaintiff must prove three things. First is the existence of an unconstitutional policy. This can be done by showing either: (a) an express policy that, when enforced, causes a constitutional deprivation; (b) a widespread practice that, although not authorized by written law or express policy, is so permanent and well settled as to constitute a custom or usage with the force of law; or (c) that the constitutional injury was caused by a person with final decision policymaking authority. Spiegel v. McClintic , 916 F.3d 611, 617 (7th Cir. 2019). Second is that the municipality is culpable, which means the municipality's policymakers were deliberately indifferent to a known or obvious risk that a policy or custom would lead to constitutional violations. Board of Comm'rs of Bryan Cty. v. Brown , 520 U.S. 397, 407, 410, 117 S.Ct. 1382, 137 L.Ed.2d 626 (1997) (citing Canton , 489 U.S. at 388, 109 S.Ct. 1197 ). Third, the municipality's policy must "directly cause[ ] a deprivation of federal rights." Id . at 415, 117 S.Ct. 1382. In other words, the county's own actions must be the "moving force" behind plaintiff's injuries. Id. at 404, 117 S.Ct. 1382. An unconstitutional policy can include implicit policies, or a gap in expressed policies. Daniel v. Cook Cty. , 833 F.3d 728, 734 (7th Cir. 2016) (citations omitted). Either way, plaintiffs must prove an actual policy is at issue, not a random event. See Calhoun v. Ramsey , 408 F.3d 375, 380 (7th Cir. 2005). Here, plaintiffs alternate between arguing Christensen's violations were the byproduct of an implicit policy, reflected in the jail's alleged widespread practice and custom of "allow[ing] and encourag[ing] inappropriate sexual behavior," and purported gaps in the county's express sexual assault policies, reflected in the absence of PREA measures. At trial, plaintiffs advanced a number of theories of the county's liability under Monell . We start with plaintiffs' claim that the county was deliberately indifferent to a known and obvious risk that its express policies would lead to, and in fact caused, Christensen's assaults. Then, we consider whether plaintiffs' general allegations against the county amount to an implicit policy, i.e., a widespread practice or custom, that permits sexual misconduct by jail staff; and if so, whether the county's deliberate indifference to that policy caused Christensen's assaults. Next, we examine plaintiffs' contention that the county was deliberately indifferent to the need for more training and supervision, causing plaintiffs' injuries. After that, we address whether Christensen's constitutional violations were a highly predictable consequence of the jail's failure to train its staff allowing for single-incident liability. Last, we consider the sufficiency of the evidence in light of the county's motion for judgment as a matter of law, which the district court denied. 1. Express policies The express terms of the jail's sexual assault policies (I-100 and C-202, described above) are not constitutionally suspect, per plaintiffs. Instead, they challenge alleged gaps or omissions in those policies. Plaintiffs contend the policies do not provide adequate measures to prevent and detect sexual assault-namely, measures suggested by PREA. "[T]he absence of a policy might reflect a decision to act unconstitutionally, but the Supreme Court has repeatedly told us to be cautious about drawing that inference." Calhoun , 408 F.3d at 380 (citing Bryan Cty. , 520 U.S. at 409, 117 S.Ct. 1382 and Canton , 489 U.S. at 388, 109 S.Ct. 1197 ). On this theory, the county's culpability hinges on whether it was deliberately indifferent to an obvious need to update or enhance its sexual assault policies. Liability could be established by showing the county adhered to a policy that it knew, or should have known, failed to prevent assaults by jail staff. But plaintiffs supplied no evidence of policy violations putting the county on notice of a sexual assault problem to resolve or act upon. Cf. Woodward v. Correctional Med. Servs. of Ill., Inc. , 368 F.3d 917, 930 (7th Cir. 2004) (affirming judgment under Monell where evidence showed contractor violated the defendant's express policies and repeatedly failed to act in the face of known violations). Even plaintiffs' expert, Eiser, conceded the Jorgenson incident did not downgrade the jail's good record on this front. To try to address this lack of evidence, on appeal plaintiffs interline a proposition in Glisson with bracketed materials: "the existence of the [PREA] Guidelines, with which [Nargis] was admittedly familiar, is evidence that could persuade a trier of fact that [Polk County] consciously chose the approach that it took." Brief of Plaintiffs-Appellees at 33, J.K.J. v. Polk Cty ., Nos. 18-1498 and 18-2170 (7th Cir. Oct. 3, 2018), quoting Glisson , 849 F.3d at 380. But Glisson concerned whether a defendant "had a policy to eschew any way of coordinating [health] care." Id . at 381. Here, the county enforced, rather than avoided, written policies prohibiting any form of sexual contact. Plaintiffs' Glisson "parallel" ignores that municipal fault still must be established. Without knowledge of an obvious risk, plaintiffs' argument is unavailing. Plaintiffs fare no better blaming the county for underuse of PREA. Their argument implies PREA binds states to implement and enforce its guidelines. But PREA, a federal statute, imposes no such obligations on county-run jails. See 34 U.S.C. §§ 30301 -09. Wisconsin law governs the county's sexual assault policies, and its prohibition of sexual contact between guards and inmates is absolute. WIS. STAT. § 940.225(2)(h) ; see also Ramos v. Hamblin , 840 F.3d 442, 444 (7th Cir. 2016) (noting Wisconsin DOC rules corresponding with PREA make the prevention of prison rape a "priority concern"). Likewise, the county's policies prohibit such contact, and the county showed that its policies fully complied with all applicable Wisconsin statutes and regulations. Wisconsin may elect to adopt or fully comply with PREA standards. See 34 U.S.C. § 30307(e)(2)(A). But where Wisconsin has not incorporated components of PREA into its laws and regulations, it is beyond the role of federal courts to render those components a constitutional requirement. See Bryan Cty. , 520 U.S. at 415, 117 S.Ct. 1382 (admonishing that "[a] failure to apply stringent culpability and causation requirements raises serious federalism concerns" and "risks constitutionalizing" requirements states have not chosen to impose). "[D]eliberate indifference is a stringent standard of fault," and not even a showing of heightened negligence will suffice to establish liability. Bryan Cty. , 520 U.S. at 407, 410, 117 S.Ct. 1382. Proof is required that a county policymaker disregarded a known or obvious consequence of his action or inaction. Id . at 410, 117 S.Ct. 1382 ; see also Connick , 563 U.S. at 61, 131 S.Ct. 1350 (citation omitted). Here, PREA might have been relevant to show a conspicuous flaw in the county's policies, but the record shows no evidence of such a flaw. When Christensen assaulted plaintiffs, the jail had no history of sexual assaults and operated under zero-tolerance sexual assault policies. Eiser's ratification of the jail's "good record" before the jail learned of Christensen's sexual assaults shows that the jail reasonably relied on the effectiveness of its express policies. In the absence of other evidence, we conclude no rational jury could infer that the jail's express policies were obviously deficient, that the county was or should have been aware that jail policies were inadequate, that assaults were imminent, or that PREA sets the norm. As for causation, plaintiffs offered no facts at trial from which the jury could conclude that a gap in the county's express sexual assault policies caused their injuries. See Bryan Cty. , 520 U.S. at 406, 117 S.Ct. 1382 (observing that challenges to a facially valid municipal policy "present much more difficult problems of proof"). Nor on appeal do plaintiffs point to any such facts. The Supreme Court demands that courts "carefully test the link between the policymaker's inadequate decision and the particular injury alleged." Id . at 410, 117 S.Ct. 1382. After examining the express policies here, we cannot conclude the county was culpable for, or its sexual assault policies caused, Christensen's assaults. 2. Implicit policy The county's real policy was to ignore its policies, according to plaintiffs, as shown by Nargis's admission of tier talk, the tizzy email, inappropriate staff comments, and the Jorgenson investigation. Plaintiffs contend the sum of these improprieties resulted in a widespread practice of allowing and encouraging sexual misconduct. Nargis was a focal point of plaintiffs' implicit policy claim, beginning with the allegation that he promoted a "toxic culture" by participating in tier talk and acceding to offensive remarks by jail staff. In our de novo review of the record, however, this allegation lacks support. The only reference at trial to "tier talk" occurred during plaintiffs' examination of Nargis. And when Nargis admitted to "tier talk," he did so only within plaintiffs' limited definition ("not necessarily flattering talk"). Plaintiffs failed to include that their definition encompassed a sexual subtext. Rather, plaintiffs grafted a sexual connotation onto the term after trial in response to the county's appeal. The record also contains no evidence that Nargis's tier talk was sexually explicit, profane, or insensitive. Despite this evidentiary void, plaintiffs mischaracterize the record in their response on appeal: "Captain Nargis routinely engaged in sexually explicit 'tier talk.' " Our dissenting colleague concludes that a reasonable jury could find that Nargis's tier talk was sexual in nature. But we believe this inference relies on plaintiffs' post-trial rebranding of the phrase. Although we view the facts in the light most favorable to the jury's verdict, we are not required to draw unreasonable inferences. Tindle v. Pulte Home Corp. , 607 F.3d 494, 496 (7th Cir. 2010). Only reasonable inferences may be considered. See Hermes v. Hein , 742 F.2d 350, 353 (7th Cir. 1984). At trial, in questioning other witnesses about indecent remarks, plaintiffs routinely and frequently used concrete and specific terms such as "sexual comments," "inappropriate comments," and "sexual harassment." None of these defined terms were used when plaintiffs questioned Nargis about tier talk. We therefore cannot assume that Nargis's admission of "not necessarily flattering talk" means "sexual talk." Instead, we take plaintiffs at their word that "tier talk" means what they told Nargis it means. That Nargis conceded to nondescript tier talk does not prove that he promoted a "toxic culture." Nor does it prove the county ignored relevant policies. The claim based on the "tizzy email" that Nargis mocked and disliked PREA also does not help plaintiffs. Even if we assume Nargis on one occasion discredited PREA, this does not constitute a policy of permitting sexual assaults. Nor can we infer that a supervisor's one-time use of a condescending noun ("tizzy") establishes a conscious disregard for measures to prevent sexual assaults. As for inappropriate remarks by staff, plaintiffs introduced the following evidence: (1) J.K.J. testified two officers overheard Christensen making flirtatious comments to inmates; (2) Christensen testified he overheard three guards make suggestive remarks to inmates; and (3) Nargis knew of two inappropriate remarks made by Christensen over a twelve-year period. We consider whether this proof reflected an implicit policy under the applicable law. Monell claims based on an unconstitutional implicit policy require proof of a "widespread" practice. See Bryan Cty. , 520 U.S. at 404, 117 S.Ct. 1382 ; see also Monell , 436 U.S. at 692, 98 S.Ct. 2018 (defining a widespread practice as one that is "persistent," "permanent," and "well settled"). "[P]roof of isolated acts of misconduct will not suffice; a series of violations must be presented to lay the premise of deliberate indifference." Palmer v. Marion Cty. , 327 F.3d 588, 596 (7th Cir. 2003) (citation omitted) ("A series of violations is necessary"; two incidents in one year is not enough). The offensive comments overheard by J.K.J. and Christensen are unclear; no evidence was adduced as to exactly what was said, the context of the remarks, or when they were said. Nor were these comments reported to Nargis or any other jail supervisors. These two allegations of unreported and undefined remarks (outside of those later learned during the Jorgenson investigation) here are insufficient to show a widespread custom or practice. See Doe v. Vigo Cty., Indiana , 905 F.3d 1038, 1045 (7th Cir. 2018) (holding three incidents of sexual contact, two instances of inappropriate remarks, two allegations of sexual harassment, and one example of cornering an employee for sex, taken together, failed to establish a widespread county practice). Likewise, Nargis's personal knowledge of two inappropriate remarks by Christensen over twelve years falls short of indicating a widespread unconstitutional practice. See Palmer , 327 F.3d at 596 (concluding personal knowledge of two incidents of misconduct by officers in a period of one year is insufficient to indicate a widespread practice). Given the law's requirements, including as to quantity and to frequency, we do not conclude these suggestive and inappropriate remarks amounted to a "widespread practice" so as to constitute an unconstitutional implicit policy. Next, plaintiffs contend the Jorgenson investigation gave the county sufficient notice that its sexual assault policies were deficient. The allegations of Jorgenson's improper contact-including putting his arm around an inmate's waist, and patting her backside-while on the same spectrum of sexual harassment and assault as Christensen's conduct, are not of the same degree as Christensen's repeated and coercive sexual abuse. See Vigo Cty. , 905 F.3d at 1045 (distinguishing between sexual harassment and the trauma of sexual assault). Nevertheless, the record shows the jail responded equally to these two incidents. After two investigations, the jail found it "probably more likely" Jorgenson engaged in "inappropriate touching," but it was unable to confirm the allegations. See id . at 1047 (holding no breach of duty to plaintiff where "the County investigated but could not substantiate one vague complaint against Gray, and it warned him for making sexualized comments to a coworker"). Unverified allegations of inappropriate touching of and humiliating comments toward one inmate over a nine-year period falls short of establishing a widespread practice or custom. See id. at 1046-47 ("A business is not alerted to the possibility that an employee might rape a member of the public by having faced the occasional, but unfortunate, predicament of employee sexual harassment, including groping."). We also note, once more, that plaintiffs' expert agreed the jail had a good record on this topic, including when considering the Jorgenson incident. The evidence gleaned from Jorgenson's human resources investigation runs into the same problems. The county does not dispute Jorgenson's comments to female coworkers were inappropriate. And plaintiffs do not dispute those comments went unreported until the N.S. investigation. Even if the county was somehow responsible for Jorgenson's boorishness, it was not deliberately indifferent to whether the problem continued. When staff notified the county of Jorgenson's behavior, an investigation ensued, showing the county's diligence, and Jorgenson resigned. See Vigo Cty. , 905 F.3d at 1046 ("[A]ccepting resignations in lieu of firings [does not] reflect[ ] the County's deliberate indifference."). Our recent decision in Vigo County is instructive on the quantum of proof necessary to establish a county's custom or practice of failing to prevent or respond to its employees' sexual misconduct. In that case, the plaintiff, Doe, volunteered at a park where Vigo County's employee, David Gray, worked. Doe alleged Gray locked her in the park's restroom area and forced her to perform oral sex and digitally penetrated her vagina. Vigo Cty. , 905 F.3d at 1041. Gray was charged with rape, criminal confinement, and official misconduct, and he was convicted of the latter two offenses. Id . Doe sued Vigo County, alleging it failed to take seriously or to address a risk of sexual violence posed by its employees. Id . at 1044-45. The record contained no evidence of any county employee having forced another to engage in a sexual act or having confined an individual to harm her. Id . at 1045. Instead, the record revealed "[s]ome involved sexual misconduct, but none resulted in coerced sexual activity, nor does the record suggest that employee misconduct occasioned impunity." Id . at 1045. This court held that "a handful of incidents of misconduct by employees of Vigo County" over the past 20 years, "is not enough to establish a custom or practice that gave rise to Doe's injuries, nor can it support a finding of indifference on the part of Vigo County officials." Id . Like Vigo County , in this trial plaintiffs failed to put forth evidence of any jail employee engaging in criminal acts like Christensen. The other allegations plaintiffs recount, though contemptible, are different from the trauma plaintiffs experienced. See id . And because plaintiffs rely on indirect proof of a widespread practice, they "must introduce evidence demonstrating that the unlawful practice was so pervasive that acquiescence on the part of policymakers was apparent and amounted to a policy decision." Dixon v. Cook Cty. , 819 F.3d 343, 348 (7th Cir. 2016) (citation omitted). Nargis's tier talk, the tizzy email, rare and unreported staff comments, and the Jorgenson investigation do not amount to evidence of persistent or pervasive sexual misconduct so obvious as to imply acquiescence of county policymakers. Above all, plaintiffs offered no evidence county officials knew or should have known that any of its practices or customs would allow or encourage-much less cause-Christensen to commit the abhorrent acts which happened to plaintiffs. Because the trial evidence contains no facts that plausibly suggest a widespread practice of sexual assaults or acquiescence to sexual conduct at Polk County Jail, the record does not support a finding that the county maintained an implicit policy that that served as the cause of their injuries. 3. Failure-to-train Plaintiffs' third liability theory is that the county's training "was entirely deficient and independently established deliberate indifference." This failure-to-train theory runs into difficulties. On the evidence presented at trial, it allows a jury to conclude liability outside the correct legal framework, and it relies on inferences expressly rejected by the Supreme Court. Further, the trial record does not show a direct causal link between the alleged failure to train and their injuries as required by Monell and its related case law. On appeal of a jury verdict, we afford a generous standard of review to avoid supplanting our view of the credibility or the weight of the evidence for that of the jury. Massey v. Blue Cross-Blue Shield of Illinois , 226 F.3d 922, 925-26 (7th Cir. 2000) (affirming judgment as a matter of law overturning verdict in plaintiff's favor because evidence failed to support the jury's finding of discrimination). But juries are not free to disregard governing legal standards. On the question of municipal liability, Monell and its limits control the calls juries are allowed to make, including for failure-to-train claims. See, e.g., Canton , 489 U.S. at 399, 109 S.Ct. 1197 (O'Connor, J., concurring in part and dissenting in part) ("Allowing an inadequate training claim such as this one to go to the jury based upon a single incident would only invite jury nullification of Monell ."). A failure-to-train claim fails without a pattern of similar violations, unless that claim "fall[s] within the narrow range of 'single-incident liability' hypothesized in Canton ." Connick , 563 U.S. at 71-72, 131 S.Ct. 1350 ("[A] pattern of violations [is] necessary to prove deliberate indifference in § 1983 actions alleging failure to train."); see also Canton , 489 U.S. at 390 n.10, 109 S.Ct. 1197. "A municipality's culpability for a deprivation of rights is at its most tenuous where a claim turns on a failure to train." Connick , 563 U.S. at 61, 131 S.Ct. 1350 (citing Oklahoma City v. Tuttle , 471 U.S. 808, 822-23, 105 S.Ct. 2427, 85 L.Ed.2d 791 (1985) (plurality opinion)) ("[A] policy of 'inadequate training' " is "far more nebulous, and a good deal further removed from the constitutional violation, than was the policy in Monell ."). Here, plaintiffs agreed there is no pattern of similar violations at the jail to establish deliberate indifference. In so doing, they deny proof of a fundamental element of failure-to-train liability. The degree of culpability in failure-to-train cases must amount to deliberate indifference. Connick , 563 U.S. at 63, 131 S.Ct. 1350. Although plaintiffs do not contend they proved a pattern of similar violations, they claim "myriad evidence" of inadequate training supported a finding of deliberate indifference. See , e.g., id . at 62, 131 S.Ct. 1350. But plaintiffs' contention here requires reliance on the same chain of inferences rejected in Connick. In Connick , the Court analyzed whether failure-to-train liability could be imposed on a district attorney's office for a rogue prosecutor's deliberate violation of Brady v. Maryland , 373 U.S. 83, 83 S.Ct. 1194, 10 L.Ed.2d 215 (1963). The Court declined to conclude four prior Brady violations was a pattern of similar violations. Connick , 563 U.S. at 54, 62-63, 131 S.Ct. 1350. Without such a pattern, the plaintiff could not prove that the district attorney had actual or constructive notice of, and was therefore deliberately indifferent to, a need for more or different Brady training. Id . at 59, 61-63, 72, 131 S.Ct. 1350. As a result, the district court's $ 14 million judgment against the district attorney was overturned. Id . at 72, 131 S.Ct. 1350. Here, we consider whether failure-to-train liability may be imposed on the county for a rogue guard's deliberate violation of jail policy, county training, and Wisconsin law. We follow Connick's approach, which required incidents "similar to the violation at issue" to "put [a policymaker] on notice that specific training was necessary to avoid this constitutional violation." Id . at 62, 131 S.Ct. 1350. "Without notice that a course of training is deficient in a particular respect"-here, the prevention of sexual assaults-"decisionmakers can hardly be said to have deliberately chosen a training program that will cause violations of constitutional rights." Id . at 62-63, 131 S.Ct. 1350. Under Connick , we conclude the evidence before the jury could not, standing alone, have prompted notice that more or different training was necessary to prevent similarly appalling violations. Jorgenson's alleged actions and behavior were wrong and degrading. Yet these twice investigated but unverified allegations, including placing his hands around N.S.'s waist and touching her backside, along with Jorgenson's other reproachable conduct, do not prompt notice that specific training was necessary to avoid Christensen's repeated sexual assaults. Because the trial evidence contained no instances or pattern of comparable actions, the county cannot be said to have adhered to an approach that it knew or should have known failed to prevent similar violations. Connick , 563 U.S. at 62, 131 S.Ct. 1350 ; Bryan Cty. , 520 U.S. at 409, 117 S.Ct. 1382. A failure-to-supervise claim fails for the same reasons. The record does not establish a likelihood of the type of harm plaintiffs suffered to have obligated the county to prevent its occurrence. See Vigo Cty. , 905 F.3d at 1046 (holding same). The dissent concludes three "primary points" prompted notice that more training was required: (1) Nargis's tier talk; (2) information gleaned from the Jorgenson investigation; and (3) irregular examples of inappropriate remarks by certain guards over a twelve-year period. But as offensive as they are, none of these points involved the clandestine and conscience-shocking repeated sexual assaults of inmates. To demonstrate deliberate indifference to the risk of constitutional violations, Connick requires "[a] pattern of similar constitutional violations by untrained employees." Connick , 563 U.S. at 62, 131 S.Ct. 1350 (emphases added). The record shows no pattern of violations similar to Christensen's conduct, and importantly, no dispute that Christensen was trained that his conduct was illegal. In fact, at trial Christensen admitted he did not require more training to know his conduct was a crime. For these reasons, we respectfully part ways with our dissenting colleague's failure-to-train evaluation. To be sure, "[i]f a program does not prevent constitutional violations, municipal decisionmakers may eventually be put on notice that a new program is called for." Bryan Cty. , 520 U.S. at 407, 117 S.Ct. 1382. To show the county ignored such notice, plaintiffs must produce evidence "of a series of constitutional violations from which deliberate indifference can be inferred." Estate of Novack ex rel. Turbin v. Cty. of Wood , 226 F.3d 525, 531 (7th Cir. 2000) ; see also Hahn v. Walsh , 762 F.3d 617, 637 (7th Cir. 2014) (holding plaintiffs failed to "show[ ] that there was a 'series of unconstitutional acts from which it may be inferred that the [sheriff] knew [correctional center] officers were violating the constitutional rights of [correctional center] inmates and did nothing") (quoting Estate of Novack , 226 F.3d at 531 ). In failure-to-train cases, the constitutional violations must be "similar to the violation at issue," Connick , 563 U.S. at 63, 131 S.Ct. 1350 ; here, sexual assaults. Yet Christensen's assaults on plaintiffs were both hidden and unprecedented. Testifying on these facts, plaintiffs' prison training expert agreed the county had a good record-even factoring in Jorgenson's misconduct-because of the lack of incidents of sexual contact between guards and inmates, let alone coercive assaults like Christensen's. Cf. Woodward , 368 F.3d at 926 (involving systemic failure to enforce a jail suicide-prevention program). There was no series of sexual assaults at the jail from which the county was aware its sexual assault training was inadequate, and chose to do nothing in the face of such knowledge. See Hahn , 762 F.3d at 637 (holding seven inmate deaths in jail from different causes than decedent's "do not show that [sheriff] was aware of any risk posed by [his] policies or that [sheriff] failed to take appropriate steps to protect [decedent]"). Nor does the trial evidence show "continued adherence" to training resulting in flaws exposed by repeated wrongdoing. Bryan Cty. , 520 U.S. at 407, 117 S.Ct. 1382. Connick presents stringent fault standards for a failure-to-train claim, and admonishes that unless an exception applies, failure-to-train liability is available only when "a pattern of similar violations" establishes a "policy of inaction." Connick , 563 U.S. at 72, 131 S.Ct. 1350. Even if the trial record showed sufficient evidence the county failed to train, that record still must contain proof of causation. When evaluating Monell claims, the Supreme Court has instructed courts to adhere to "rigorous" causation requirements. Bryan Cty. , 520 U.S. at 415, 117 S.Ct. 1382. The trial evidence showed that the "moving force," id . at 404, 117 S.Ct. 1382, behind the assaults on plaintiffs was not a failure to give jail guards additional training. Rather, it was a predatory employee-who does not merit the term "guard"-who furtively abused his power and preyed upon inmates. From the witness stand, Christensen confessed his behavior was irrepressible: he admitted he was trained his actions were criminal and violated jail policy; he agreed he did not require more training to know his assaults were a crime; he knew he was placing plaintiffs at risk; and he operated under the delusion that his conduct was welcome, consensual, and a product of "mutual voluntary attraction." When juxtaposing these facts with the absence of any similar violations at the jail, Christensen-not a failure to train-was the moving force behind the deprivation of plaintiffs' federal rights. At trial and on appeal, plaintiffs have offered no more than conclusory assertions that Christensen's lack of training caused their injuries. The trial record also does not reveal an affirmative link between a failure to train and plaintiffs' injuries. The dissent identifies this gap-"[w]hat was missing"-and cites the need for more training on "the inherent vulnerability" of the confinement setting, as well as the harm caused by sexual abuse. "But showing merely that additional training would have been helpful in making difficult decisions does not establish municipal liability." Connick , 563 U.S. at 68, 131 S.Ct. 1350. "Proving that an injury or accident could have been avoided if an employee had had better or more training, sufficient to equip him to avoid the particular injury-causing conduct will not suffice." Id . (internal brackets and quotation marks omitted). "In virtually every instance where a person has had his or her constitutional rights violated by a city employee, a § 1983 plaintiff will be able to point to something the city 'could have done' to prevent the unfortunate incident." Canton , 489 U.S. at 392, 109 S.Ct. 1197 (citing Tuttle , 471 U.S. at 823, 105 S.Ct. 2427 (plurality opinion)). Even plaintiffs' expert on prison training standards, Eiser, conceded no proof exists that better or more training could have dissuaded Christensen from his predatory behavior and established causation. Our dissenting colleague warns of the risk of sexual attacks at jails "employing male guards to supervise female inmates." But Connick requires more than "the broader context" of male guards supervising female inmates to establish causation. It does not follow that all male guards will "so obviously make wrong decisions that failing to train them amounts to 'a decision by the [county] itself to violate the Constitution.' " Connick , 563 U.S. at 71, 131 S.Ct. 1350 (quoting Canton , 489 U.S. at 395, 109 S.Ct. 1197 ). Connick sets the bar higher: To prove deliberate indifference, Thompson needed to show that Connick was on notice that, absent additional specified training, it was 'highly predictable' that the prosecutors in his office would be confounded by those gray areas and make incorrect Brady decisions as a result. In fact, Thompson had to show that it was so predictable that failing to train the prosecutors amounted to conscious disregard for defendants' Brady rights. Connick , 563 U.S. at 71, 131 S.Ct. 1350 (citing Bryan Cty., 520 U.S. at 409, 117 S.Ct. 1382, and Canton , 489 U.S. at 389, 109 S.Ct. 1197 ) (emphases in original). This trial evidence does not clear that bar, and there are no gray areas to the zero-tolerance policy in question. A finding of liability on a failure-to-train theory here cannot be reconciled with the reasoning of Connick , and would rest more on good policy prescriptions than proof of municipal fault and causation. And whether noteworthy recommendations for jail training are legal requirements is something for the people of Wisconsin and their elected officials to decide, rather than our court in this context. 4. Single-incident theory In the absence of a pattern of similar assault violations, another liability theory is that the county failed to train its guards in light of foreseeable sexual assaults. "In limited circumstances, a local government's decision not to train certain employees about their legal duty to avoid violating citizens' rights may rise to the level of an official government policy for purposes of § 1983." Connick , 563 U.S. at 61, 131 S.Ct. 1350 ; see also Canton , 489 U.S. at 390 n.10, 109 S.Ct. 1197. Canton left open the possibility that a plaintiff might succeed in a failure-to-train claim without showing a pattern of constitutional violations. See Connick , 563 U.S. at 63, 71-72, 131 S.Ct. 1350 (describing as "single-incident liability"). The Supreme Court "hypothesized" in Canton that "a violation of federal rights may be a highly predictable consequence" for failing to train police officers about constitutional limits on the use of deadly force. Id . at 63-64, 71-72, 131 S.Ct. 1350. In Canton , the Court required that "for liability to attach in this circumstance the identified deficiency in a city's training program must be closely related to the ultimate injury." 489 U.S. at 391, 109 S.Ct. 1197. The single-incident theory of liability described in Canton "assumes ... no knowledge at all" of the required constitutional standards. Connick , 563 U.S. at 67, 131 S.Ct. 1350. Without specific training, explained the Court, a police officer would not be "equipped with the tools to interpret and apply legal principles." Id . at 64, 131 S.Ct. 1350. In Connick , the Court contrasted this hypothetical with an attorney asked to make a Brady determination. In that situation, and in the absence of a pattern of similar Brady violations, a district attorney "is entitled to rely" on prosecutors' law school or bar exam training, ethical obligations, and on-the-job experience, to deal with Brady decisions. Id . at 66-67, 131 S.Ct. 1350 ("A licensed attorney making legal judgments, in his capacity as a prosecutor, about Brady material simply does not present the same 'highly predictable' constitutional danger as Canton's untrained officer."). "In light of this regime of legal training and professional responsibility, recurring constitutional violations are not the 'obvious consequence' of failing to provide prosecutors with formal in-house training about how to obey the law." Id . at 66, 131 S.Ct. 1350. Accordingly, the failure of Connick's office to alert its prosecutors of all reasonably conceivable legal duties did not subject it to failure-to-train liability, even if "additional training would have been helpful." Id . at 68, 131 S.Ct. 1350. Here, the proof at trial does not fit within Canton's single-incident hypothetical. First, we cannot assume "no knowledge at all," because Christensen was trained and knew that his actions were criminal. Given this knowledge and training, Christensen's assaults-in which he was a lone and surreptitious actor-were not a "highly predictable consequence" ( Bryan Cty. , 520 U.S. at 409, 117 S.Ct. 1382 ) of the county's sexual assault policies. Second, unlike the nuanced and compound legal standards contemplated in Canton (involving constitutional limits on the use of deadly force) and Connick (involving evidentiary disclosure obligations), the legal standard here involved a direct, non-discretionary rule: no sexual contact with inmates. Third, the record shows the county's guards-including Christensen-were trained on their legal and professional obligations to avoid the constitutional violation at issue, sexual relationships with inmates. Christensen was thus "equipped with the tools" to obey the law, as Canton requires. Connick , 563 U.S. at 64, 131 S.Ct. 1350. On this record, we decline to find single-incident liability. 5. Sufficiency of the evidence Judgment as a matter of law should not be granted unless the evidence, viewed in the light most favorable to the jury's verdict, shows that no rational jury could return a verdict against the moving party. Woodward , 368 F.3d at 926. Although the district court found sufficient evidence to sustain the verdict in Nargis's "awareness of tier talk," facts culled from the Jorgenson investigation, and the single county-sponsored PREA training session, ultimately we disagree. In our de novo review, plaintiffs failed to present enough evidence to support their claims. Houskins , 549 F.3d at 493. The facts do not show Christensen "was highly likely to inflict the particular injury suffered by the plaintiff[s]." Bryan Cty. , 520 U.S. at 412, 117 S.Ct. 1382 (emphasis in original). The trial record offers no support that the same or similar constitutional injuries were foreseeable. So far as the record reveals, the county had no reason to believe, before the events giving rise to this case, that its training or supervision of Christensen was inadequate. Besides, there is no evidence of a pattern of constitutional violations making it "known or obvious" that additional training or enhanced policies were necessary. Id . at 407, 410, 117 S.Ct. 1382 ; see also Connick , 563 U.S. at 61, 131 S.Ct. 1350. The trial record does not contain evidence from which a rational jury could find the county knew of, but was deliberately indifferent to, a risk that inmates' constitutional rights would be violated. Even if the record contained evidence of culpability, plaintiffs needed to show a direct causal connection between a county policy, practice, or custom and their injuries. Rice ex rel. Rice v. Corr. Med. Servs. , 675 F.3d 650, 675 (7th Cir. 2012) ("[I]n other words that the policy or custom was the moving force behind the constitutional violation."). Given the absence of causation evidence, the record can only be read that Christensen-not any county policy or failure to train-was the moving force behind plaintiffs' injuries. To hold otherwise