In the

     United States Court of Appeals
                     For the Seventh Circuit
                       ____________________
Nos. 18-1498, 18-1499, 18-2170 & 18-2177
J.K.J. and M.J.J.,
                                                Plaintiffs-Appellees,
                                  v.

POLK COUNTY and DARRYL L. CHRISTENSEN,
                                   Defendants-Appellants.
                       ____________________

          Appeals from the United States District Court for the
                     Western District of Wisconsin.
    Nos. 3:15-cv-00428 & 3:15-cv-00433 — William M. Conley, Judge.
                       ____________________

     ARGUED DECEMBER 5, 2019 — DECIDED MAY 15, 2020
                ____________________

   Before WOOD, Chief Judge, and BAUER, EASTERBROOK,
KANNE, ROVNER, SYKES, HAMILTON, BARRETT, BRENNAN,
SCUDDER, and ST. EVE, Circuit Judges.
    SCUDDER, Circuit Judge. While confined in the Polk County
Jail, two female inmates, J.K.J. and M.J.J., endured repeated
sexual assaults at the hands of correctional officer Darryl
Christensen. The two women brought suit in federal court
against Christensen and Polk County. A trial ensued, and the
jury heard evidence of Christensen’s horrific misconduct over
2                                           Nos. 18-1498, et al.

a three-year period. The County’s written policy prohibited
sexual contact between inmates and guards but failed to ad-
dress the prevention and detection of such conduct. Nor did
the County provide any meaningful training on the topic.
What is more, toward the beginning of the relevant period,
the County learned that another guard made predatory sex-
ual advances toward a different female inmate. The trial evi-
dence showed that the County imposed minor discipline on
the guard but from there took no institutional response—no
review of its policy, no training for guards, no communication
with inmates on how to report such abuse, no nothing. In the
end, the jury returned verdicts for J.K.J. and M.J.J.
   The case against Christensen was open and shut. But a di-
vided panel of this court overturned the jury’s verdict against
Polk County, determining that the trial evidence failed to
meet the standard for municipal liability under Monell v. De-
partment of Social Services, 

(1978). We decided to
rehear the case en banc and now affirm the jury’s verdicts
against both Christensen and Polk County. While the stand-
ard for municipal liability is demanding—designed to ensure
that a municipality like Polk County is liable only for its own
constitutional torts and not those of employees like Christen-
sen—the evidence was sufficient to support the verdict
against the County.
                               I
    J.K.J. and M.J.J. sued Christensen and Polk County under
42 U.S.C. § 1983, alleging that the defendants violated the
Eighth and Fourteenth Amendments by acting with deliber-
ate indifference to a serious risk of harm to their safety and
well-being. They also brought a negligence claim under Wis-
consin law against the County. The district court consolidated
Nos. 18-1498, et al.                                           3

the cases for trial. The five-day trial ended with the jury find-
ing both defendants liable on all claims, and we recount the
facts in the light most favorable to that verdict. See Martin v.
Milwaukee County, 

, 547 n.1 (7th Cir. 2018).
                               A
    J.K.J. and M.J.J. suffered from addictions and committed
crimes that landed them in the Polk County Jail intermittently
between 2011 and 2014. Located in northwest Wisconsin, the
institution houses up to 160 inmates, including a small num-
ber of women, and employs about 27 correctional officers.
Christensen worked for 19 years as one of the guards tasked
with protecting the inmates—a duty he severely betrayed.
    J.K.J. and M.J.J.’s experiences with Christensen were
unique but shared a basic pattern. Christensen began by com-
menting on their appearances—remarks like “nice ass” and
“you’re looking good”—with the verbal harassment then es-
calating to explicit sexual overtures. In time came physical
contact, which began with Christensen groping and kissing
the women and from there advanced to oral sex and digital
penetration and eventually to intercourse. J.K.J. could not pin-
point the total number of times Christensen assaulted her but,
by way of example, stated that, during a two-month period in
the summer of 2012, he insisted on sexual contact every time
he was on duty. For her part, M.J.J. estimated that Christensen
engaged in sexual contact with her 25 to 75 times. These
events spanned about three years.
   Christensen took steps to conceal his misconduct within
the jail. While making inappropriate sexual comments in
front of others, he always made sure to take J.K.J. and M.J.J. to
hidden areas to engage in the physical contact. Christensen
4                                             Nos. 18-1498, et al.

also instructed both women not to tell anyone of the encoun-
ters because, if word got out, he would lose his job and family.
For the most part, the women heeded his admonishment and
kept the abuse to themselves during their incarceration. J.K.J.
and M.J.J. explained their silence in terms familiar to many
victims of sexual harassment and assault—shame, doubt an-
yone would believe them, and fear of retaliation.
    But the truth eventually came out. Another county’s inves-
tigator called Polk County to report an allegation that Chris-
tensen had sexual contact with an inmate. Polk County re-
sponded by commencing an internal investigation, and Chris-
tensen resigned upon being confronted. A criminal investiga-
tion followed and led the Wisconsin Department of Justice to
J.K.J. and M.J.J. After expressing initial reluctance to talk, both
women eventually felt safe enough to trust the investigators
with their stories. Christensen later pleaded guilty to criminal
charges and is now serving a 30-year sentence.
                                B
    Christensen’s conduct was not the only evidence of sexual
misconduct at the Polk County Jail that the jury heard. In
2012, toward the beginning of Christensen’s assaults of J.K.J.
and M.J.J., complaints surfaced that correctional officer Allen
Jorgenson had an inappropriate relationship with a female in-
mate known as N.S. Sergeant Steven Schaefer brought the
complaints to Captain Scott Nargis, the day-to-day head (ef-
fectively the warden) of the jail. Schaefer reported that Jorgen-
son had touched N.S. on her waist and rear end, adding that
the complaints did not come as a surprise because “[w]e have
all heard complaints about [Jorgenson’s] inappropriate com-
ments to both inmates and staff.”
Nos. 18-1498, et al.                                               5

    Captain Nargis responded by partnering with Deputy
Sheriff Steven Moe to investigate the contentions. Although
Jorgenson and N.S. denied any wrongdoing, Nargis and Moe
believed lines had been crossed. Indeed, the investigation re-
vealed that Jorgenson not only flirted with female inmates,
but also focused video cameras on the female housing pod for
an inordinate amount of time, and fostered an inappropriate
relationship with N.S. But Moe testified that he initially did
not believe Jorgenson had a sexual relationship with N.S.
     Based on those findings, Moe and Captain Nargis decided
that the right response was to issue a written reprimand to
Jorgenson. As part of doing so they assured Jorgenson that the
reprimand was not a “major deal” and he could move on from
it. “After having confronted Allen,” Moe testified, “we felt
that it was important that we recognize and support Allen’s
prior work history. He was a good employee. He was a go-to
employee. We appreciated his efforts and his work, so we
wanted to salvage him as an employee.”
    But the issue reawakened when N.S. sent Captain Nargis
a letter, dated January 19, 2012, explaining that she had lied
in denying the allegations about Jorgenson. At J.K.J. and
M.J.J.’s trial, the district court admitted N.S.’s letter not for its
truth, but for the non-hearsay purpose of informing the jury
of allegations of sexual misconduct that Polk County received
during the relevant period.
    N.S. began her letter by saying “I’m sorry for lying” and
“I would like to tell the truth about the allegations made
against Allen Jorgenson” because “[t]here are many things
[Jorgenson] has said & done that have been inappropriate in
a sexual manner towards me” and other inmates. Before de-
tailing her own account, N.S. emphasized that “I did not tell
6                                            Nos. 18-1498, et al.

the truth [earlier] because [Jorgenson] has told me to keep
quiet & said he didn’t wanna get in trouble.” From there N.S.
described the following misconduct that “started during my
last stay here from 10-27-10 til 7-6-11 & is continuing through
my incarceration now”:
       Jorgenson “always makes comments about see-
        ing us in the shower. He always calls it [a] ‘nice
        show.’”
       He has asked me “what the color of the day
        was”—a question about the color of “my under-
        clothes.”
       “He has told me he wants me to ride topless in
        his boat, [and] he has wanted me to lift my shirt
        for him while I’ve been here [in the jail] both
        times.”
       “Many times he’s leaned over the cart to look
        down my shirt.”
       “Recently he has started touching me.” “Every-
        one knows he’s doing these things” and
        “[w]hen he walks me back from the nurses of-
        fice, visiting anywhere he shoves me & pushes
        me” and “very recently … he grabbed me
        around my waist & kept his hand there til the K-
        Pod door opened then he slapped my butt.”
       “[W]hen giving me meds, he’ll look to see if the
        camera is on us. If not he comes around the cart
        & touches my back & butt as I go back in. One
        time the camera was on our direction, he said
        dam[n], I was gonna go in for the kill. Whatever
        that meant.”
Nos. 18-1498, et al.                                              7

    N.S. closed her letter by underscoring that “another reason
I did not tell the truth [in the initial investigation] is because I
don’t want problems with [Jorgenson] or any of the other jail-
ers who will be mad @ me for confirming these accusations.”
She also explained that she was not candid with her fellow
inmates either because “I didn’t want Allen [Jorgenson] in
trouble or mad @ me, making it hell for me here.”
    Upon receiving N.S.’s letter, Deputy Sheriff Moe and Cap-
tain Nargis reopened the prior investigation to take a fresh
look at Jorgenson’s conduct. Sergeant Schaefer also got in-
volved, spoke with N.S. to verify her report, and concluded
that she may have been telling the truth at that point. Moe,
too, acknowledged at trial that, upon receiving N.S.’s letter,
he found it “more likely” that Jorgenson had inappropriately
or even illegally touched her.
     From there, however, Polk County chose not to revisit its
prior disciplinary decision and determined that Jorgenson’s
conduct still merited only the prior written reprimand. The
jail took no further action in response to N.S.’s new allega-
tions. Jorgenson later resigned after an unrelated investiga-
tion regarding his female co-workers.
    Sexual harassment had appeared in the jail in other ways
too. Christensen testified that he witnessed at least two other
jailers, including Allen Jorgenson, make sexual comments to
inmates. And then there was the issue of “tier talk,” a term
that Captain Nargis agreed reflected “not necessarily flatter-
ing talk amongst co-workers in the tier.” By way of example,
Captain Nargis confirmed hearing that Christensen had made
inappropriate sexual comments about women in general and,
on one occasion, about an inmate’s breasts. Nargis even
acknowledged that “on occasion” he too participated in tier
8                                             Nos. 18-1498, et al.

talk, in an effort to be viewed as part of the group and a
trusted leader of the officers.
                                C
    What Polk County had done (and not done) to prevent the
sexual abuse of inmates was a key focus at trial. The County
established written policies against the sexual harassment
and assault of inmates. Policy I-100 of the jail’s Policy and Pro-
cedures Manual listed inmates’ rights and stated that they
were never to be subjected to “verbal, physical, emotional,
psychological or sexual harassment” by staff. Any harassing
officer was “subject to disciplinary charges and/or termina-
tion.”
    Another of the Manual’s provisions, Policy C-202, stated
that jail employees were prohibited from fraternizing with in-
mates, including “[b]eing in an intimate social or physical re-
lationship with a prisoner.” In July 2012, Polk County up-
dated the Manual to include some language from the Prison
Rape Elimination Act, a federal statute enacted in 2003 to de-
ter the sexual assault of prisoners. The new language in-
structed that any staff member or inmate “who knows or rea-
sonably suspects” sexual misconduct was to inform the “Jail
Administrator” or, if the complainant was an inmate, she
could inform a staff member, and went on to describe how
such reports would be handled. The section noted that “Wis-
consin State Statutes make it a criminal offense for correc-
tional staff members to have sexual intercourse or contact
with an individual confined in a correctional institution.” See
WIS. STAT. § 940.225(2)(h).
   The 12-page Inmate’s Handbook also mentioned sexual
misconduct, providing: “[e]very inmate has the right to be
Nos. 18-1498, et al.                                             9

safe from sexual abuse and harassment. No one has the right
to pressure you to engage in sexual acts. If you are being pres-
sured[,] threatened, or extorted for sex, you should report this
to staff immediately.” As the plaintiffs’ expert would later de-
scribe it, however, this information appeared in “very small
font; one paragraph in the middle not even headlined, not
even with a title on it.” Inmates received the Handbook dur-
ing the intake process and were told to read it. At no point,
though, did the jail provide the inmates with any further in-
formation on how to report sexual misconduct.
    Aside from these written policies, Polk County Jail staff
received no training (in any sense of the word) focused on the
sexual harassment or assault of female inmates. Few though
they are, the details are important. Consider Polk County’s
program that required officers to read a specified policy each
day from the jail’s Policy and Procedures Manual and then to
initial a piece of paper and write the policy’s title as proof they
did so. For his part, Christensen told the jury that most of the
time he just went through the motions of writing down a pol-
icy’s title and signing without reviewing anything. Even
more, the trial record contained no evidence showing that
Captain Nargis or anyone from the County dedicated any
portion of any live training session to reviewing the jail’s writ-
ten policies or underscoring the necessity of reporting any
known or suspected sexual misconduct.
    Any in-person training that occurred was hidden among
the jail’s general training and completely silent on preventing
and detecting the sexual assault of female inmates. There was,
for example, a county-wide (but not jail-specific) training on
sexual harassment that addressed how employees should
maintain proper co-worker relationships but which Sergeant
10                                           Nos. 18-1498, et al.

Schaefer clarified was “not anything regarding inmates.” The
jury also heard evidence about the jailers’ training on the
vague topic of maintaining distance from inmates, with no
testimony to suggest that training ever touched the topic of
sexual assault. Sergeant Schaefer, who helped oversee the
training of new officers, expressly admitted that much. He re-
called being taught in “jail school” that he should not “be-
come too close” with inmates or share personal information
with them, though he explicitly denied any memory of being
told that it was improper for jail officers to have sexual rela-
tionships with inmates. Those vague cautions were repeated
to him during on-the-job training in the Polk County Jail. And
when Sergeant Schaefer trained others, he gave the same ad-
monishments. While Schaefer agreed that having sex with an
inmate would qualify as being “too familiar,” he did not tes-
tify that he ever addressed this topic in any way, including in
any training session.
     Beyond learning that training on sexual abuse was nearly
nonexistent, the jury heard affirmative evidence revealing the
County’s dismissive attitude about preventing and detecting
it. The prime example came in the “tizzy email.” On February
21, 2014, near the end of Christensen’s abuse of J.K.J. and
M.J.J., Captain Nargis sent an email to many staff members
summarizing the contents of a training held the day before,
which included the Prison Rape Elimination Act as one of its
several topics. Nargis wrote that it “[s]eems that everyone is
in a tizzy to train their staff on PREA.” Nargis testified that he
used the word “tizzy” to mean “that there’s a bit of a scramble
for, in this particular case, time and attention that seemed to
be misplaced.” His email went on to state that, although
“[t]here is no requirement for [the jail] to be compliant with
Nos. 18-1498, et al.                                         11

everything that [PREA] calls for,” the training would “hit the
basics.”
    For their part, J.K.J. and M.J.J. presented evidence on the
inadequacy of Polk County’s policies and training. They did
so through the testimony of Jeffrey Eiser, an expert on jail op-
erations. Eiser explained that Congress enacted PREA in re-
sponse to an “evident” and “prevalent” problem with sexual
assault and abuse in jails. He added that PREA’s threefold ob-
jectives are “to prevent sexual abuse and harassment, to de-
tect it, and then to respond to it.” But Eiser’s review of Polk
County’s policies left him of the conviction that the jail had
sufficiently covered only the third base—responding to sex-
ual abuse complaints—but otherwise inadequately addressed
prevention and detection.
    Eiser did not stop there. He then offered concrete exam-
ples of ways Polk County could improve its policies. To pre-
vent abuse, a policy could make clear that the institution op-
erates under a zero-tolerance policy on sexual abuse and har-
assment. It likewise could designate a PREA coordinator,
train staff on what to look for and how to report abuse as well
as how to make inmates feel comfortable coming forward,
take added care with job assignments within the facility, and
ensure that all inmates understand their right to be free from
sexual abuse and harassment. Similarly, Eiser testified that, to
detect sexual misconduct, a policy could make sure that the
inmates understand what abuse entails, since they may come
from life experiences that have blurred the lines of abnormal
and normal relationships. Eiser added that a sound policy
also would provide a safe, confidential way for inmates to re-
port abuse (through, for example, the use of a locked drop-
box), instead of putting inmates in the position of having to
12                                           Nos. 18-1498, et al.

hand a grievance to an officer who may be friends with the
abuser. Polk County’s policy lacked all of these features.
                               D
    The jury returned a verdict in favor of J.K.J. and M.J.J. on
their claims against both Christensen and Polk County. The
case proceeded to the damages phase, and the jury then heard
testimony about the impact the defendants’ conduct has had
on the women’s lives. The jury translated that evidence into
compensatory damages awards of $2 million each for J.K.J.
and M.J.J. The jury further determined that Christensen’s con-
duct warranted his paying punitive damages of $3.75 million
to each woman.
    Both defendants challenged the jury’s verdicts in post-trial
motions. They moved under Rule 59 for a new trial based on
errors that they contended the district court made. Polk
County also moved under Rule 50 for judgment as a matter of
law, arguing that the trial evidence was legally insufficient to
prove J.K.J. and M.J.J.’s § 1983 claims and that it was immune
from liability on the state-law negligence claim. The district
court agreed on the latter point and dismissed the negligence
claim, but otherwise left intact the jury’s verdicts on the con-
stitutional claims and denied the request for a new trial.
    Having presided over the trial, Judge Conley determined
that the evidence sufficed to allow the jury to find from the
“tier talk” alone that “jail officials not only turned a blind eye,
and perhaps even fostered, a culture where inappropriate sex-
ual comments were accepted as the norm.” Highlighting the
showing J.K.J. and M.J.J. made of both Captain Nargis’s learn-
ing in 2012 of Allen Jorgenson’s sexual misconduct and his
dismissive handling of PREA training in February 2014, Judge
Nos. 18-1498, et al.                                           13

Conley explained that the jury had ample evidence from
which to conclude that the County “downplayed the im-
portance of preventing sexual assault and harassment within
the jail.” Considered in its entirety, Judge Conley continued,
the evidence supported the jury’s finding by a preponderance
of the evidence that “if the County had provided adequate
notice and training to correctional officers and inmates on
what constitutes sexual harassment and abuse, and how to re-
port it, plaintiffs [J.K.J. and M.J.J.] may not have been sexually
assaulted and harassed” from 2011 to 2014.
   This appeal followed.
                                II
   Christensen gives us no good reason to upset the jury’s
verdict against him. To establish that his conduct violated
their Eighth Amendment rights, J.K.J. and M.J.J. had to prove
that Christensen acted with deliberate indifference to an ex-
cessive risk to their health or safety. See Farmer v. Brennan, 

, 837 (1994); Sinn v. Lemmon, 

, 419 (7th Cir.
2018). It was more than reasonable for the jury to conclude
that the trial evidence met that standard. To say that the sex-
ual assaults he committed against J.K.J. and M.J.J. objectively
imposed serious risk to their safety would be an understate-
ment. And the evidence was equally sufficient to show that
Christensen knew of that danger. Indeed, he admitted at trial
that he knew he was putting the plaintiffs at risk and that his
conduct not only violated prison policy but was criminal.
    Christensen’s only defense was to try to somehow per-
suade the jury that J.K.J. and M.J.J. consented to the sexual re-
lations. The effort failed and now on appeal he contends that
the district court erred in not giving the jury a special
14                                          Nos. 18-1498, et al.

instruction on his consent defense. Our review of the trial
transcript shows that the district court completely and accu-
rately instructed the jury on the elements of the Eighth
Amendment claim. No further explanation was necessary to
tell the jury how to consider the consent issue. If the jury had
bought Christensen’s story that J.K.J. and M.J.J. were willing
participants (and, for that matter, even capable of being will-
ing participants under the circumstances), it would have
found that the women had not met their evidentiary burdens
of proving that he acted with deliberate indifference to their
safety and well-being. But the jury reached no such conclu-
sion. The instructions were sufficient.
    Christensen’s last challenge is to the damages awards. He
finds them problematic because the jury gave the same
amounts to both women. To be sure, the sexual abuse had
unique effects on J.K.J. and M.J.J., who each came to Polk
County Jail from distinct lives and suffered their own per-
sonal tragedies. But that does not mean that they necessitated
different compensatory damages amounts, particularly given
the psychology expert’s recommendation of identical courses
of treatment for both women. Nor was the jury’s punitive
damages award so great as to be unreasonable or outside the
bounds of due process.
     The judgment against Christensen is easily affirmed.
                              III
                               A
   We now turn to the more difficult question of Polk
County’s liability. The County raises a few issues on appeal
but only one merits discussion—whether the trial evidence
was sufficient to sustain the jury’s verdict. Polk County is not
Nos. 18-1498, et al.                                            15

automatically on the hook for Christensen’s unconstitutional
acts just because it employed him. Under the familiar holding
of Monell v. Department of Social Services, local governments
like Polk County can be held responsible for constitutional vi-
olations only when they themselves cause the deprivation of
rights. 

, 691–92 (1978). Time and again the Su-
preme Court has reinforced the strict prohibition against al-
lowing principles of vicarious liability to establish municipal
liability under § 1983. See

see also Bd. of Comm’rs
of Bryan County v. Brown, 

, 403 (1997) (collecting
cases and reinforcing that the doctrine of respondeat superior
does not apply under § 1983).
    Monell liability is difficult to establish precisely because of
the care the law has taken to avoid holding a municipality re-
sponsible for an employee’s misconduct. A primary guardrail
is the threshold requirement of a plaintiff showing that a mu-
nicipal policy or custom caused the constitutional injury. See

–91. “Locating a ‘policy,’” the Supreme
Court has emphasized, “ensures that a municipality is held
liable only for those deprivations resulting from the decisions
of its duly constituted legislative body or of those officials
whose acts may fairly be said to be those of the municipality.”
Bryan 

–04; see also Glisson v. Indiana
Dep’t of Corrs., 

, 381 (7th Cir. 2017) (“The critical
question under Monell remains this: is the action about which
the plaintiff is complaining one of the institution itself, or is it
merely one undertaken by a subordinate actor?”). A munici-
pal action can take the form of an express policy (embodied,
for example, in a policy statement, regulation, or decision of-
ficially adopted by municipal decisionmakers), an informal
but established municipal custom, or even the action of a
16                                           Nos. 18-1498, et al.

policymaker authorized to act for the municipality. See Glis-

.
    More is required before Monell liability can attach, how-
ever. “The plaintiff must also demonstrate that, through its
deliberate conduct, the municipality was the ‘moving force’ be-
hind the injury alleged.” Bryan 

. The
plaintiff, in short, “must show that the municipal action was
taken with the requisite degree of culpability and must
demonstrate a direct causal link between the municipal action
and the deprivation of federal rights.” Id.; see also 

(explaining that only when a municipality’s policy
or custom “inflicts the injury” is the entity responsible under
§ 1983).
    The most straightforward Monell claims are those in which
a plaintiff alleges that an affirmative municipal action is itself
unconstitutional. See Bryan 

–05. In
those cases, inferences of culpability and causation are easy,
for they follow directly from the municipality’s intentional
decision to adopt the unconstitutional policy or custom or to
take particular action. See

Consider, for example, a
city with a policy authorizing its employees to take some un-
constitutional act in connection with traffic stops after mid-
night. Deliberate conduct is easily inferred from the inten-
tional adoption of the offending policy. And if a victim of the
unconstitutional act emerges as a Monell plaintiff, there will
be little doubt that it was the city’s express instruction—not
the employee’s independent choice—that caused the injury.
    Here, however, J.K.J. and M.J.J. do not claim that Polk
County took affirmative action to harm them. To the contrary,
their theory of Monell liability roots itself in inaction—in gaps
in the County’s sexual abuse policy and its failure to properly
Nos. 18-1498, et al.                                            17

train the jailers in the face of obvious and known risks to fe-
male inmates. These failures to act, J.K.J. and M.J.J. contend,
were deliberate and together caused their constitutional inju-
ries. The Supreme Court has recognized that Monell liability
can arise from such decisions because a “city’s ‘policy of inac-
tion’ in light of notice that its program will cause constitu-
tional violations ‘is the functional equivalent of a decision by
the city itself to violate the Constitution.’” Connick v. Thomp-
son, 

, 61–62 (2011) (quoting City of Canton v. Harris,

, 395 (1989) (O’Connor, J., concurring in part and
dissenting in part)); see also 

(“[I]n sit-
uations that call for procedures, rules or regulations, the fail-
ure to make policy itself may be actionable.”).
    But the path to Monell liability based on inaction is steeper
because, unlike in a case of affirmative municipal action, a
failure to do something could be inadvertent and the connec-
tion between inaction and a resulting injury is more tenuous.
For these reasons, “[w]here a plaintiff claims that the munici-
pality has not directly inflicted an injury, but nonetheless has
caused an employee to do so, rigorous standards of culpabil-
ity and causation must be applied to ensure that the munici-
pality is not held liable solely for the actions of its employee.”
Bryan 

.
   Rigorous though these standards may be, they are not in-
surmountable. The question before us is whether the evidence
presented to the jury was legally sufficient to support the ver-
dict against Polk County. The law affords great respect to jury
verdicts. As a court of review, our role is limited to policing
the evidentiary threshold necessary as a legal matter to meet
Monell’s demands. In doing so, we do not reweigh evidence,
assess the credibility of any trial witness, or otherwise attempt
18                                               Nos. 18-1498, et al.

to usurp the jury’s role as factfinder. See Ruiz-Cortez v. City of
Chicago, 

, 601 (7th Cir. 2019) (“In our Rule 50 re-
view, we give the nonmovant ‘the benefit of every inference’
while refraining from weighing for ourselves the credibility
of evidence and testimony.”). To the contrary, we must affirm
unless there is “no legally sufficient evidentiary basis for a
reasonable jury to find for the non-moving party.” Woodward
v. Corr. Med. Servs. of Ill., Inc., 

, 926 (7th Cir. 2004).
    Against this standard of review, and ever mindful of Mo-
nell’s exacting liability requirements, we turn to the evidence
J.K.J. and M.J.J. put before the jury here.
                                  B
   All agree that Polk County’s written policies categorically
prohibited sexual contact with inmates and required re-
sponses to alleged violations. But J.K.J. and M.J.J. presented
evidence that the policy contained material gaps. The jury
heard expert testimony from Jeffrey Eiser about the im-
portance of a policy that does not wait for reports of sexual
abuse to trigger an institutional response, but instead contains
measures both to prevent the wrongdoing in the first instance
and to detect it if it does occur. Eiser spotlighted for the jury
that Polk County’s policy, although addressing incident re-
sponse, fell far short on prevention and detection.
    Eiser then explained that any number of policy measures
could have filled the gaps at little or no cost to Polk County.
Consider the need to detect sexual abuse. The importance of
a safe and confidential reporting channel—even something as
simple as a lockbox available to inmates—cannot be over-
stated. Under the County’s policy, an inmate seeking to report
abuse is left to inform one of 27 employees in a small jail that
Nos. 18-1498, et al.                                          19

she suffered a sexual assault at the hands of his coworker.
Given the perceived comradery among the male guards and
acceptance of sexual harassment at the jail’s highest levels (in-
ferred perhaps foremost from the “tier talk”), the jury could
have found that this was not a viable reporting option and in-
deed reflected a meaningful policy gap. Put most simply, the
jury could have credited Eiser’s expert testimony as part of
finding that Polk County’s policy deficiency affirmatively de-
terred the reporting and detection of sexual abuse of female
inmates.
    The policy gaps only widen when the focus turns to the
County’s sexual abuse training. Training is important because
it can educate and sensitize guards as well as shape and rein-
force institutional values, bringing to life words that other-
wise exist only on paper. The trial evidence showed that the
County’s training on preventing and detecting the sexual har-
assment and abuse of inmates was all but nonexistent. The
training consisted almost exclusively of informing guards of
the easy and evident—that the jail’s policies prohibited sexual
contact with inmates. The only training even addressing the
sexual assault of inmates by guards came in a single session
on the Prison Rape Elimination Act in 2014, well after much
of Darryl Christensen’s abuse of J.K.J. and M.J.J. had occurred,
and which he did not even attend. And even then the jury—
ever mindful of Captain Nargis’s dismissive “tizzy email”—
could have found that the County itself hardly took the PREA
training seriously.
    What was missing stands out. The jury heard no evidence
of the County informing guards of the inherent vulnerability
the confinement setting presents to female inmates, educating
jailers on the symptoms of an inmate suffering from the
20                                            Nos. 18-1498, et al.

trauma of abuse, requiring officers to report each other’s mis-
conduct, or taking any time to otherwise instruct guards on
matters of prevention and detection, whatever form that
might have taken.
    The trial evidence makes the bottom line plain: the jury
could have found that Polk County’s sexual abuse prevention
program was entirely lacking. The policy stated nothing but
the obvious—do not sexually abuse inmates. The County then
exacerbated the gap by failing to use training as the means of
making the policy prohibition a reality (or, at the very least,
mitigating risk) within the institution. The jury could have tal-
lied these gaps as part of finding the conscious, deliberate mu-
nicipal inaction upon which to rest Monell liability.
                                C
    Identifying municipal action—or, as it were, inaction—is
only part of the requisite inquiry under Monell. The Supreme
Court has made plain that a failure to act amounts to munici-
pal action for Monell purposes only if the County has notice
that its program will cause constitutional violations. See Con-

–62. Demonstrating that notice is essential
to an ultimate finding and requires a “known or obvious” risk
that constitutional violations will occur. Bryan 

.
    In many Monell cases notice requires proof of a prior pat-
tern of similar constitutional violations. See

This case
presents no such pattern. The district court declined to in-
struct the jury on the theory because it found insufficient evi-
dence of previous instances of sexual assault known to the
County. In so concluding, however, the district court
Nos. 18-1498, et al.                                           21

recognized that J.K.J. and M.J.J. had available another path to
show Polk County had the requisite notice.
    The alternative path to Monell liability comes from a door
the Supreme Court opened in City of Canton v. Harris, 

(1989). The Court observed that there may, as here, be cir-
cumstances in which “the need for more or different training
is so obvious, and the inadequacy so likely to result in the vi-
olation of constitutional rights” that a factfinder could find
deliberate indifference to the need for training.

. 
“In
that event, the failure to provide proper training may fairly be
said to represent a policy for which the city is responsible, and
for which the city may be held liable if it actually causes in-
jury.”
 Id. Put another 
way, a risk of constitutional violations
can be so high and the need for training so obvious that the
municipality’s failure to act can reflect deliberate indifference
and allow an inference of institutional culpability, even in the
absence of a similar prior constitutional violation.
    The Court did not leave the liability point in any way ab-
stract. To the contrary, it gave the express example of “city
policymakers [who] know to a moral certainty that their po-
lice officers will be required to arrest fleeing felons.”
 
Id. at 390
n.10. “The city,” the Court continued, “has armed its officers
with firearms, in part to allow them to accomplish the task.”

Id. The Court 
concluded that under those circumstances, “the
need to train officers in the constitutional limitations on the
use of deadly force can be said to be ‘so obvious,’ that failure
to do so could properly be characterized as ‘deliberate indif-
ference’ to constitutional rights.”
 Id. (internal citation 
omit-
ted).
   Drawing upon this precise example from City of Canton,
the Court has since reinforced that this doorway to Monell
22                                          Nos. 18-1498, et al.

liability remains ajar. In Bryan County, the Court confirmed
that City of Canton “did not foreclose the possibility that evi-
dence of a single violation of federal rights, accompanied by
a showing that a municipality has failed to train its employees
to handle recurring situations presenting an obvious potential
for such a violation, could trigger municipal 
liability.” 520
F.3d at 409
. The Court explained that “in a narrow range of
circumstances,” deliberate indifference could be found when
the violation of rights is a “highly predictable consequence”
of a failure to provide officers what they need to confront “re-
curring” situations.
 Id. And even 
more recently, in Connick v.
Thompson, the Court renewed its prior observations that City
of Canton “sought not to foreclose the possibility, however
rare, that the unconstitutional consequences of failing to train
could be so patently obvious that a city could be liable under
§ 1983 without proof of a pre-existing pattern of 
violations.”
563 U.S. at 64
.
    Though the Supreme Court has yet to confront a case that
presents a viable Monell claim based on a municipality’s fail-
ure to act in absence of a pattern, our court has done so twice.
Our en banc decision in Glisson v. Indiana Department of Cor-
rections relied upon City of Canton to hold that an institution
could be liable for failing to adopt protocols for the coordi-
nated and comprehensive treatment of chronically ill inmates.
See 849 F.3d at 382
. The inaction came at a time when the in-
stitution “had notice of the problems posed by a total lack of
coordination,” but then “despite that knowledge, did nothing
for more than seven years to address that risk.”
 Id. In reason-
ing fully applicable here, we concluded that a jury could find
that the prison knew for certain that its health providers
“would be confronted with patients with chronic illnesses,
and that the need to establish protocols for the coordinated
Nos. 18-1498, et al.                                            23

care of chronic illnesses is obvious,” just like it is obvious that
police officers would encounter situations where they would
need protocols on the use of excessive force.
 Id.
    And in 
Woodward v. Correctional Medical Services of Illinois,
Inc., we upheld a jury’s Monell verdict against a jail’s
healthcare provider for an inmate’s suicide. 
See 368 F.3d at
929
. The training on suicide prevention—a requirement un-
der the Eighth Amendment—was so inadequate and so
widely ignored that the contractor was on notice that a con-
stitutional violation was a “highly predictable consequence of
[its] failure to act.”
 Id. at 929 
(citing Bryan 
County, 520 U.S. at
409
). The trial evidence allowed the jury to infer that policy-
makers “noticed what was going on and by failing to do any-
thing must have encouraged or at least condoned” the mis-
conduct that caused the inmate’s death.
 Id. at 927. 
It did not
matter that no one had been hurt before, as the law allowed
no “one free suicide” pass.
 Id. at 929
.
    These teachings from the Supreme Court and our court
make plain that Monell liability based on a failure to act, at its
core, follows from a showing of constitutional violations
caused by a municipality’s deliberate indifference to the risk
of such violations. Sometimes the notice will come from a pat-
tern of past similar violations; other times it will come from
evidence of a risk so obvious that it compels municipal action.
But at all times and in all Monell cases based on this theory,
the Supreme Court has directed the focus on the presence and
proof of “a known or obvious” risk.
                                D
   The jury had ample evidence to find that Polk County’s
policy failures—both the prevention and detection gaps in its
24                                           Nos. 18-1498, et al.

written policies and the absence of training—occurred in the
face of an obvious and known risk that its male guards would
sexually assault female inmates.
    Start with the County’s affirmative obligation to protect its
inmates: “[W]hen the State takes a person into its custody and
holds [her] there against [her] will, the Constitution imposes
upon it a corresponding duty to assume some responsibility
for [her] safety and general well-being.” DeShaney v. Winne-
bago County Depʹt of Soc. Servs., 
489 U.S. 189
, 199–200 (1989).
Just as the healthcare contractor in Woodward shouldered a
constitutional duty to protect inmates from suicide, Polk
County bore the constitutional responsibility to protect its in-
mates from sexual assault. This requirement comes from the
Eighth Amendment, because “[b]eing violently assaulted in
prison is simply not part of the penalty that criminal offenders
pay for their offenses against society.” 
Farmer, 511 U.S. at 834
.
   Now consider the delicate setting for J.K.J. and M.J.J. They
were confined in circumstances where they depended on
male guards for nearly everything in their lives—their safety
as well as their access to food, medical care, recreation, and
even contact with family members. With this authority and
control for the guards came power and, in turn, access and
opportunity to abuse it. It is difficult to conceive of any setting
where the power dynamic could be more imbalanced than
that between a male guard and a female inmate. The jury
knew that from common sense—the reality was as obvious as
obvious could be—and they heard the point reinforced and
underscored through the testimony of J.K.J., M.J.J., and their
expert, Jeffrey Eiser. As J.K.J. aptly explained to the jury,
“there’s a male figure standing in front of me in a uniform
Nos. 18-1498, et al.                                           25

with a badge that has authority to do whatever he wants to
me.” The confinement setting is a tinderbox for sexual abuse.
    But there was more. J.K.J. and M.J.J. presented evidence
that the County was aware of sexual misconduct happening
within its jail, rendering the risk to female inmates far from
hypothetical. The jury learned that Captain Nargis knew of
sexual comments male guards made about female inmates.
This was especially consequential because Nargis was respon-
sible for creating and implementing the jail’s policies and
standards, and his actions therefore could be attributed to the
County for the purpose of Monell liability. See Gernetzke v. Ke-
nosha Unified Sch. Dist. No. 1, 
274 F.3d 464
, 468 (7th Cir. 2001)
(“The question is whether the promulgator, or the actor, as
the case may be—in other words, the decisionmaker—was at
the apex of authority for the action in question.”); see also
Pembaur v. City of Cincinnati, 
475 U.S. 469
, 481 (1986) (employ-
ing similar reasoning). Captain Nargis even admitted to him-
self participating in “tier talk,” jailhouse chatter that the jury
easily could have found amounted to sexually inappropriate
banter—in a word, harassment. The tier talk bespoke volumes
about the jail’s culture—the exact point the district court un-
derscored as part of rejecting Polk County’s Rule 50 motion
challenging the jury’s verdict. A reasonable jury could have
viewed the jail’s denigrating culture as confirming the unde-
niable risk that a guard would grow too comfortable, lose his
better angels, and step over the clear line marked in Polk
County’s written policies.
   If the County had looked the other way until this point,
the notice became undeniable when Captain Nargis learned
of Allen Jorgenson’s sexual misconduct against inmate N.S.
Recall the timing. Nargis learned of Jorgenson’s reported
26                                            Nos. 18-1498, et al.

wrongdoing no later than January 2012. It was then that N.S.
wrote her letter explaining the predatory and escalating na-
ture of Jorgenson’s sexual pursuit of her within the jail. Recall,
too, the details: N.S. informed Nargis that Jorgenson’s con-
duct began with watching her (and other female inmates)
shower, grew to requests to expose her body for him, and in
time intensified to forcibly touching her in a sexual manner—
all the while ordering her to “keep quiet.”
    A reasonable jury could have viewed the County’s learn-
ing of Jorgenson’s sexual exploitation of N.S. as sounding an
institutional alarm, making it “highly predictable,” if not cer-
tain, that a male guard would sexually assault a female in-
mate if the County did not act. Bryan 
County, 520 U.S. at 409
.
By that point the risk was not only obvious, but blatantly so.
To be certain, the accusations of Jorgenson’s reprehensible
conduct fell short of rape. But it would be naive in the extreme
to dismiss the misconduct as no more than boorish behavior
or, more to it, providing no incremental notice of an obvious
risk.
    The jury was not compelled to see Jorgenson’s conduct as
jailhouse horseplay, as guards somehow just being guards, or
anything of the sort. The evidence allowed the opposite con-
clusion: the jury was entitled to conclude that, separate and
apart from whatever discipline should befall Jorgenson, the
County had a plain example of predatory sexual behavior
staring it in the face. It took no imagination for the jury to see
parallels between Jorgenson’s escalating actions, cut short as
they were, and Christensen’s early abuse of J.K.J. and M.J.J.
See Cash v. County of Erie, 
654 F.3d 324
, 337 (2d Cir. 2011) (up-
holding a Monell jury verdict because “knowledge that an es-
tablished practice has proved insufficient to deter lesser
Nos. 18-1498, et al.                                         27

[sexual] misconduct can be found to serve [as] notice that the
practice is also insufficient to deter more egregious miscon-
duct”). That Jorgenson’s grooming of N.S. did not end with
rape is no liability shield; it was good fortune. See 
Woodward,
368 F.3d at 929 
(“That no one in the past committed suicide
simply shows that [the jail’s healthcare contractor] was fortu-
nate, not that it wasn’t deliberately indifferent.”).
    And with red lights flashing, Polk County chose the one
unavailable option—doing nothing. It did not change its sex-
ual abuse policy, institute a training, inquire of female in-
mates, or even call a staff meeting. With the writing on the
wall, Polk County deliberately chose to stand still, or at least
a reasonable jury could have so concluded.
    Having taken no action despite the obvious and known
risk of sexual assaults in its jail, Polk County could not claim
a lack of notice, much less surprise, upon learning that Chris-
tensen sexually assaulted J.K.J. and M.J.J. The County’s writ-
ten policies were lacking and its training on the topic was
barely existent. Even if the County somehow harbored a dif-
ferent perspective, that view became untenable upon learning
of Jorgenson’s misconduct. The jury could have viewed the
allegations of Jorgenson’s wrongdoing as exposing as false
any belief the County may have had that its barebones written
policy and training were enough to protect its female inmates
from sexual abuse. The County’s inaction following
knowledge that the existing program was not working—
Jorgenson’s sexual misconduct underscored that reality—was
sufficient to demonstrate deliberate indifference. Or, as the
Supreme Court put the same point in City of Canton, the ne-
cessity to act, whether by different training or new preventa-
tive measures, was “so obvious” and “the [existing]
28                                          Nos. 18-1498, et al.

inadequacy so likely to result in a violation of constitutional
rights” that a deliberate choice to stay the course could be
seen as a policy for which Polk County bears legal responsi-
bility. 489 U.S. at 390
. We likewise reasoned in Glisson that
“the key is whether there is a conscious decision not to take
action” and the record allowed a reasonable jury to so find
because “the existence of the INDOC Guidelines [addressing
the coordination of medical care], with which Corizon [as the
prison’s healthcare provider] was admittedly familiar, is evi-
dence that could persuade a trier of fact that Corizon con-
sciously chose the approach that it took”—“not to adopt the
recommended polices—not for Glisson, not for 
anyone.” 849
F.3d at 380
, 381.
     All of this is doubly true when coupled with other evi-
dence of Polk County’s deliberate indifference to sexual
abuse. Remember that the County’s investigation of Jorgen-
son ended with the considered conclusion that a reprimand
was adequate discipline. But even the reprimand came with
jail officials assuring Jorgenson that the censure was “not a big
deal.” The jury could have viewed this slap on the wrist as
confirming the jail’s broken culture, as explaining why not
only the “tier talk” was allowed to go on—with Captain
Nargis himself participating in it as a way of fitting in and
earning the confidence of the guards under his supervision—
but also why Nargis’s “tizzy email” evinced such a dismissive
attitude toward sexual abuse training. See 
Cash, 654 F.3d at
338 
(holding the jury could infer deliberate indifference to the
risk of inmate sexual assault because the jail’s issuance of a
one-page memorandum was a “token response” to a prior in-
stance of lesser sexual misconduct).
Nos. 18-1498, et al.                                         29

    Viewing this evidence all together, the jury could have
found that Polk County did little to reinforce the dignity and
respect owed female (and indeed all) inmates and instead
seemed to enable a culture that condoned the sexual objectifi-
cation of the women in its custody. Through this lens, Chris-
tensen’s repeated sexual assaults of J.K.J. and M.J.J. were not
the result of a “single instance of flawed conduct” but rather
“based on repeated failures to ensure [the inmates’]
safety … as well as a culture that permitted and condoned vi-
olations of policies that were designed to protect inmates.”
Woodward, 368 F.3d at 929
. Risk that started as obvious (from
the confinement setting and power dynamic between male
guards and female inmates) was fully on display (following
the Jorgenson incident) within an institution that scoffed at
PREA, denigrated female inmates, and devoted not a word of
its policies or a minute of any training session to concrete
measures to prevent, detect, and respond to sexual assault.
The jury stood on solid evidentiary ground seeing the
County’s dormancy as more than oversight, but instead as de-
liberate inaction.
    We recognize that policies can always be more robust, and
training can always be more thorough. PREA is not a consti-
tutional standard, and jails are not required to adopt it. Our
federal structure leaves the choices to state and local authori-
ties.
    Our conclusion is more limited: the risks to female inmates
in the confinement setting are obvious—indeed, PREA owes
its very existence to that reality—and N.S.’s report of Jorgen-
son’s misconduct reinforced for Polk County that the risks
were real and acute in the jail. Faced with that notice, the
County had a legal obligation to act—to take reasonable steps
30                                            Nos. 18-1498, et al.

to reduce the obvious and known risks of assaults on inmates.
See 
Farmer, 511 U.S. at 844
–45; Ortiz v. Jordan, 
562 U.S. 180
, 190
(2011) (reinforcing Farmer’s requirement of a reasonable re-
sponse). Just as a municipality cannot issue firearms to new
police academy graduates, wish them Godspeed on the
streets, and hope the new officers exercise sound judgment
when deciding whether circumstances warrant the use of le-
thal force—the precise example the Supreme Court provided
in City of Canton—Polk County could not, knowing all that it
did about the risk within its jailhouse walls, dispatch male
guards to stand watch over its female inmates equipped with
nothing more than a piece of paper with a flat instruction not
to abuse those under their care. The jury had enough to con-
clude that Polk County deliberately chose a path of inaction
when that option was off the table.
                                E
    Much of the same evidence proving Polk County deliber-
ately indifferent to the constitutional consequences of its inac-
tion likewise illustrates that its indifference was the moving
force behind J.K.J. and M.J.J.’s injuries. “The high degree of
predictability” that constitutes notice, the Supreme Court has
emphasized, “may also support an inference of causation—
that the municipality’s indifference led directly to the very
consequence that was so predictable.” Bryan 
County, 520 U.S.
at 409
–10. Having established that the jury could conclude
that the risk of constitutional injury—here, the sexual as-
saults—was obvious, it took but a small inferential step for the
jury to find causation. And inferences were required, for find-
ing causation is not a mechanical exercise like working a math
problem and getting an answer, but instead requires jurors to
view evidence in its totality, draw on their life experiences
Nos. 18-1498, et al.                                          31

and common sense, and then reach reasonable conclusions
about the effects of particular action and inaction.
    The circumstantial evidence paved multiple roads for the
jury to travel to find that Polk County’s inaction caused J.K.J.
and M.J.J.’s constitutional injuries. The assaults did not end
until Christensen was reported, giving rise to an eminently
reasonable inference that if the County had put in place some
of Eiser’s proposed policies and training to prevent but espe-
cially detect sexual abuse, Christensen’s conduct would have
been exposed sooner, perhaps by one of his fellow guards. Or
it may have been reported by another inmate or even by the
victims themselves. J.K.J. and M.J.J. testified that they did not
feel comfortable reporting the abuse while they were still
within the jail. N.S. similarly waited before revealing the full
extent of Jorgenson’s conduct because she worried he would
make it “hell” for her there. If Polk County had different pol-
icies or training, its culture would have changed, including its
dismissive and flippant attitude toward sexual assault, and
these women or someone else may have felt able to report the
abuse at some point during the three-year period of Christen-
sen’s conduct.
    Because any of these inferences from the evidence would
have been reasonable, the jury was entitled to conclude that if
Polk County had taken action in response to the glaring risk
that its female inmates’ health and safety were in danger,
J.K.J. and M.J.J.’s assaults would have stopped sooner, or
never happened at all.
   The County presses a different view. It sees a guard’s sex-
ual abuse of an inmate as so patently wrong and so plainly
prohibited by Wisconsin law and the jail’s policy that no
amount of training and no enhancements to the institution’s
32                                            Nos. 18-1498, et al.

code of conduct could have made any difference. And this is
especially so, the County urges, given the lengths to which
Christensen went to hide his conduct. Put most bluntly, no
amount of training, no policy, no monitoring—nothing, liter-
ally nothing—could have prevented or detected what he did
to J.K.J. and M.J.J., or so the County would have it.
   The County’s narrow fixation on Christensen exposes its
error. Monell liability did not hinge on predictions about
whether Christensen would have brought himself to stop
abusing J.K.J. and M.J.J. Maybe more robust policies could
have fostered a zero-tolerance culture in which Christensen
would not have felt free to openly harass female inmates,
thereby opening the door to his escalating abuse. Or they
could have caused Christensen to curb his conduct because of
a greater risk of detection—whether from closer monitoring,
more frequent guard rotations, or a policy preventing male
officers from being alone with female inmates. But maybe not.
    The point need not detain us because the evidence allowed
the jury to conclude that the County’s acting to institute more
robust policies—foremost addressing prevention and detec-
tion—and then training on those policies would have resulted
in another correctional officer, an inmate, or even J.K.J. and
M.J.J. taking some step to stop Christensen’s sexual assaults.
    The evidence did not require the jury to accept as inevita-
ble that Christensen’s conduct was unpreventable, undetect-
able, and incapable of giving rise to Monell liability. Admit-
tedly, “[p]redicting how a hypothetically well-trained officer
would have acted under the circumstances may not be an easy
task for the factfinder.” City of 
Canton, 489 U.S. at 391
. Nor was
the jury compelled to conclude that the sexual abuse suffered
by J.K.J. and M.J.J. had one and only one cause. See Whitlock
Nos. 18-1498, et al.                                        33

v. Brueggemann, 
682 F.3d 567
, 583 (7th Cir. 2012) (“[T]here is
no rule demanding that every case have only one proximate
cause.”). The law allowed the jury to consider the evidence in
its entirety, use its common sense, and draw inferences as part
of deciding for itself. Indeed, the Supreme Court envisioned
precisely that approach in City of Canton, observing that the
causation inquiry often will be complicated but emphasizing
that “judge and jury, doing their respective jobs, will be ade-
quate to the 
task.” 489 U.S. at 391
.
                           *   *   *
    Darryl Christensen’s long-term abuse of J.K.J. and M.J.J.
more than justified the jury’s verdict against him. And the
jury was furnished with sufficient evidence to hold Polk
County liable not on the basis of Christensen’s horrific acts
but rather the County’s own deliberate choice to stand idly by
while the female inmates under its care were exposed to an
unmistakable risk that they would be sexually assaulted—a
choice that was the moving force behind the harm inflicted on
J.K.J. and M.J.J. The jury so concluded, and we AFFIRM.
34                                           Nos. 18-1498, et al.

    HAMILTON, Circuit Judge, concurring. I join Judge Scud-
der’s opinion for the court. In light of comments in the dis-
senting opinions, it is worth emphasizing that the Monell
claims against the county are based on much more than
whether guards knew right from wrong or knew that it was a
crime to have sex with inmates. The Monell claims are also
based on the county’s failure to monitor its guards and its fail-
ure to provide effective channels for complaints so as to dis-
courage such abuse.
    To illustrate the point, consider an analogy involving only
greed, rather than lust and a guard’s horrific abuse of power
over inmates. Any bank will train its tellers that they should
not steal and that theft is a crime. All tellers know that
whether they receive the training or not. Suppose, though,
that a bank’s managers fail to conduct regular audits of tellers’
cash drawers. Most tellers are honest despite the lack of over-
sight, but one gives in to temptation. Managers later discover
that the one teller has been stealing money for years. The risk
of embezzlement, even by tellers who know the law and the
rules, is obvious. So is the need for audits. The risk and need
are so obvious that the bank’s stockholders could easily find
that its managers (i.e., its policymakers) were not merely neg-
ligent but deliberately indifferent (i.e., reckless) toward this
obvious and known risk, even if only one teller gave in to the
temptation. The same logic applies here to Christensen, who
repeatedly gave in to the temptation to abuse his power over
inmates.
Nos. 18-1498, et al.                                         35

    EASTERBROOK, Circuit Judge, dissenting in part. I agree
with the majority that the verdict against Christensen is
sound and with Judge Brennan that the verdict against Polk
County is not. Because this appeal has occasioned so much
ink, and my assessment differs somewhat from that of my
colleagues, I have concluded that it would be helpful to state
briefly why I find the claim against the County lacking.
    The majority recognizes that the County’s stated policy—
no sexual contact between guards and inmates—satisfies the
Constitution. It faults the County for failing to train guards
about that policy. Yet the Constitution does not require
training. See Connick v. Thompson, 
563 U.S. 51 
(2011). Nor
does the Constitution require every municipality to imple-
ment current understandings of best practices, such as the
Prison Rape Elimination Act of 2003, 34 U.S.C. §§ 30301–09
(PREA). The duty is to avoid unconstitutional policies. We
are supposed to assess the validity of the policies—that is to
say, the policymakers’ decisions—not how well subordinates
implement those policies.
    Connick is the only decision in which the Justices assessed
on the merits a contention that a unit of government violated
the Constitution by inadequate training that failed to avert
one particular bad outcome. It rejected the claim. The rea-
sons the Court gave are true of the Jail as well. Christensen is
the one and only rapist among the guards; no prior, similar
incidents notified the County about looming problems. And
as soon as supervisors learned of Christensen’s misconduct,
the County ended his employment and put him in prison
himself. See slip op. 4. This could well have avoided liability
for an employer under Title VII of the Civil Rights Act of
1964, a statute that unlike 42 U.S.C. §1983 allows vicarious
36                                            Nos. 18-1498, et al.

liability. See Burlington Industries, Inc. v. Ellerth, 
524 U.S. 742
(1998); Faragher v. Boca Raton, 
524 U.S. 775 
(1998). Liability
under Monell v. New York City Department of Social Services,
436 U.S. 658 
(1978), is supposed to be harder and depend on
the validity of the policy.
   One major reason why Connick (and every other decision
by the Supreme Court in which failure to train was ad-
vanced as a theory of liability) found no municipal liability is
that the Court sees knowledge as the proper goal of training.
Canton v. Harris, 
489 U.S. 378
, 390 n.10 (1989), observed:
       [C]ity policymakers know to a moral certainty
       that their police officers will be required to ar-
       rest fleeing felons. The city has armed its offic-
       ers with firearms, in part to allow them to ac-
       complish this task. Thus, the need to train of-
       ficers in the constitutional limitations on the
       use of deadly force, see Tennessee v. Garner, 
471
       U.S. 1 
(1985), can be said to be “so obvious,”
       that failure to do so could properly be charac-
       terized as “deliberate indifference” to constitu-
       tional rights.
In other words, a policy such as “comply with the Fourth
Amendment” is useless to non-lawyers without information
about what compliance entails. Is it lawful to shoot a fleeing
felon? If the answer depends on ongoing danger, how much
danger justifies deadly force? A city that stops with “obey
the Constitution” lacks a genuine policy.
    Contrast that with Polk County. One statewide rule, re-
flected in a criminal law, forbids sexual intercourse or con-
tact between guards and prisoners. Wis. Stat. §940.225(2)(h).
Nos. 18-1498, et al.                                        37

The Jail reinforces this by forbidding “an intimate social or
physical relationship with a prisoner”. Another of the Jail’s
policies says:
       under no circumstances will any inmate be the
       object of verbal, physical, emotional, psycho-
       logical, or sexual harassment by facility staff.
       Any officer engaged in such actions is subject
       to disciplinary charges and/or termination.
I can see a need for explication about “emotional” or “psy-
chological” harassment, but anyone can understand the rule
against intimate physical relations between guards and in-
mates. The Jail made sure that every guard knew about this
rule. What training is required to get guards to grasp it? The
problem is not a want of comprehension (as in Canton’s hypo-
thetical) but a want of compliance. Yet subordinate employ-
ees’ failure to comply with a valid policy is not a ground of
liability against a municipality.
    The difference between comprehension and compliance
is vital to understanding when training is required. Plaintiffs
do not argue that training is essential to guards’ comprehen-
sion. It ought to follow that the County is not liable.
    I could imagine the possibility of liability—under a theo-
ry that a policy is irrelevant if it is nothing but a paper
tiger—when evidence shows that training makes all the dif-
ference between a policy that works and a policy that does
not. But plaintiffs have not made such an argument. Consid-
er two possibilities that might have been relevant if the goal
were to expand liability beyond anything the Supreme Court
has yet recognized.
38                                          Nos. 18-1498, et al.

     First, was Polk County Jail rife with sexual crimes com-
mitted by guards? Plaintiffs concede that the answer is no.
They acknowledge that there had not been an instance of
improper sexual contact before Christensen. Plaintiffs also
acknowledge that the Polk County Jail was no worse on re-
lated subjects than other institutions, some of which used the
training programs that the Jail did not. Plaintiffs’ expert tes-
tified that the Jail had a “good” record overall. My col-
leagues in the majority refer to “the jail’s broken culture”
(slip op. 28) but do not compare the Jail with the results
achieved by other institutions that require guards to spend
more time in training. If training does not improve out-
comes, its absence cannot spoil an otherwise valid policy.
    Second, does the social science literature, or any expert
evidence in this record, show that more training achieves
better compliance with simple rules such as “no intimate
contact between guards and inmates”? The plaintiffs did not
offer any such evidence, their expert conceded that there is
none, and my search through scholarly sources did not turn
any up. Unless evidence shows that more training markedly
decreases the prevalence of sexual misconduct by guards (or
by employees at private firms), liability based on the lack of
such training is just vicarious liability by another name.
    Law schools must offer courses in legal ethics, because
education can be vital to understand complex doctrines such
as the attorney-client privilege and the rules against conflict
of interest. But law schools do not try to “train” law students
not to steal from clients’ trust funds or rape people who
come to them in search of advice. Implementing rules
against theft and rape does not depend on imparting infor-
Nos. 18-1498, et al.                                         39

mation to would-be lawyers, so training is not how to
achieve compliance.
    Threats of criminal prosecution or losing one’s livelihood
offer better prospects of deterring malicious conduct. Polk
County threatened guards with both kinds of punishment,
and it carried through against Christensen. Those steps
show vividly that the Jail does not tolerate sexual abuse of
prisoners—that the policy is not just a cynical attempt to de-
flect liability. I do not see anything in the Constitution that
prevents a county from electing deterrence and incapacita-
tion as the means of enforcing its policies.
    A belief that people can be trained (or perhaps condi-
tioned) not to commit crimes underlies the rehabilitation
model of criminal punishment. But many years of scholarly
study failed to produce support for that model, which has
been abandoned. Today punishment for the purpose of re-
habilitation is forbidden, at least in the national courts. See
18 U.S.C. §3582(a); Tapia v. United States, 
564 U.S. 319 
(2011).
Faith in the power of training to reduce crime is no more ap-
propriate when applied to suits under §1983.
   In sum, plaintiffs have not tried to show that the rule
against guards’ intimate contact with prisoners is hard to
understand (in general, or for the Jail’s guards in particular).
That leaves nothing for a jury to consider. The suit fails for
legal reasons.
    Evidence that earlier violations of the Jail’s policy were
tolerated by slaps on the wrist would be better proof that the
“real policy” differed from the written one, but only if the
toleration were attributable to the County rather than to
subordinates. If policymakers create a valid rule that is sabo-
40                                           Nos. 18-1498, et al.

taged by persons lower in the hierarchy, liability is supposed
to fall on those persons rather than the governmental entity.
That is how Monell differs from respondeat superior.
    At all events, in a case based on a theory of single-
incident liability, which is how my colleagues treat this suit,
evidence about laxity concerning less-serious matters is ir-
relevant. Consider Connick once again: The Justices recog-
nized that the prosecutor’s office in Orleans Parish had vio-
lated the Brady doctrine repeatedly but held that this did not
show a toleration of wrongdoing. If that was not enough in
Connick, the Jail’s failure to control lewd talk or do more in
response to one guard’s sexual harassment is categorically
insufficient to make the County liable for a different guard’s
rapes.
    The question under Monell is not whether the County
could have done better at inducing compliance with its
rules. With the benefit of hindsight, that’s always possible.
The question is whether the County had a constitutional pol-
icy. If Monell is to be overruled, and vicarious liability estab-
lished, that should be done forthrightly (and by the Supreme
Court), rather than via the roundabout route the majority
has devised.
Nos. 18-1498, et al.                                          41

    BRENNAN, Circuit Judge, with whom BAUER and SYKES, Cir-
cuit Judges, join, dissenting in part.
    The majority opinion holds a municipal employer liable
under § 1983 for a failure to train and a failure to supplement
policies because its employee did what those policies and
training expressly forbade him to do.
    Liability is based on the single-incident theory hypothe-
sized in City of Canton, Ohio v. Harris, 
489 U.S. 378
, 390 n.10
(1989). The “rare” and “narrow circumstances” under which
that theory applies do not fit here. Connick v. Thompson, 
563
U.S. 51
, 63, 64 (2011). Nor does this case meet the stringent
fault and causation requirements set by the Supreme Court to
prove § 1983 liability.
    The majority opinion upholds a jury verdict finding a
county liable for a jail guard’s repeated rapes of two inmates.
It does so without any evidence that Polk County actually and
directly caused the plaintiffs’ terrible injuries, and no affirm-
ative link between the County’s policies and the guard’s
crimes.
    It is undisputed that these horrible crimes were perpe-
trated without the County’s knowledge. It is also undisputed
that no pattern of similar violations put the County on notice
of a need for specific training that would have prevented
these sexual assaults. Yet the majority opinion concludes the
same evidence that failed to show notice under pattern liabil-
ity shows notice under single incident liability, as well as cau-
sation.
   This court’s decision stands alone, unaided by precedent.
The Supreme Court has never ruled that a Monell claim based
on a municipality’s failure to act is viable in the absence of a
42                                                 Nos. 18-1498, et al.

pattern. No federal appellate court has extended the single-
incident exception to the sexual assault context. Nor has a fed-
eral appellate court applied that exception when the
employee’s compliance with the municipality’s policy and
training would have prevented the injuries. And no federal
appellate court has held that specialized training is required
for an employee to know that rape is wrong.
   Because the legally deficient constitutional claim against
the County never should have gone to the jury, I respectfully
dissent in part.1
I. Background
    The facts as I understand them are described below, re-
counting the evidence in a light most favorable to the verdict
against the County, Martin v. Marinez, 
934 F.3d 594
, 596 (7th
Cir. 2019), and noting disagreement with how some facts are
set forth in the majority opinion.
   M.J.J. and J.K.J. were female inmates at Polk County Jail at
various times between 2011 and 2014, and guard Darryl
Christensen admits he covertly and repeatedly sexually as-
saulted them. Plaintiffs suffered serious injustices and trau-
matic injuries because of Christensen. The question is
whether Polk County bears legal responsibility for Christen-
sen’s actions.
     A. Trial Evidence
   Plaintiffs made four principal allegations against the
County at trial: (1) the jail’s sexual assault policies and train-
ing were inadequate; (2) the jail tolerated sexually offensive

1The majority opinion is correct that Christensen’s appeal fails, and this
dissent is limited to the question of the County’s liability.
Nos. 18-1498, et al.                                          43

comments by guards; (3) a former guard’s alleged misconduct
shows the jail’s sexual assault policy was deficient; and (4) the
jail underutilized recommendations under the Prison Rape
Elimination Act, 34 U.S.C. §§ 30301–09. After considering the
evidence in support of these allegations, the jury reached a
verdict against the County.
       1. Policies and training
    The jury considered several County policies prohibiting
sexual contact between guards and inmates. Two stand out.
Policy I-100 forbids any mistreatment or harassment of in-
mates, explains inmates’ rights, and instructs staff 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 termina-
tion.” Similarly, policy C-202 prohibits any “intimate social or
physical relationship with a prisoner.” That policy also in-
forms jail staff that sexual contact with any inmate is a crimi-
nal 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 inter-
course by a correctional staff member with an inmate as a
Class C felony).
   In language consistent with these policies, inmates are
provided a handbook upon arrival at the jail that says:
       Every inmate has the right to be safe from sex-
       ual 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.
44                                            Nos. 18-1498, et al.

    Plaintiffs’ opinion 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. Christensen said
the same. At trial he acknowledged the jail trained him that
sexual contact with inmates is against jail policies and a fel-
ony. Specifically, Christensen testified:
        He knew his assaults violated jail policy;
        He was trained his assaults were a crime;
        He knew he was putting plaintiffs at risk;
        He never forgot that sex with inmates was a crime; and
        He did not require more training to know his assaults
         were a crime.
     The jail’s onboarding and continuing education programs
also instruct employees that sexual contact with prisoners is a
crime and never permitted. The Wisconsin Department of
Corrections annually approved these programs, requiring: (1)
eight to ten weeks of “field training,” during which a new cor-
rections 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.
   The majority opinion expresses concern that “the trial rec-
ord … contained no evidence showing that … anyone from
the County dedicated any portion of any training session to
reviewing the jail’s written policies.” Majority op. at p. 9. But
Sergeant Steven Schaefer of the Polk County’s Sheriff’s de-
Nos. 18-1498, et al.                                          45

partment, who worked at the jail from 2002 until 2015, testi-
fied that “[o]ne of the methods of training within [the] jail was
to have jail staff review policies on a routine basis.” “[T]rain-
ing on policies related to improper sexual relations or im-
proper relationships with inmates,” Schaefer continued, were
part of that “routine training,” including policies I-100 and
C-202. Despite Schaefer’s testimony to the contrary, 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.’” Plaintiffs’ counsel con-
tinued that misrepresentation in their appeal brief which de-
clares: “Sergeant Steven Schaefer also testified to never receiv-
ing any training regarding sexual assault”; and Schaefer
“agreed that [he] received no training on sexual assault at any
time.” Appellees’ Br. at 13.
    In addition to the training above, Schaefer testified “[cor-
rectional officers] were all required to attend” countywide
training on sexual harassment, and he provided the training
to new employees from time to time. This training, according
to Schaefer, instructed guards on the jail’s numerous prohibi-
tions on conduct between staff and inmates, including im-
proper comments, becoming too close or too familiar, sharing
personal information, sexual relationships, and sexual as-
saults. Schaefer agreed that improper relationships between
inmates and guards was “something that the jail as a whole
took very seriously.” Consistent with that concern, Schaefer
testified that “Christensen had dishonored and disgraced eve-
rything the sheriff’s department was about” and that
Christensen’s behavior was “a betrayal of the [jail’s] ethical
standards.”
46                                            Nos. 18-1498, et al.

    The majority opinion describes the County’s training as
“vague cautions” and “nearly nonexistent.” Majority op. at p.
10. It states the jury heard evidence about jailers’ training on
maintaining distance from inmates, but “no testimony to sug-
gest that training touched the topic of sexual assault” and that
Schaefer “expressly admitted that much.”
 Id. It also 
states that
Schaefer “denied any memory of being told that it was im-
proper for jail officers to have sexual relationships with in-
mates.”
 Id.
    Schaefer testified 
differently than that. He stated when he
trained new jail officers, he trained them not to engage in too
familiar relationships with inmates, and that “too familiar”
would include an officer having sex with an inmate. He also
agreed the jail’s training included training on policies about
“improper sexual relations or improper relationships with in-
mates.” Schaefer also attested he received training on Polk
County Jail’s policies including C-202 (fraternization with in-
mates) and specifically those provisions prohibiting physical
relationships with prisoners.
       2. Inappropriate speech
     Plaintiffs alleged that jail staff routinely made sexually in-
appropriate comments about female inmates. They pointed to
Captain Scott Nargis, who oversaw daily operations of the
jail. During Nargis’s adverse examination at trial, plaintiffs’
counsel asked if he ever “engaged in tier talk which is not nec-
essarily flattering talk amongst co-workers,” and Nargis re-
sponded “yes.” Nargis also agreed that he participated in tier
talk “on occasion” to establish trust among subordinate offic-
ers. When Nargis admitted to “tier talk,” he did so within
plaintiffs’ definition (“not necessarily flattering talk”). Plain-
tiffs did not present evidence that tier talk connoted “sexual
Nos. 18-1498, et al.                                                       47

talk,” that Nargis’s “tier talk” was sexually explicit, or that
Nargis made comments sexual in nature with, about, or
around inmates or guards. The only reference at trial to “tier
talk” occurred during plaintiffs’ cross-examination of
Nargis.2
   The majority opinion accepts plaintiffs’ new, sexually-
charged definition of “tier talk.” Majority op. at p. 19 (finding
that “tier talk” implied “acceptance of sexual harassment at
the jail’s highest levels”).3 While this court views the facts in
the light most favorable to the jury’s verdict, it need not draw
unreasonable inferences, see Tindle v. Pulte Home Corp., 
607
F.3d 494
, 496 (7th Cir. 2010), or define terms differently than
they were used at trial.
     The evidence at trial of sexually inappropriate remarks by
jail staff was as follows: (1) J.K.J. testified two officers over-
heard Christensen making flirtatious comments to inmates;
(2) Christensen testified he overheard three guards make sex-
ual comments to inmates; and (3) Nargis testified that over a
twelve-year period Christensen made one sexually inappro-
priate comment and “it could have happened” that Christen-
sen made another.




2 Plaintiffs’ counsel in their appeal brief were less than candid with this
court when they wrote: “Captain Nargis routinely engaged in sexually ex-
plicit ‘tier talk.’” Appellees’ Br. at 14.
3 The district court’s order on the County’s Rule 50 motion also assumes
“tier talk” had a sexual implication despite the lack of any trial evidence
or definition that “tier talk” included a sexual component. See Opinion
and Order at 8–9, J.K.J. v. Polk Cty., No. 15-CV-428 (W.D. Wis. Feb. 5, 2018),
ECF No. 279.
48                                               Nos. 18-1498, et al.

    The specifics of these comments are as follows: The flirta-
tious comments overheard by J.K.J. and Christensen were
unreported, unspecific, and undated. Neither J.K.J. nor Chris-
tensen offered details on the alleged comments or the
timeframe in which they occurred, and there was no evidence
the County had notice of these comments. Nargis testified
that during Christensen’s twelve-year employment, he did
not recall but “it could have happened” that Christensen com-
mented on a female’s “rear end.” Nargis also recalled being
told that Christensen once remarked about an inmate’s
breasts. The majority opinion extrapolates from these two in-
stances when it states: “Captain Nargis confirmed hearing
that Christensen had made inappropriate sexual comments
about women in general.” Majority op. at p. 7.
       3. Investigation of former guard
   No one disputes that Christensen’s assaults were unprec-
edented and hidden from jail officials. Apart from Christen-
sen, the jail’s history contains only one other allegation of
sexual contact between a jail guard and an inmate: another
inmate saw guard Allen Jorgeson put his arm around inmate
N.S.’s waist and “pat her on the butt.” This occurred in 2012,
two years before Christensen’s assaults were discovered.4
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 Ste-
ven Moe to further investigate Jorgenson.



4 Christensen’s assaults began in 2011, and the County first learned of
them on October 29, 2014.
Nos. 18-1498, et al.                                          49

    The findings of that investigation revealed that another in-
mate believed Jorgenson and N.S. had an “inappropriate re-
lationship” but “no physical relationship,” and that Jorgenson
allegedly misused a jail camera to focus on female inmates
longer than necessary. Nargis and Moe did not limit their in-
vestigation to internal inquiries, reaching out to former in-
mates as part of their review. Because of inconsistent witness
accounts, including N.S.’s denial, Nargis and Moe could not
confirm that Jorgenson engaged in any sexual contact with
N.S. They did conclude that Jorgenson’s affiliation with N.S.
violated jail policy. As a result, Jorgenson was issued a writ-
ten reprimand. Up until this point, N.S. continued to deny
Jorgenson acted improperly.
    Days after Jorgenson’s reprimand, N.S. recanted her deni-
als in a letter to Nargis. N.S.’s letter stated that Jorgenson
made sexually harassing gestures and indecent remarks, and
said Jorgenson put his arm around N.S.’s waist and touched
her ”back and butt.” As a result of this letter Nargis and Moe
reopened the investigation “to take a whole fresh look at the
situation.” After this second review, Nargis and Moe again
could not confirm N.S.’s allegations and decided the repri-
mand of Jorgenson remained the proper 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.
    Jorgenson made sexually inappropriate comments to fe-
male coworkers, which the County does not dispute. When
staff first notified jail authorities of Jorgenson’s behavior to-
wards coworkers, an investigation ensued, leading to Jorgen-
son’s resignation. It is undisputed that Jorgenson’s comments
to staff went unreported until the N.S. investigation. It is also
50                                          Nos. 18-1498, et al.

undisputed that in the nine years preceding trial, 14,100 in-
mates came though the jail, and the Jorgenson circumstance
was the only known allegation of an improper relationship
between a guard and an inmate.
    The majority opinion recounts N.S.’s letter in detail. The
letter was admitted, however, only for the limited, non-hear-
say purpose of notice. The letter could not be considered on
its merits, which the majority opinion appears to do. Majority
op. at pp. 26–27 (“That Jorgenson’s grooming of N.S. did not
end with rape is no liability shield; it was good fortune.”).
    On Schaefer’s testimony about Jorgenson, the majority
opinion states “Schaefer reported that Jorgensen had touched
N.S. on her waist and rear end, adding that the complaints did
not come as a surprise because ‘[w]e have all heard com-
plaints about [Jorgenson’s] inappropriate comments to both
inmates and staff.’” Majority op. at p. 4. But Schaefer’s quoted
testimony spoke of improper comments, not improper touch-
ing. As to those comments, Schaefer could not recall an exam-
ple of Jorgenson making an improper comment to an inmate.
Schaefer also testified that none of the Jorgenson comments to
staff or inmates that he knew of rose to a level requiring dis-
cipline.
       4. Prison Rape Elimination Act (PREA)
    PREA played a large role in plaintiffs’ case. They argued
the jail failed to implement its policy recommendations,
showing a lack of concern for preventing and detecting sexual
assaults. Plaintiffs claimed Nargis openly “deni-
grated … PREA standards.” For this assertion, plaintiffs cited
a 2014 email from Nargis to jail staff which stated:
Nos. 18-1498, et al.                                               51

       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 called this the “tizzy” email. Plaintiffs ar-
gued Nargis’s choice of the word “tizzy” was “mocking” of
PREA and “indicat[ed] that he disliked PREA.” Plaintiffs also
claimed the email never discussed any specific PREA
measures, but merely restated the jail’s current anti-sexual as-
sault policies. Plaintiffs argued the “tizzy” email proves that
Nargis and the jail “consciously disregarded” PREA stand-
ards and thus disregarded the risk of sexual assaults at the
jail.
    Plaintiffs’ opinion witness Eiser grounded his testimony
on PREA’s recommendations, opining that the jail’s sexual as-
sault policies and training were inadequate because they did
not fully adopt several components of PREA. Eiser also testi-
fied there is no empirical data that compliance with PREA
yields better results. And he conceded that PREA-compliance
is not mandatory for county jails in Wisconsin, and that its
standards are better viewed as “best practices.” Eiser agreed
the County had a good record on sexual assaults—even fac-
toring in Jorgenson’s misconduct—because of the lack of inci-
dents of sexual contact between guards and inmates, much
less criminal sexual assaults like Christensen’s.
    Plaintiffs agreed that state law, not PREA, governs county
jails in Wisconsin, but they offered no evidence that the jail’s
sexual assault policies or training fell below state legal or ad-
ministrative standards. As for compliance with state law, the
County presented evidence that the Wisconsin Department of
52                                           Nos. 18-1498, et al.

Corrections annually reviews the jail’s policies, including its
policy prohibiting fraternization with inmates. In each year of
plaintiffs’ incarcerations, that Department 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.
II. Municipal Liability and the Law of Monell
    Municipal liability under § 1983 as set by the Supreme
Court has traveled a winding route. But that route has a con-
stant beacon to courts: each case examines what federal power
may be exercised over state and municipal governments and
considers the Court’s desire to harmonize § 1983 with the
structural limits of federalism. These precedents are disposi-
tive here and warrant detailed review before their applica-
tion. They prescribe a different outcome than reached by my
colleagues in the majority.
    The first is Rizzo v. Goode, 
423 U.S. 362
, 365–66 (1976), in
which the Court struck down a district court order requiring
police training reforms for violating federalism principles.
The Court held “there was no affirmative link between the oc-
currence of the various incidents of police misconduct and the
adoption of any plan or policy by petitioners … showing their
authorization or approval of such misconduct.”
 Id. at 371. 
The
Court also dismissed plaintiffs’ argument that “even without
a showing of direct responsibility for the actions of a small
percentage of the police force,” the local government’s failure
to act in the face of a pattern is actionable.
 
Id. at 375–76. 
The
Court distinguished the mere failure to act in the face of a pat-
tern of incidents from “active conduct,” rejecting the argu-
ment that a pattern alone creates a “constitutional duty” to
Nos. 18-1498, et al.                                            53

develop preventative procedures to deter speculative abuses.

Id. at 372, 
375–76, 378. The Court also rejected “invocation of
the word ‘pattern’ in a case where … the defendants are not
causally linked to it.”
 
Id. at 375–76.
    Federalism, the Court made clear in Rizzo, governed its
holding.
 Id. at 377–80. 
The district court’s injunctive relief, the
Court ruled, had “departed from the[] precepts” of federalism
“[w]hen it injected itself … into the internal disciplinary af-
fairs of this state agency.”
 Id. at 380. 
“[T]he principles of fed-
eralism,” the Court admonished, “play such an important
part in governing the relationship between federal courts and
state governments … .”
 Id. And those 
principles apply when
relief is sought against local governments.
 Id.
    Rizzo formulated 
the heightened causation requirement
between a policy and a constitutional violation now integral
to all § 1983 claims. See Polk Cty. v. Dodson, 
454 U.S. 312
, 326
(1981) (citing 
Rizzo, 423 U.S. at 370
–77, for the proposition that
an official policy “must be the moving force” of the constitu-
tional violation). It also constrained federal courts from “sec-
ond-guessing municipal employee-training programs” to
avoid “implicat[ing] serious questions of federalism.” 
Canton,
489 U.S. at 392 
(citing 
Rizzo, 423 U.S. at 378
–80).
    In Monell v. New York City Dept. of Social Servs., 
436 U.S. 658
(1978), the Court adopted Rizzo’s federalism-influenced
requirements and set parameters for establishing municipal
liability. In Monell, the Court considered “[w]hether local gov-
ernmental officials and/or local independent school boards
are ‘persons’ within the meaning of 42 U.S.C. § 1983.”
 Id. at
662. 
Based on “[a] fresh analysis of the debate on the Civil
Rights Act of 1871,”id. at 665, the Court held that the Act
54                                            Nos. 18-1498, et al.

“compell[ed] the conclusion that Congress did intend munic-
ipalities and other local government units to be included
among those persons to whom § 1983 applies.”
 Id. at 690. 
So
local governments could properly be sued as “persons”
within the meaning of § 1983.
    Although Monell exposed municipalities to § 1983 law-
suits, the Court circumscribed how that statute applied to
them. In particular, the Court limited claims against munici-
palities in two ways by holding that liability under § 1983 can-
not be predicated on a theory of respondeat superior, and that a
municipality’s own policy or custom must be the “moving
force” behind the constitutional violation.
 Id. 691–92. The
Court interpreted the statute’s use of the verbs “subject[s], or
causes to be subjected” as requiring a municipality’s direct in-
volvement, as opposed to liability on a vicarious basis.
 Id. at
692. 
Thus, the court ruled that the “mere right to control with-
out any control or direction being exercised” cannot support
§ 1983 liability. Id. (citing 
Rizzo, 423 U.S. at 370
–71). For those
reasons, a local government is liable under § 1983 only “when
execution of a government’s policy or custom … inflicts the
injury.”
 Id. at 694. 
The policy or custom must be the “moving
force” behind, or actual cause of, the constitutional injury.
 Id.
    The municipality 
in Monell officially adopted an unconsti-
tutional policy, so the municipality itself “unquestionably”
caused the constitutional violation.
 Id. at 690, 
694–95. Monell
left unanswered whether plaintiffs could establish municipal
liability by alleging unofficial policies, especially those of mu-
nicipal inaction or inadequate training, such as here. Over
three decades the Court has filled in those blanks. Each time
the Court has set the requirements to establish municipal lia-
bility for failure to train, liability has become more difficult to
Nos. 18-1498, et al.                                           55

find. These decisions always steer clear of respondeat superior.
Underlying those decisions, and the requirements they im-
pose, are the federalism principles articulated in Rizzo and the
Court’s intent to harmonize § 1983 with those principles.
    First of the municipal training cases was City of Oklahoma
City v. Tuttle, 
471 U.S. 808 
(1985). In Tuttle, seven Justices re-
jected an instruction that permitted the jury to infer liability
attributable to inadequate training from a single incident of
unconstitutional activity by a non-policymaking employee.

Id. at 821 
(plurality opinion) (“We think this inference unwar-
ranted … [because it] allows a § 1983 plaintiff to establish mu-
nicipal liability without submitting proof of a single action
taken by a municipal policymaker.”);
 id. at 830 
(Brennan, J.,
concurring in part and concurring in the judgment) (“Under
the instruction in question, the jury could have found the city
liable solely because [the officer]’s actions on the night in
question were so excessive and out of the ordinary. A jury
finding of liability based on this theory would unduly
threaten petitioner's immunity from respondeat superior liabil-
ity.”).
    Four years later, the Court further shaped the contours of
inadequate training liability in Canton. There, the plaintiff
sued the city for failing to adequately train its police when to
summon medical care for injured 
detainees. 489 U.S. at 381
.
On the question of fault, the Court held that a claim of inade-
quate training triggers municipal liability “only where the
failure to train amounts to deliberate indifference to the rights
of persons with whom [municipal employees] come into con-
tact.”
 Id. at 388. 
A municipality can be liable under § 1983
“[o]nly where a failure to train reflects a ‘deliberate’ or ‘con-
scious’ choice” not to fully train employees.
 
Id. at 389
. 
Such a
56                                                   Nos. 18-1498, et al.

deliberate choice could be shown when “in light of the duties
assigned to specific officers or employees the need for more
or different training is so obvious, and the inadequacy so likely
to result in the violation of constitutional rights, that the poli-
cymakers of the city can reasonably be said to have been de-
liberately indifferent to the need.”
 
Id. at 390 
(emphasis
added).
    Canton offered two examples of what “so obvious” means.
First, police may “so often violate constitutional rights that
the need for further training must have been plainly obvious
to the city policymakers.”
 
Id. at 390 
n.10.5 In the second exam-
ple, the Court left open the possibility—now known as the
“single-incident theory”—that a pattern of similar violations
might not be necessary to show deliberate indifference. To
prevail under this theory, a plaintiff must prove that munici-
pal “policymakers know to a moral certainty” that its employ-
ees will confront a given situation and fail to train for it.
 Id.
   To illustrate 
a single incident scenario, the Court posed the
hypothetical of a city that arms its police force with guns and
deploys those officers into the public without training. As the



5 Expounding on this principle, and foreshadowing the Court’s holdings
in Bryan Cty., Okl. v. Brown, 
520 U.S. 397
, 407–08 (1997), and Connick v.
Thompson, 
563 U.S. 51
, 62 (2011), Justice O’Connor wrote in concurrence:
“[M]unicipal liability for failure to train may be proper where it can be
shown that policymakers were aware of, and acquiesced in, a pattern of
constitutional violations … .” City of Canton, Ohio v. Harris, 
489 U.S. 378
,
397 (1989) (O’Connor, J., concurring in part and dissenting in part). “Al-
lowing an inadequate training claim” without proof of a pattern “to go to
the jury based upon a single incident,” Justice O’Connor continued,
“would only invite jury nullification of Monell.”
 Id. at 399.
Nos. 18-1498, et al.                                           57

Court explained, given the known frequency with which po-
lice encounter fleeing felons plus the “predictability that an
officer will lack specific tools to handle those situations,” “the
need to train officers in the constitutional limitations on the
use of deadly force … can be said to be ‘so obvious’ that the
failure to do so could properly be characterized as ‘deliberate
indifference’ to constitutional rights.”
 
Id. at 390 
n.10; see also
Connick, 563 U.S. at 63 
(explaining same) (quoting Bryan 
Cty.,
520 U.S. at 409
).
    But in either scenario, a plaintiff cannot establish that a
municipality acted with deliberate indifference for failing to
train employees for rare or unforeseen events. Relying on
Rizzo, the Court in Canton stressed the need for a rigorous
fault standard because:
       [A] lesser standard of fault would result in de
       facto respondeat superior liability on municipali-
       ties … . It would also engage federal courts in
       an endless exercise of second-guessing munici-
       pal employee-training programs. This is an ex-
       ercise we believe the federal courts are ill suited
       to undertake, as well as one that would impli-
       cate serious questions of federalism.

Id. at 392 
(citing 
Rizzo, 423 U.S. at 378
–80).
   As for causation, Canton cautioned that just because “a
particular officer may be unsatisfactorily trained will not
alone suffice to fasten liability on the city, for the officer’s
shortcomings may have resulted from factors other than a
faulty training program.”
 
Id. at 390
–91 (citations omitted).
“And plainly, adequately trained officers occasionally make
mistakes; the fact that they do says little about the training
58                                           Nos. 18-1498, et al.

program or the legal basis for holding the city liable.”
 
Id. at
391. 
“Neither will it suffice,” the Court stated, “to prove that
an injury or accident could have been avoided if an officer had
had better or more training, sufficient to equip him to avoid
the particular injury-causing conduct.”
 Id. Only when 
a plain-
tiff “prove[s] that the deficiency in training actually caused
the [employee]’s indifference” can liability attach.
 Id. Even
more, 
“the identified deficiency in a city’s training program
must be closely related to the ultimate injury.”
 
Id. at 391.
    In Canton, the Court gave reasons for this stringent causa-
tion standard. A failure-to-train claim “could be made about
almost any encounter resulting in injury, yet not condemn the
adequacy of the program to enable officers to respond
properly to the usual and recurring situations with which
they must deal.”
 Id. And “[i]n 
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 some-
thing the city ‘could have done’ to prevent the unfortunate
incident.”
 
Id. at 392 
(quoting 
Tuttle, 471 U.S. at 823 
(plurality
opinion)).
    Eight years later, the Court imported Canton’s failure-to-
train principles into the hiring context in Bd. of Cty. Comm’rs
of Bryan Cty. v. Brown, 
520 U.S. 397 
(1997). There a deputy
sheriff who caused the plaintiff’s injury had received no train-
ing in proper pursuit and arrest techniques.
 Id. at 400–02. 
The
Court addressed whether the county’s decision to hire the of-
fending deputy sheriff triggered liability under Monell:
“Where a plaintiff claims … the municipality has not directly
inflicted an injury, but nonetheless has caused an employee to
do so, rigorous standards of culpability and causation must
be applied to ensure that the municipality is not held liable
Nos. 18-1498, et al.                                            59

solely for the actions of its employee.”
 Id. at 405 
(citing 
Canton
489 U.S. at 391
–92, and 
Tuttle, 471 U.S. at 824 
(plurality opin-
ion)). And “[w]here a court fails to adhere to rigorous require-
ments of culpability and causation, municipal liability
collapses into respondeat superior liability.”
 
Id. at 415.
    “Canton makes clear,” the Court explained in Bryan
County, that “deliberate indifference is a stringent standard of
fault, requiring proof that a municipal actor disregarded a
known or obvious consequence of his action.”
 
Id. at 410 
(in-
ternal quotation marks omitted). A “plaintiff must demon-
strate that a municipal decision reflects deliberate indiffer-
ence to the risk that a violation of a particular constitutional or
statutory right will follow the decision.”
 Id. at 411 
(emphasis
added). In other words, the deprivation of a federally pro-
tected right must be the “plainly obvious consequence” of the
municipality’s decision.
 Id. Bryan County 
instructs courts to
focus on the particular constitutional violation that occurred,
not constitutional violations in general.
    Subject to this particularity requirement, Bryan County
identified three ways a municipality could be liable for
inadequate training. First, “[i]f a program does not prevent
constitutional violations, municipal decisionmakers may
eventually be put on notice that a new program is called for.”

Id. at 407. 
Second, “the existence of a pattern of tortious con-
duct by inadequately trained employees may tend to show
that the lack of proper training, rather than a one-time negli-
gent administration of the program or factors peculiar to the
officer involved in a particular incident, is the ‘moving force’
behind the plaintiff's injury.”
 Id. at 407–08. 
And third, the sin-
gle-incident theory as laid out in 
Canton, 489 U.S. at 390 
n.10:
in a “narrow range of circumstances, a violation of federal
60                                            Nos. 18-1498, et al.

rights may be a highly predictable consequence of a failure to
equip law enforcement officers with specific tools to handle
recurring situations.” Bryan 
Cty., 520 U.S. at 409
. In this third
scenario, liability hinges on the likelihood that “a violation of
a specific constitutional or statutory right” will recur and “the
predictability that an officer lacking specific tools to handle
that situation will violate citizens’ rights.”
 Id. at 409.
    Finally, we come to Connick v. Thompson, 
563 U.S. 51
(2011). There, the plaintiff spent 18 years in prison, including
14 years on death row, because an assistant district attorney
willfully suppressed blood evidence that exculpated plaintiff.

Id. at 55–56. 
The plaintiff alleged that violation was caused by
the district attorney’s deliberate indifference to an obvious
need to train prosecutors to avoid violations of Brady v.
Maryland, 
373 U.S. 83
, 87 (1963), to disclose evidence favorable
to the defense. The jury agreed, awarding plaintiff $14 million
in damages, and the district court added more than $1 million
in attorney’s fees and 
costs. 563 U.S. at 57
.
   The Supreme Court reversed the jury verdict in Connick.
In fact, in each of the post-Monell cases discussed—Tuttle,
Canton, Bryan County, and Connick—the Court reversed a jury
verdict for the plaintiff.
    Connick reflects the Court’s most recent and most exhaus-
tive assessment of inadequate training liability. Because such
a claim treads so closely to vicarious liability, the Court ad-
monished: “A municipality’s culpability for a deprivation of
rights is at its most tenuous where a claim turns on a failure
to train.”
 Id. at 61 
(citing 
Tuttle, 471 U.S. at 822
–23). The Court
recognized two types of inadequate training claims: those that
require a pattern of similar constitutional violations, and
Nos. 18-1498, et al.                                          61

those that may succeed without a pattern under single-inci-
dent theory.
 
Id. at 71
–72.
    Per Connick, to prevail under a pattern theory a plaintiff
must prove “[a] pattern of similar constitutional violations by
untrained employees.”
 
Id. at 62
. 
“[C]ontemporaneous or sub-
sequent conduct cannot establish a pattern of violations that
would provide ‘notice to the cit[y] and the opportunity to con-
form to constitutional dictates … .’”
 Id. at 63 
n.7 (quoting
Canton, 489 U.S. at 395
). A pattern of constitutional violations
means prior violations “similar to” the “specific” violation
suffered by the plaintiff.
 
Id. at 62
–63, 
63 n.7, 67. In Connick,
despite at least four prior Brady violations in the same district
attorney’s office, and that up to four prosecutors may have
been responsible for the nondisclosure of favorable evidence
that exculpated plaintiff, the Court held that the plaintiff had
failed to establish a pattern of similar violations.
    Because the plaintiff in Connick could not rely on a pattern
of similar Brady violations, the Court addressed whether he
could prevail under the “single-incident” theory hypothe-
sized in Canton.
 Id. at 63
–71. In Connick, the Court set three
requirements to establish liability under Canton’s single-inci-
dent hypothetical:
   First, single-incident liability applies only when dealing
with “untrained employees.”
 Id. at 61
–62, 67; see also
 
id. at 91
(Ginsburg, J., dissenting) (agreeing with majority on this re-
quirement).
   Second, once it is established that the offending employee
was untrained, a plaintiff alleging single-incident liability
must prove “police officers have no knowledge at all” of the
required constitutional standards.
 Id. at 67. 
That means “in
62                                            Nos. 18-1498, et al.

the absence of training,” there must be “no way for novice
officers to obtain the legal knowledge they require.”
 Id. A
plaintiff 
meets this requirement when there is “no reason to
assume” the officer is “familiar with the constitutional con-
straints” of the prohibited conduct.
 
Id. at 64. 
That a policy has
“gray areas,” the Court cautioned, does not mean an em-
ployee “will so obviously make wrong decisions that failing
to train them amounts to ‘a decision by the city itself to violate
the Constitution.’”
 Id. at 71 
(quoting 
Canton, 489 U.S. at 395
(O’Connor, J., concurring in part and dissenting in part)). In
those situations, a plaintiff must show it was “highly predict-
able” that employees would be confounded by those gray ar-
eas and make incorrect decisions as a result.
 Id.
    Third, an 
absence of formal or specialized training does
not show deliberate indifference because “failure-to-train lia-
bility is concerned with the substance of the training, not the
particular instructional format.”
 
Id. at 68
. 
Put more simply,
under the single-incident theory there is not inquiry into the
subtleties of training. Instead, the key question to qualify un-
der Canton’s hypothetical is whether the municipal employee
is “equipped with the tools to interpret and apply legal prin-
ciples.”
 Id. at 64; 
Bryan 
County, 520 U.S. at 409
. The question is
not whether better or more training might have prevented the
violation.
    “[S]howing merely that additional training would have
been helpful in making difficult decisions does not establish
municipal liability.” 
Connick, 563 U.S. at 68
; see also
 id. at 73–
74 (Scalia, J., concurring) (rejecting liability “under the rubric
of failure to train simply because the municipality does not
have a professional educational program covering the specific
violation in sufficient depth”). Said another way, “’proving
Nos. 18-1498, et al.                                            63

that an injury or accident could have been avoided if an em-
ployee had had better or more training, sufficient to equip
him to avoid the particular injury-causing conduct’ will not
suffice.” Id. (quoting 
Canton, 489 U.S. at 391
).
    Applying these three requirements, the Court in Connick
contrasted Canton’s hypothetical with an attorney asked to
make a Brady determination. 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 obli-
gations, and on-the-job experience, to deal with Brady deci-
sions.
 Id. at 66-67. 
“In light of this regime of legal training and
professional responsibility, recurring constitutional viola-
tions are not the ‘obvious consequence’ of failing to provide
prosecutors with formal in-house training about how to obey
the law.”
 Id. at 66. 
The Court therefore concluded in Connick
that “this case does not fall within the narrow range of ‘single-
incident’ liability hypothesized in Canton.”
 
Id. at 71
–72.
    These precedents control the consideration of this case.
The Supreme Court has never held a municipality liable for a
failure to act in the absence of a pattern of prior similar viola-
tions. See Majority op. at p. 22 (recognizing same). This case
does not present such a claim either, for the reasons that fol-
low.
III. Theories of Liability
    The majority opinion holds the County employer liable for
the crimes of its employee Christensen under the single-inci-
dent theory hypothesized in Canton, specifically for failure to
train and for failure to supplement County policies. Majority
op. at 18-23. Those holdings rest on three conclusions:
64                                            Nos. 18-1498, et al.

    1. In the absence of a pattern of prior similar sexual as-
saults at the jail, the rapes of J.K.J. and M.J.J. by Christensen
pose one of those “rare” and “narrow range of circumstances”
(Connick, 563 U.S. at 64
, 71–72) the Court hypothesized in
Canton’s footnote 10, in which the need for training in consti-
tutional requirements is “so obvious ex ante” (Id. at 72) (con-
currence). Majority op. at pp. 20–30.
    2. The jail’s omission of sexual assault prevention and de-
tection measures in its written policies amounted to unconsti-
tutional inaction under Monell. Majority op. at pp. 18–20.
   3. The failure to train about sexual assault prevention and
detection measures, or the omission of such measures from
written policies, caused plaintiff’s injuries. Majority op. at pp.
30-33.
   In reaching these conclusions, the majority opinion de-
parts from the Supreme Court’s requirements in Canton,
Bryan County, and Connick and oversteps the culpability and
causation rules governing § 1983 claims, resulting in re-
spondeat superior liability, an outcome forbidden since Monell.
     A. Failure to Train
        1. Single-incident theory
    For a failure-to-train claim the standard of municipal fault
is deliberate indifference. Con
nick, 563 U.S. at 61 
(labeling this
“a stringent standard of fault”); 
Canton, 489 U.S. at 388
–89.
Only when a municipality has “actual or constructive notice
that a particular omission in their training program causes
city employees to violate citizens’ constitutional rights” may
the municipality be deemed deliberately indifferent for re-
taining that program. Con
nick, 563 U.S. at 61
. “Without notice
Nos. 18-1498, et al.                                            65

that a course of training is deficient in a particular respect, de-
cisionmakers can hardly be said to have deliberately chosen a
training program that will cause violations of constitutional
rights.”
 Id. at 62 
(emphasis added).
    The Court has recognized two ways to establish notice.
The first requires a prior pattern of similar constitutional vio-
lations to prove deliberate indifference, the second does not.
See 
Connick, 563 U.S. at 62
–63. Per Connick, the second—the
“single-incident theory”—is distinct from and serves as an ex-
ception to the pattern theory.
 Id. at 62
–63; 71–72. These two
theories share the same objective of discerning whether “the
need for further training” was “so obvious.” 
Canton, 489 U.S.
at 390 
n.10. Bryan County phrased this question as whether “a
municipal actor disregarded a known or obvious conse-
quence of his 
action.” 520 U.S. at 410 
(describing Canton’s
standard of fault); see also Con
nick, 563 U.S. at 61 
(quoting
same). Either way, the key term is “obvious.” See Farmer v.
Brennan, 
511 U.S. 825
, 841 (1994) (describing Canton’s applica-
tion of deliberate indifference as an “obviousness test”).
    Under the pattern theory, the obviousness of a risk is de-
termined from “[a] pattern of similar constitutional violations
by untrained employees.” 
Connick, 563 U.S. at 62
. Under the
single-incident theory, the question is whether the risk is “ob-
vious in the abstract.” Bryan 
County, 520 U.S. at 410 
(applying
Canton’s standard of fault) (internal quotation marks omit-
ted). The majority opinion says “[t]he Court did not leave the
[single-incident] liability point in any way abstract.” Majority
op. at p. 21. The Court in Bryan County said the opposite.
When evaluating “the risk from a particular glaring omission
in a training regimen” the question is whether the risk is “ob-
vious in the abstract.” Bryan 
County, 520 U.S. at 410 
(internal
66                                           Nos. 18-1498, et al.

quotation marks omitted). Under either theory an objective
test is applied. 
Farmer, 511 U.S. at 841 
(“It would be hard to
describe the Canton understanding of deliberate indifference,
permitting liability to be premised on obviousness or con-
structive notice, as anything but objective.”).
    The majority opinion recognizes that this case does not
present a pattern of misconduct which gave notice to the
County that its training program would cause constitutional
violations. Majority op. at p. 20. Nevertheless, it concludes the
need to supplement the County’s sexual assault training was
“so obvious” that the failure to do so amounted to deliberate
indifference under Canton’s single-incident theory. Majority
op. at p. 27. I respectfully part ways with my colleagues in the
majority that the requirements to establish single-incident li-
ability have been met here.
    As noted above, to illustrate “so obvious” notice of a need
for training under the single-incident theory, Canton hypoth-
esized a city deploying armed officers, untrained on the con-
stitutional limits of the use of deadly force, to capture fleeing
felons. 489 U.S. at 390 
n.10. In Connick, the Court distilled
Canton’s hypothetical into three single-incident liability re-
quirements. None of these are met here.
    First, single-incident liability applies only when dealing
with “untrained employees.” Con
nick, 563 U.S. at 61
–62, 67;
 
id.
at 91 
(Ginsburg, J., dissenting) (agreeing with the majority
that failure-to-train liability attaches “only when the failure”
involves “untrained employees”). It is undisputed that the jail
trained Christensen that sexual contact with inmates was
against jail policies and a felony, so this prerequisite is not
met.
Nos. 18-1498, et al.                                           67

    To meet Connick’s second requirement there must be “no
reason to assume” the County’s jail guards were “familiar
with the constitutional constraints” of the prohibited conduct.
Connick, 563 U.S. at 64
. But Christensen testified to five ways
he was familiar with the policy prohibiting sexual contact: (1)
he knew his conduct violated jail policy; (2) he was trained his
conduct was a crime; (3) he knew he was putting plaintiffs at
risk; (4) he never forgot that sex with inmates was a crime;
and (5) he did not require more training to know his conduct
was a crime. Unlike the nuanced and compound legal stand-
ards imposed by deadly force limits (Canton) and Brady’s
evidentiary obligations (Connick), even the majority opinion
recognizes this requirement cannot be met, as it characterizes
the jail’s zero-tolerance abuse policy as “categorical[],” a
“clear line,” as well as an “obvious” and “easy and evident”
constitutional parameter. Majority op. at pp. 18–19, 25. The
record does not show that County guards, let alone Christen-
sen, had “no knowledge at all” of the relevant constitutional
standard. 
Connick, 563 U.S. at 67
.
    On this second requirement, plaintiffs must prove the
County’s jail guards “ha[d] no knowledge at all” of the rele-
vant constitutional standard,
 id. at 67, 
here, a blanket prohibi-
tion against sexual assault. That means a court must find there
is “no way for novice [guards] to obtain the legal knowledge
they require” unless they are trained.
 
Id. at 64. 
The majority
opinion concludes that without training “male guards would
sexually assault female inmates.” Majority op. at p. 24. This
conclusion assumes that jails present an “opportunity to
abuse” and male guards will “obvious[ly]” exploit that op-
portunity unless trained otherwise.
 Id. As the 
majority opin-
ion puts it: “It is difficult to conceive of any setting where the
power dynamic could be more imbalanced than that between
68                                                    Nos. 18-1498, et al.

a male guard and female inmate,” so it is “obvious as obvious
could be” that male guards will “sexual[ly] abuse” female in-
mates. Id.6
    An opportunity in which abuse may be committed does
not establish deliberate indifference. At most it establishes
negligence, which “will not suffice” to establish § 1983 liabil-
ity. Bryan 
County, 520 U.S. at 407
. The record does not support
the assumption that, unless trained to refrain from sexual as-
saults, every male guard at Polk County Jail posed a risk to
the bodily integrity of female inmates. Nor can it be assumed
that absent training, male guards would have “obviously” vi-
olated behavioral norms and commit sex crimes.7 For these
reasons, Connick’s second requirement also is not met.


6 The majority opinion’s conclusion that the “power dynamic” of confine-
ment places municipalities on notice of the risk of sexual assaults is un-
likely to prove workable for district courts. If the unalterable nature of
confinement places municipalities on perpetual notice sufficient to estab-
lish § 1983 liability, how do courts determine if a municipality failed to act
in the face of that obviousness?
7 The majority opinion’s conclusion of obviousness also rests on a theory
rejected by other federal appellate courts. Whitson v. Stone Cty. Jail, 
602
F.3d 920
, 927 (8th Cir. 2010) (“We … could scarcely hold as a matter of
course that every male guard is a risk to the bodily integrity of a female
inmate whenever the two are left alone. Absent evidence to the contrary,
we assume that jailers will not violate models of social decorum or other-
wise commit a punishable offense.”); Barney v. Pulsipher, 
143 F.3d 1299
,
1310–11 (10th Cir. 1998) (refusing to find that the defendants knew of a
substantial risk of assault based on the proposition that every male guard
is a risk to the bodily integrity of a female inmate whenever the two are
left alone because in the record there was no evidence of sexual miscon-
duct in the offending jailer’s background nor was there evidence of previ-
ous incidents of sexual misconduct by jailers generally); Hovater v.
Nos. 18-1498, et al.                                                    69

    Third, “failure-to-train liability is concerned with the sub-
stance of the training, not the particular instructional format.”

Id. at 68
. 
Under Connick, this poses the question whether
Christensen was “equipped with the tools to interpret and ap-
ply legal principles.”
 
Id. at 64. 
The majority opinion frames
the applicable principle as “do not sexually abuse inmates,”
calling it an “easy and evident” constraint. Majority op. at pp.
19–20. Christensen thought so too, testifying he knew his as-
saults were crimes that put plaintiffs at risk, and that more
training would not have altered that knowledge. Despite
Christensen’s avowals, the majority opinion holds the County
liable for Christensen’s conduct because its sexual assault
training failed to educate and sensitize guards to the follow-
ing:
       the dignity and respect owed female inmates;
       the inherent vulnerability the confinement set-
        ting presents to female inmates;
       the symptoms of an inmate suffering from the
        trauma of abuse;
       report[ing] each other’s misconduct; and
       matters of prevention and detection.
Majority op. at pp. 19–20, 28.
   This scrutiny of existing policies and training goes too far,
violating the third requirement of Connick. The single-inci-
dent liability inquiry under Connick stops after the question of
whether Christensen knew his conduct was a crime. It does
not ask whether better or more training might have prevented


Robinson, 
1 F.3d 1063
, 1066–68 (10th Cir. 1993) (“[T]here is no evidence in
the present case of an obvious risk that male detention officers will sex-
ually assault female inmates if they are left alone.”).
70                                            Nos. 18-1498, et al.

his crimes. On this exact question the Court in Connick stated:
“’[P]rov[ing] that an injury … 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. at 68 
(quoting 
Canton, 489 U.S. at 391
). See also
Connick, 563 U.S. at 73
–74 (Scalia, J., concurring) (rejecting lia-
bility “under the rubric of failure to train simply because the
municipality does not have a professional educational pro-
gram covering the specific violation in sufficient depth”);
 id.
at 78 
(Scalia, J., concurring) (“[A] bad-faith, knowing viola-
tion, could not possibly be attributed to lack of training. …”)
    This third requirement ends where it does for good rea-
sons. As the Court ruled in Canton, a municipality cannot act
with deliberate indifference for failing to train employees for
rare or unforeseen events. A less than rigorous fault standard
“result[s] in de facto respondeat superior liability on municipali-
ties—a result we rejected in Monell.” 
Canton, 489 U.S. at 392
(citing 
Monell, 436 U.S. at 693
–94; 
Rizzo, 423 U.S. at 378
–80).
That result also intrudes upon the federalism interests that the
Monell failure-to-train cases strongly reinforce. In Canton, the
Court noted how “[i]n virtually every instance where a per-
son 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 inci-
dent.”
 
Id. at 392 
(quoting 
Tuttle, 471 U.S. at 823 
(plurality opin-
ion)). While the education and sensitization steps the majority
opinion lists fit into that category, the stringent fault require-
ments of Canton do not demand them.
    None of the three requirements in Connick are met here to
satisfy Canton’s single-incident hypothetical.
Nos. 18-1498, et al.                                           71

    The single-incident theory fails here for two other reasons.
Bryan County requires a “direct causal link” link between
Christensen’s crimes and an ineffective training 
regimen. 520
U.S. at 404
. See also 
Tuttle, 471 U.S. at 823 
(requiring “affirma-
tive link”); 
Rizzo, 423 U.S. at 371 
(same). The record here
shows no such link. Before the discovery of Christensen’s
crimes, there were no prior instances of similar sexual as-
saults; Eiser agreed the jail, including the Jorgenson incident,
had a “good” record on sexual assaults. And Eiser testified
that no empirical data shows compliance with PREA would
have yielded a better result. These same facts, however, per-
suade a majority of my colleagues that the need for more or
better sexual assault training was “so obvious.” To the con-
trary, the record shows the County’s training “equipped
[Christensen] with the tools” to understand its uncomplicated
zero-tolerance assault policy, Connick’s third 
requirement. 563
U.S. at 64
. Christensen chose repeatedly to disregard that pol-
icy and, hiding his actions, assault the plaintiffs.
    To establish single incident liability, a plaintiff also must
prove that municipal “policymakers know to a moral cer-
tainty” that its employees will confront a given situation and
fail to train for it. 
Canton, 489 U.S. at 390 
n.10. Here, the
County had not previously faced the circumstance of a guard
raping inmates. Even so, the jail had a zero-tolerance sexual
assault policy and trained its guards on it. In Connick, the
Court ruled that “[a] district attorney is entitled to rely on
prosecutors’ professional training and ethical obligations in
the absence of specific reason, such as a pattern of violations,
to believe that those tools are insufficient to prevent future
constitutional 
violations.” 563 U.S. at 67
. Connick also requires
similarity among violations. See
 id. at 62–63. 
Following
Connick’s reasoning, the County was also entitled to rely on
72                                              Nos. 18-1498, et al.

its policies and training given the absence of any sexual as-
saults similar to Christensen’s at the jail. See Majority op. at p.
26. (“To be certain, the accusations of Jorgenson’s reprehensi-
ble conduct fell short of rape.”).
       2. The “Obvious” Contradiction
    All agree this case turns on notice. The majority opinion
states at p. 20: “The Supreme Court has made plain that a fail-
ure to act amounts to municipal action for Monell purposes
only if the County has notice that its program will cause con-
stitutional violations. See Con
nick, 563 U.S. at 61
–62.” Canton
in footnote 10 provides that the need to train officers on con-
stitutional limits “can be said to be ‘so obvious’ that failure to
do so could properly be characterized as ‘deliberate indiffer-
ence’ to constitutional 
rights.” 489 U.S. at 390 
n.10. The evi-
dence the majority opinion relies on to conclude notice falls
outside the sphere of single-incident liability.
     The majority opinion concludes repeatedly and with cer-
tainty that the jail posed an obvious risk that male guards
would sexually assault female inmates. Majority op. at p. 24
(“obvious and known risk”),
 id. (“the reality 
was as obvious
as obvious can be”),
 

27 (“obvious and known risk of
sexual assaults in its jail”), and
 

30 (“the risk of consti-
tutional injury—here, the sexual assaults—was obvious”). In
support of this conclusion it reviews “evidence that the
County was aware of sexual misconduct happening within its
jail.”
 
Id. at p. 
25. That notice evidence falls into four categories:
(1) Nargis’s tier talk (
Id. at pp. 
19, 25, and 28); (2) the “tizzy”
email (
id. at pp. 
19 and 28); (3) infrequent, undated, and
mostly unreported inappropriate comments by four guards
over a twelve-year period (

25); and (4) information
gleaned from the Jorgenson investigation (
id. at pp. 
25–28).
Nos. 18-1498, et al.                                          73

On this evidence the majority opinion holds that the County
disregarded obvious notice of risk and the need for “different
policies or training” and “more robust policies.”
 
Id. at pp.
31-32.
   But in its consideration of and conclusion based on this
evidence, the majority opinion embraces a contradiction. The
identical four categories of evidence served as plaintiffs’ evi-
dence of notice in the district court. As in every appeal, we are
limited by the record. The district court expressly ruled “that
plaintiffs failed to put forth sufficient evidence to support
finding a pattern of constitutional violations known to policy-
makers.” Opinion and Order at 7, J.K.J. v. Polk Cty., No. 15-
CV-428, 
2018 WL 708390 
(W.D. Wis. Feb. 5, 2018), ECF No.
279. And plaintiffs have not challenged that holding on ap-
peal. Appellees’ Br. at 45 n.10 (“Plaintiffs do not argue on ap-
peal that deliberate indifference is established here due to a
pattern of similar past incidents.”).
    These four categories of evidence fail as “so obvious” no-
tice under a pattern theory, as found by the district court.
Nevertheless, in the majority opinion the same four categories
provide “so obvious” notice under a single-incident theory.
The majority opinion turns the single-incident theory on its
head, using the same facts to produce the opposite conclu-
sion. On this same record, the need for more training and bet-
ter policies is simultaneously “not at all obvious” and “so
obvious.” The lack of a pattern deprived the County of notice
of a need, but the same evidence now provides notice of such
a need. For example, the majority opinion allows Jorgenson’s
misconduct to provide “incremental notice of an obvious
74                                                 Nos. 18-1498, et al.

risk.” Majority op. at p. 26. Unexplained is why—if four cate-
gories of evidence did not provide that notice, as the district
court held—one category could.8
     If the response is that the “so obvious” notice necessary
for single-incident liability differs from the “so obvious” no-
tice for pattern liability, no authority so provides. Nor would
it. The same evidence is not being considered as of different
times, or in a different sense. Pattern and single incident the-
ories of liability differ in their requirements, but they share
the identical objective of notice. Under either theory, a munic-
ipality’s policymakers must have notice of omissions before
being deemed deliberately indifferent. Con
nick, 563 U.S. at
61
-62; Bryan 
Cty., 520 U.S. at 410
; 
Canton, 489 U.S. at 390 
n.10.
Evidence that falls short under one theory falls short under
the other as well.
    Do not construe my dissent as concluding that the four
categories of evidence listed above are always irrelevant.
Those categories would be relevant under other avenues of
liability, such as implied policy or pattern theories. But the
single-incident theory requires that the need for training be
obvious without consideration of prior violations. See
Connick, 563 U.S. at 63 
n.7 (rejecting contemporaneous or sub-
sequent evidence to prove notice). When evaluating “the risk
from a particular glaring omission in a training regimen” the
question is whether the risk is “obvious in the abstract.” Bryan
County, 520 U.S. at 410 
(internal quotation marks omitted).


8 Also unexplained is how notice of an “obvious risk” could be “incremen-
tal.” Incremental denotes gradual, or progressive, the opposite of a “so
obvious” need necessary for deliberate indifference under a “single inci-
dent” theory. 
Canton, 489 U.S. at 390 
n.10.
Nos. 18-1498, et al.                                          75

That is for good reason. Single incident liability exists for a
limited circumstance in which a municipality faced no prior
instances of similar constitutional violations. See 
Connick, 563
U.S. at 64 
(explaining that with the single-incident exception
“[t]he Court [in Canton] sought not to foreclose the possibility,
however rare, … that a city could be liable under § 1983 with-
out proof of a preexisting pattern of violations”). The single-
incident theory is a way to show liability, without evidence of
past constitutional violations. Here, the district court
concluded that those same categories of evidence failed to
provide notice under a pattern theory, and an implied policy
theory of liability is not before us. So the four categories can-
not be considered here.
    There is only one way out of this contradiction. If notice is
no longer required, what remains is liability without notice:
respondeat superior. Monell and Supreme Court precedents for-
bid that result. Because the “so obvious” standard from the
Canton footnote 10 hypothetical cannot be satisfied here, the
plaintiffs’ constitutional claim is legally deficient.
   B. Failure to Supplement Policies
    The second basis on which the majority opinion concludes
liability—that a failure to supplement County policies
amounted to unconstitutional inaction under Monell—also
relies on Canton’s single-incident theory. For a number of rea-
sons, that approach cannot be squared with the Court’s ad-
monitions in Canton, Bryan County, and Connick either.
   Canton’s single-incident hypothetical expressly considers
only “the need to train” 
officers. 489 U.S. at 390 
n.10. The
Court has never extended single-incident liability outside fail-
76                                           Nos. 18-1498, et al.

ure to train. See 
Connick, 563 U.S. at 64 
(limiting single-inci-
dent theory to “the obvious need for specific … training”);
Bryan 
County, 520 U.S. at 410 
(limiting single-incident theory
to “a particular glaring omission in a training regimen”). The
claim at issue involves something different—purported gaps
in the County’s sexual assault express policies—not failure to
train.
    As just noted, the majority opinion’s unconstitutional in-
action holding relies exclusively on the same four categories
of evidence that failed to provide “obvious” notice of a need
for more action. Cf. Con
nick, 563 U.S. at 61
, 63, 71 (requiring
“obvious” notice to establish single-incident liability); Bryan
County, 520 U.S. at 410 
(same); 
Canton, 489 U.S. at 390 
n.10 (re-
quiring “so obvious” and “plainly obvious” notice). So ap-
pears the contradiction: the same evidence the district court
found did not provide notice, a finding left unchallenged on
appeal, gives notice in the majority opinion of a need for more
action. The need for additional policies is simultaneously “not
at all obvious” and “so obvious.”
    The majority opinion concludes that the County’s zero-tol-
erance sexual assault policies contained “material gaps”
because those policies did not include information about pre-
venting and detecting sexual assaults. Majority op. at pp.
18-19. To prove deliberate indifference, plaintiffs must show
it was “highly predictable” that, absent supplements to its
zero-tolerance assault policies, male Polk County Jail guards
would inflict the particular constitutional injuries suffered by
plaintiffs. Bryan 
Cty., 520 U.S. at 411
. This means plaintiffs
must show that those guards lacked the tools to know that
sexual assault is a crime and violated County policies. See
 id.
But as 
explained above, there were no prior instances of rape
Nos. 18-1498, et al.                                                      77

at the jail and plaintiffs’ own opinion witness agreed the jail
had a “good” record on sexual assaults. As with all of its
guards, the County gave Christensen the tools to stop himself;
he chose not to.
   The majority opinion instead grounds fault solely on three
generalized risks:
       That “[t]he confinement setting is a tinderbox
        for sexual abuse.”
       That any “guard would grow too comfortable,
        lose his better angels, and step over the clear
        line marked in Polk County’s written policies.”
       That it was “’highly predictable,’ if not certain,
        that a male guard would sexually assault a fe-
        male inmate if the County did not act.”
Majority op. at pp. 24–26 (emphases added).9 Absent from
these generalizations is any mention of how purported gaps
in County policies amounted to deliberate indifference. Left
unexplained is specifically how the decision to rely on policies
that, until Christensen, produced “good” results on the par-
ticular risk of sexual assaults amounted to deliberate indiffer-
ence. It cannot be said a situation that has never occurred is
likely to recur. See Bryan 
County, 520 U.S. at 409 
(basing pre-
dictability on “[t]he likelihood that the situation will recur“).
   The deliberate indifference standard in Canton has an im-
portant corollary. Under the single incident theory, Canton
requires employers “know to a moral certainty” that their em-
ployees will face a “difficult choice of the sort that training or

9 If such a generalized risk suffices to trigger notice here, the same gener-
alized risk could apply for jail fights, medical attention, and other aspects
of confinement.
78                                            Nos. 18-1498, et al.

supervision will make less difficult.” Walker v. City of New
York, 
974 F.2d 293
, 297 (2d Cir. 1992) (citing 
Canton, 489 U.S.
at 390 
n.10). Case law has referred to this as the “difficult de-
cision” or “difficult choice” requirement.
    However phrased, it is not a “difficult choice” or a “diffi-
cult decision” for a guard not to rape an inmate. Such a deci-
sion is mandated by the law, written policies and training
here, as well as any moral code. Just so, Christensen had no
“difficult choice.” He was instructed by the written policies,
training, and the law not to sexually assault, but he willfully
and surreptitiously ignored that training and instruction. For
all these reasons, this case does not fit within the narrow and
rare single-incident exception to the pattern requirement for
municipal liability.
     C. Causation
    To recover from the County under § 1983, plaintiffs also
must prove “a direct causal link between the municipal action
and the deprivation of federal rights.” Bryan 
County, 520 U.S.
at 404
. A municipality cannot be liable “unless deliberate action
attributable to the municipality directly caused a deprivation
of federal rights.”
 
Id. at 415. 
Even if training or a policy omis-
sion was in some manner deficient, “the identified defi-
ciency … must be closely related to the ultimate injury” such
that “the deficiency … actually caused the police officers’ in-
difference.” 
Canton, 489 U.S. at 391
. Said slightly differently,
“a municipality can be liable under § 1983 only where its pol-
icies are the moving force behind the constitutional viola-
tion.”
 Id. at 389 
(citations, internal brackets, and quotation
marks omitted). These causation requirements are “stringent”
and “rigorous” to prevent expansions of constitutional tort li-
ability and the use of federal courts to restructure municipal
Nos. 18-1498, et al.                                          79

institutions. See Bryan 
County, 520 U.S. at 415 
(holding a lesser
causation standard embroils “serious federalism concerns”
and permits “municipal liability [to] collapse[] into respondeat
superior liability.”).
    This case presents a glaring causation problem: When
Christensen assaulted the plaintiffs, he knowingly and will-
fully acted in defiance of his training and County policies. No
evidence shows that Christensen decided to assault plaintiffs
for any reason related to inadequate training or policies. The
majority opinion cites no “affirmative link” between the oc-
currence of Christensen’s crimes and the County’s alleged in-
action. 
Tuttle, 471 U.S. at 823 
(plurality opinion) (“At the very
least there must be an affirmative link between the policy and
the particular constitutional violation alleged.”); 
Rizzo, 423
U.S. at 371 
(rejecting liability because “there was no affirma-
tive link between the occurrence of the various incidents of
police misconduct and the adoption of any plan or policy”).
Without an affirmative link, the causation requirement is not
met.
    The majority opinion rests its conclusion of causation on
the obviousness of a risk: “’The high degree of predictability’
that constitutes notice … ‘may also support an inference of
causation …’” Majority op. at p. 30 (citing Bryan 
County, 520
U.S. at 409
–10). Because the risk of constitutional injury was
obvious, the majority opinion decides, only a “small inferen-
tial step” therefrom is necessary to find causation. Majority
op. at p. 30. But there must be a foundation from which to take
that step. That is missing here.
   First, Connick emphasizes the notice a municipality’s poli-
cymakers must have of the omissions in their training
program causing municipal employees to violate a citizen’s
80                                            Nos. 18-1498, et al.

constitutional rights before the municipality may be deemed
deliberately 
indifferent. 563 U.S. at 61
–62. “Without notice
that a course of training is deficient in a particular respect, de-
cisionmakers can hardly be said to have deliberately chosen a
training program that will cause violations of constitutional
rights.”
 
Id. at 62
.
    But the district court ruled, and it is unchallenged on
appeal, that the misconduct evidence did not constitute a pat-
tern of constitutional violations. The lack of a pattern de-
prived the County of notice of a need for additional training
and policies. Nevertheless, the majority opinion relies on that
same evidence to conclude obviousness and notice. But it can-
not be both ways; notice of a risk cannot simultaneously be
not present and present. Respect for the district court’s find-
ing on an issue not appealed should control, and not form the
foundation for a conclusion of causation.
    Second is the type of causation found. Certainly, the ab-
sence of supplemental training and policies may increase the
likelihood that a guard might assault an inmate. The inquiry
here, though, is whether male Polk County Jail guards posed
a risk to female inmates such that reliance on the County’s
existing policies and training would result in guard-on-in-
mate rape. Bryan 
County, 520 U.S. at 411
. The district court did
not perform this “legal inquiry,”
 id., for failure 
to train or un-
constitutional inaction liability, and what remains is just a
conclusion of potential causation.
    But liability under § 1983 requires actual causation of an
injury. Bryan 
County, 520 U.S. at 404
–05; 
Canton, 489 U.S. at
389
; 
Monell, 436 U.S. at 692 
(holding the “mere right to control
without any control or direction having been exercised” will
not support § 1983 liability (citing 
Rizzo, 423 U.S. at 370
–71));
Nos. 18-1498, et al.                                           81

Rizzo, 423 U.S. at 372
, 375–76, 378 (distinguishing the mere
failure to act from “active conduct”). “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.” 
Connick,
563 U.S. at 68 
(quoting 
Canton, 489 U.S. at 391 
(internal quota-
tion marks and brackets omitted)). The question is not
whether the inaction potentially caused plaintiffs’ injuries; it
is whether the inaction “actually” (
Canton, 489 U.S. at 391
) and
“directly” (Bryan 
County, 520 U.S. at 415
) caused the injuries.
Here, no evidence shows that County inaction actually and
directly caused plaintiffs’ injuries.
     Third, the majority opinion permits a finding of liability in
jail and prison settings based not on proof of causation, but a
presumption of negligence. Male jail guards have exclusive
“authority and control” over “nearly everything in [inmates’]
lives.” Majority op. at p. 24. “With this authority and con-
trol … c[o]me[s] power and, in turn, access opportunity to
abuse it.”
 Id. Under this 
“delicate setting,” “the risk of consti-
tutional injury” “was as obvious as obvious could be.”
 
Id. at
pp. 
23, 28. “The jury knew that from common sense[,]” so “it
took but a small inferential step for the jury to find causation.”

Id. at 28.
    Far from Monell’s “rigorous” causation requirements, the
majority opinion’s reasoning echoes res ipsa loquitur, presum-
ing negligence from Christensen’s crimes. See 
Tuttle, 471 U.S.
at 831 
n.6 (Brennan, J., concurring) (admonishing application
of res ipsa loquitur doctrine in municipal liability claims). Res
ipsa loquitur applies when “the facts of the occurrence warrant
the inference of negligence” and “the defendant [has] exclu-
sive control of all the things used in an operation which might
82                                           Nos. 18-1498, et al.

probably have caused injury.” Jesionowski v. Bos. & M. R. R., 
329
U.S. 452
, 456 (1947) (emphases added) (citation and quotation
marks omitted). Compare the elements of res ipsa loquitur with
the majority opinion’s reasoning: The jail’s “authority and
control” through “male guards” over “nearly everything in
[inmates’] lives,” supports an inference of causation. Majority
op. at p. 24. Instead of causation or deliberate indifference,
negligence is presumed. Monell requires more: deliberate in-
difference and moving force causation, not “simple or even
heightened negligence.” Bryan 
County, 520 U.S. at 407
.
    Plaintiffs’ appalling injuries were not caused by a lack of
specific training and policy language about sexual assault
prevention and detection. They were caused by a miscreant
guard’s hidden, willful, and criminal defiance. There is no ev-
idence that Christensen made the decision to assault plaintiffs
for any reason related to inadequate training or policies. For
example, no evidence shows that Christensen calculatedly ex-
ploited training and policy gaps. Nor does any evidence show
that such gaps emboldened, let alone caused, Christensen to
commit rapes. The record shows that when Christensen as-
saulted plaintiffs he knew he was acting contrary to his train-
ing and in violation of County policies. From this undisputed
evidence, any reasonable fact finder would have to conclude
that Christensen’s bad-faith conduct, in conflicting with his
employer’s policies and training, caused plaintiffs’ injuries.
Christensen, not the County, was the “moving force” that
caused plaintiffs’ injuries. 
Monell, 436 U.S. at 691
–92.
Nos. 18-1498, et al.                                           83

IV. Federalism
   The majority opinion encroaches on the federalism princi-
ples that the Supreme Court’s Monell cases hold so essential.
    Monell's policy requirement was intended to balance
§ 1983’s remedial purpose with the principle that the power
exercised by federal courts must not be so broad as to upend
the delicate balance of powers among federal, state, and local
governments. See Michael J. Gerhardt, The Monell Legacy: Bal-
ancing Federalism Concerns and Municipal Accountability under
Section 1983, 62 S. Cal. L. Rev. 539 (1989). The majority opinion
upsets that balance. It extends federal civil liability by
allowing a federal statute to dictate municipal liability, imper-
missibly infringing on state and local independence and self-
governance. It also defines notice of constitutional violations
so broadly that liability may now arise in any confinement
setting, without limit.
    The Supreme Court has emphasized the important
federalism interests implicated by municipal liability cases in-
volving failure to train, like this one. Allowing § 1983 cases to
advance against municipalities for failure to train or inaction
“on a lesser standard of fault would result in de facto respondeat
superior liability on municipalities—a result [] rejected in
Monell.” 
Canton, 489 U.S. at 392 
(citation omitted). “This is an
exercise … federal courts are ill suited to undertake, as well
as one that would implicate serious questions of federalism.”
Canton, 489 U.S. at 392 
(citing 
Rizzo, 423 U.S. at 378
–80).
   The Court sounded the same warning in Bryan County:
       Where a court fails to adhere to rigorous re-
       quirements of culpability and causation, munic-
84                                             Nos. 18-1498, et al.

       ipal liability collapses into respondeat superior li-
       ability. As [] recognized in Monell and [] repeat-
       edly reaffirmed [since], Congress did not intend
       municipalities to be held liable unless deliberate
       action attributable to the municipality directly
       caused a deprivation of federal rights. A failure
       to apply stringent culpability and causation re-
       quirements raises serious federalism concerns,
       in that it risks constitutionalizing … require-
       ments that States have themselves elected not to
       
impose.
520 U.S. at 415 
(citing 
Canton, 489 U.S. at 392
).
    Connick raised the same cautionary flag: “[Section 1983]
does not provide plaintiffs or courts carte blanche to mi-
cromanage local governments throughout the United 
States.”
536 U.S. at 68
. While Justice Scalia joined the Connick opinion
in full, his concurrence explained that the dissent’s “theory of
deliberate indifference would repeal the law of 
Monell.” 536
U.S. at 73 
(Scalia, J., concurring):
       [W]e do not have de facto respondeat superior lia-
       bility, for each such violation under the rubric
       of failure to train simply because the municipal-
       ity does not have a professional educational
       program covering the specific violation in suffi-
       cient depth. …
       These restrictions are indispensable because
       without them, failure to train would become a
       talismanic incantation producing municipal lia-
       bility in virtually every instance where a person
       has had his or her constitutional rights violated
Nos. 18-1498, et al.                                                        85

        by a city employee—which is what Monell re-
        jects. Worse, it would engage the federal courts
        in an endless exercise of second-guessing mu-
        nicipal employee-training programs, thereby
        diminishing the autonomy of state and local
        governments.
Id at 73–74 (quotes, citations, and footnotes omitted).
    This case presents the infringement warned of in Canton,
Bryan County, and Connick by allowing a federal statute to dic-
tate a municipality’s liability without proof of fault or causa-
tion.
     The majority opinion declares the County was deficient in
preventing and detecting sexual assaults. This standard is
drawn directly from the trial testimony of plaintiff’s opinion
witness, Eiser, that the jail had inadequately addressed the
prevention and detection of sexual assaults of inmates. Eiser
testified PREA “set[s] the standard for jails very clearly.” Ac-
cording to Eiser, the PREA standard has three components:
prevention, detection, and training. The prevention and de-
tection components are lifted directly from the text of PREA.10
The changes he recommended to allow for easier reporting of
sexual assaults—including the lockbox for inmate complaints
mentioned in the majority opinion—are all teachings of
PREA. Eiser further testified the jail had addressed only one
of the PREA components in its policy, and what should occur
if a facility cannot meet all the PREA components.



10At 34 U.S.C. § 30302 PREA provides: “The purposes of this chapter are
to—(6) increase the accountability of prison officials who fail to detect, pre-
vent, reduce, and punish prison rape.” (emphasis added).
86                                          Nos. 18-1498, et al.

    Per the majority opinion “PREA is not a constitutional
standard, and jails are not required to adopt it.” Majority op.
at p. 29. Yet, it concludes that the county’s liability rests on
Eiser’s testimony that the municipality failed to prevent and
detect the sexual assaults by a guard of two inmates. Strik-
ingly, an opinion witness who advanced a standard the ma-
jority opinion says is not required now sets the fault standard
in this circuit, even though that witness agreed the County
had a “good” record on sexual assaults.
    What type of action must municipalities take to satisfy this
new standard and avoid liability? The majority opinion says
“reasonable steps to reduce the obvious and known risks of
assaults on inmates.”
 
Id. at p. 
29. But what those reasonable
steps should be requires notice, which the County did not
have. Recall, the district court concluded that the four catego-
ries of evidence, including the Jorgenson incident, did not col-
lectively make a risk of rape known to the County. Per the
majority opinion, the only direction given to courts and mu-
nicipal policymakers on those reasonable steps are the detec-
tion and protection measures of PREA. If Wisconsin munici-
pal jails should be held to the PREA standard, that decision
should be made by the people of Wisconsin through their leg-
islature, or by the Wisconsin Department of Corrections
through administrative rulemaking. Instead, PREA’s adop-
tion effectively comes by way of this court.
    This case presents another federalism concern: One of the
mechanisms by which the majority opinion concludes the jail
was on notice of policy and training gaps applies to all con-
finement settings. As the Court explained in Bryan County, es-
tablishing that a municipality acted with “deliberate indiffer-
ence … require[es] proof that a municipal actor disregarded a
Nos. 18-1498, et al.                                            87

known or obvious consequence of his action.”
 
Id. at 410 
(in-
ternal quotation marks omitted). The Court specifically re-
jected fault based on generalized risk.
 Id. But the 
majority
opinion permits fault based on such a generalized risk by de-
fining “obvious” so expansively as to render the term effec-
tively without meaning.
    Because the majority finds the risk of sexual assault is “ob-
vious” in all confinement settings involving male guards and
female inmates, Majority op. at p. 24, and inaction in the face
of that “obvious” risk establishes deliberate indifference, any
municipality in the Seventh Circuit operating a confinement
setting with male guards and female inmates may now be
subject to liability and federal oversight if it does not enact
new policies or training. This formulation, founded on an in-
curable standard of notice, is unworkable. Regardless of what
steps a municipality takes, it is on notice of the risk of sexual
assaults so long as male guards supervise female inmates in a
confinement setting. And even if a municipality enacts poli-
cies to prevent sexual assaults and trains employees on those
policies (like Polk County), it is deliberately indifferent unless
it enacts additional policies and training programs.
    This formulation also gives no guidance to district courts
or municipal policymakers as to what training is or is not re-
quired to avoid a constitutional deprivation. This creates lia-
bility potentially unlimited in scope; if it has limits, those con-
tours are ill-defined, and certain to confound and divide
courts and policymakers. And in creating potentially unlim-
ited liability, this court disregards each of the warnings and
cautions on this topic passed to us from the Supreme Court
and repeated above.
88                                              Nos. 18-1498, et al.

    For these reasons, our law interpreting § 1983 should fol-
low the Court’s lead and avoid intrusion into an area of tradi-
tional state authority such as this. See, e.g., Fla. Dep’t of Revenue
v. Piccadilly Cafeterias, Inc., 
554 U.S. 33
, 50 (2008) (holding the
“federalism canon … obliges us to construe [a statute]’s ex-
emption narrowly”); BFP v. Resolution Trust Corp., 
511 U.S.
531
, 544 (1994) (refusing to interpret a statute “[t]o displace
traditional state regulation” unless Congress’s intent to do so
was clear from the statute (citing Gregory v. Ashcroft, 
501 U.S.
452
, 460–61 (1991)); 
Gregory, 501 U.S. at 468 
(“By its terms, the
Fourteenth Amendment contemplates interference with state
authority: ‘No State shall … deny to any person within its ju-
risdiction the equal protection of the laws.’ … But this Court
has never held that the Amendment may be applied in com-
plete disregard for a State’s constitutional powers.” (quoting
U.S. Const. amend. XIV § 1)).
    The majority opinion heralds federal involvement in what
historically has been an area of state and local decision-mak-
ing. Subsidiarity should be respected, especially with the Su-
preme Court repeatedly and specifically noting important
federalism interests in precisely this type of case.
V. Rule 50 and the Jury Verdict
    Courts are the gatekeepers for the questions a jury may
properly answer. In permitting a jury to determine whether a
municipality is liable under § 1983, courts are constrained by
the Supreme Court cases relayed above in II., the parameters
of failure to train liability as explained in III., and the federal-
ism principles noted in IV.
   Federal Rule of Civil Procedure 50 sets the standard and
procedure courts are to employ in this role. “Judgment as a
Nos. 18-1498, et al.                                           89

matter of law is proper ‘if a reasonable jury would not have a
legally sufficient evidentiary basis to find for the party on that
issue.’” Lawson v. Sun Microsystems, Inc., 
791 F.3d 754
, 761 (7th
Cir. 2015) (quoting FED. R. CIV. P. 50(a)(1)).
    This court reviews de novo a district court’s ruling on a
post-trial motion for judgment as a matter of law under Rule
50. Andy Mohr Truck Center, Inc. v. Volvo Trucks N.A., 
869 F.3d
598
, 602 (7th Cir. 2017) (citation omitted). The evidence is
viewed in the light most favorable to the verdict. Empress
Casino Joliet Corp. v. Balmoral Racing Club, Inc., 
831 F.3d 815
,
822 (7th Cir. 2016). This standard applies regardless of
whether the verdict under review was for the plaintiff or the
defendant and regardless of the case’s underlying legal is-
sues. Schandelmeier-Bartels v. Chicago Park Dist., 
634 F.3d 372
,
376 (7th Cir. 2011).
    The showing required for judgment as a matter of law un-
der Rule 50 has been equated with a grant of summary judg-
ment under Federal Rule of Civil Procedure 56. The Advisory
Committee Note to the 1991 amendments to Rule 50 explains
that, in part, those amendments are intended to call attention
to the similarity between judgment as a matter of law under
Rule 50 and summary judgment under Rule 56. The Supreme
Court has also “noted that the ‘genuine issue’ summary judg-
ment standard is ‘very close’ to the ‘reasonable jury’ directed
verdict standard … .” Anderson v. Liberty Lobby, Inc., 
477 U.S.
242
, 251 (1986); see also Miller v. Fisher, 381 F. App’x 594, 596
(7th Cir. 2010) (“The standard [for reviewing a Rule 50 mo-
tion] mirrors our appellate analysis of a motion for summary
judgment.”).
   When considering whether a claim survives a motion for
judgment as a matter of law, courts decide whether the claim
90                                                    Nos. 18-1498, et al.

is legally sufficient. See FED. R. CIV. P. 50(a)(1) (stating a judge
may grant judgment as a matter of law if the movant shows
“a reasonable jury would not have a legally sufficient eviden-
tiary basis to find for the party on that issue”). This question
differs from the factual finding asked of and made by a jury.
District courts must determine the controlling law and exer-
cise “responsibility to assure the fidelity of [their] judgment
to the controlling law, a responsibility imposed by the Due
Process Clause of the Fifth Amendment.” FED. R. CIV. P. 50
advisory committee’s note to 1991 amendment (citation omit-
ted). That note provides good guidance: “In ruling on such a
[Rule 50] motion, the court should disregard any jury deter-
mination for which there is no legally sufficient evidentiary
basis enabling a reasonable jury to make it.”
 Id.
    The constitutional 
claim against the County never should
have reached the jury. Summary judgment should have been
granted to the County because the claim was legally deficient,
as shown above.11 The legal deficiency of the claim, more than
the lack of evidence, should have led to its demise. If sum-
mary judgment had not been granted on the claim, judgment
as a matter of law should have been granted to the County.
    If this court had reversed the district court’s denial of the
County’s motion for judgment as a matter of law, it certainly
would not have “taken a case from the jury.” Rather, it would
have exercised its proper role by deciding that the jury never
should have considered the claim against the County in the
first place. Putting to the side the understandable sympathy


11The County moved for summary judgment on plaintiffs’ § 1983 claims
on this issue (Dist. Ct. D.E. 54), which the district court denied. (Dist. Ct.
D.E. 160 at 30).
Nos. 18-1498, et al.                                                     91

factor—the desire to make someone pay for the assaults these
plaintiffs endured—the County should not be held liable on
plaintiffs’ legally deficient constitutional claim. Liability
standards exist to govern behavior. Every tort case has a
standard for fault and for causation. The court’s role is to pre-
clude a finding of liability if there is no legal basis and no
causal link to hold the defendant at fault. Otherwise courts
fail to ensure that the law is properly applied. A jury’s verdict
must be respected, but the jury can only be posed a question
the law allows. This jury was asked a question it should not
have been.
    This case presents the downside of the mantra “let a ques-
tionable claim go to the jury.” A claim that should have been
denied at summary judgment, or resolved after the evidence
was presented, or even post-trial, took on inertia. Rule 50 ex-
ists to prevent that. The Supreme Court and federal appellate
courts consistently overturn verdicts that rest on incorrect le-
gal grounds. As noted above, the Court’s failure-to-train de-
cisions that control here each reversed a jury verdict for the
plaintiff: 
Tuttle, 415 U.S. at 813
; 
Canton, 489 U.S. at 391
; Bryan
Cty., 520 U.S. at 400
; and 
Connick, 563 U.S. at 57
.
   In addition to these reversals, in the last twenty years this
court has overturned jury verdicts, for plaintiffs and for de-
fendants, in at least twelve cases when a prisoner inmate sued
under § 1983.12 In about the same time frame this court has


12Martin v. Milwaukee Cty., 
904 F.3d 544 
(7th Cir. 2018) (reversed jury ver-
dict for plaintiff on a prison sexual assault claim because rape not within
scope of employment); Ruiz-Cortez v. City of Chicago, 
931 F.3d 592 
(7th Cir.
2019) (vacated jury verdict for defendant on a Brady violation claim, or-
dered a new trial on that claim, and affirmed summary judgment for town
92                                                      Nos. 18-1498, et al.

twice overturned a jury’s finding of an unconstitutional pol-
icy in Monell cases. See Thomas v. Cook Cty. Sheriff’s Dept., 
604
F.3d 293
, 298 (7th Cir. 2010) (concluding insufficient evidence
existed to hold sheriff liable because causal connection be-
tween his policies and death was tenuous in light of jury’s
finding that individual correctional officers deliberately dis-
regarded plaintiff’s medical needs); Palmquist v. Selvik, 
111
F.3d 1332
, 1347 (7th Cir. 1997) (holding municipality was not
deliberately indifferent for failure to train because the lack of


on Monell claim); Nelson v. City of Chicago, 
810 F.3d 1061 
(7th Cir. 2016)
(reversed jury verdict for plaintiff on an unlawful search and seizure claim
and ordered a new trial); Barber v. City of Chicago, 
725 F.3d 702 
(7th Cir.
2013) (reversed jury verdict for defendant on claims of arrest without
probable cause and excessive force and ordered a new trial); Phillips v.
Cmty. Ins. Corp., 
678 F.3d 513 
(7th Cir. 2012) (reversed jury verdict for de-
fendants on a claim of excessive force and remanded to enter judgment
for plaintiff); Duran v. Town of Cicero, 
653 F.3d 632 
(7th Cir. 2011) (reversed
denial of defendant’s new trial motion on claims of excessive force, false
arrest, and equal protection violations against police officers and town);
Fox v. Hayes, 
600 F.3d 819 
(7th Cir. 2010) (reversed jury verdict for plaintiffs
due process claim after plaintiffs won on false arrest, malicious prosecu-
tion, violation of due process, and state law claims because plaintiff’s state
law theory was the same as violation of due process claim); Watkins v.
Kasper, 
599 F.3d 791 
(7th Cir. 2010) (reversed jury verdict for incarcerated
plaintiff’s exercise of free speech claim because inmate’s speech was in-
consistent with legitimate penological interests); Thomas v. Cook Cty. Sher-
iff’s Dep't, 
604 F.3d 293 
(7th Cir. 2009) (reversed jury verdict for plaintiff’s
wrongful death action against sheriff under Monell and remanded to enter
judgment in sheriff’s favor); Campbell v. Miller, 
499 F.3d 711 
(7th Cir. 2007)
(reversed jury verdict for defendant officer because no jury could find his
public strip search of the plaintiff in a driveway was reasonable); Riccardo
v. Rausch, 
375 F.3d 521 
(7th Cir. 2004) (reversed jury verdict for incarcer-
ated plaintiff on a failure to protect claim); Campbell v. Peters, 
256 F.3d 695
(7th Cir. 2001) (reversed jury verdict for incarcerated prisoner on qualified
immunity grounds).
Nos. 18-1498, et al.                                                       93

training was not sufficiently obvious and any inadequacy was
not sufficiently likely to result in a constitutional violation).
Because the constitutional claim against the County had no
legal basis, judgment as a matter of law should have been
granted.
VI. Other Case Law
    The majority opinion expands municipal liability under
§ 1983 beyond the boundaries established by federal appellate
courts, including that of this circuit:
        No federal appellate court has ever extended
         the single-incident exception to the sexual as-
         sault context;
        No federal appellate court has ever extended
         the single-incident exception when the em-
         ployee’s compliance with the municipality’s
         policy and training would have prevented the
         injuries; and
        Specialized training is not required to know that
         rape is wrong.
     A. Single-Incident Liability for Sexual Assaults
    The other federal appellate courts to have ruled on this is-
sue have rejected the “so obvious” single-incident exception
in the sexual assault context:13




13 The majority opinion is also at odds with this court’s order in Johnson v.
Cook Cty., 526 F. App’x 692 (7th Cir. 2013), in which this court rejected sin-
gle-incident liability and affirmed the dismissal of an inmate’s Monell
claim after the inmate was sexually assaulted by a Cook County Jail em-
ployee.
 Id. at 697.
94                                            Nos. 18-1498, et al.

    Eighth Circuit: Parrish v. Ball, 
594 F.3d 993
, 999 (8th Cir.
2010) (rejecting plaintiff’s single-incident theory because
“tell[ing] officers to not sexually assault detainees [] is not so
obvious that not doing so would result in an officer actually
sexually assaulting a female detainee"); S.J. v. Kansas City
Missouri Pub. Sch. Dist., 
294 F.3d 1025
, 1029 (8th Cir. 2002) (re-
jecting plaintiff’s single-incident claim after finding there is
no patently obvious need for public schools or principals to
train volunteers not to commit sexual abuse); Andrews v.
Fowler, 
98 F.3d 1069
, 1077 (8th Cir. 1996) (rejecting plaintiff’s
single-incident theory after determining there is no “patently
obvious need for the city to specifically train officers not to
rape young women”).
    Ninth Circuit: Flores v. Cty. of Los Angeles, 
758 F.3d 1154
,
1160 (9th Cir. 2014) (rejecting plaintiff’s single-incident theory
of liability because “[t]here is [] every reason to assume that
police academy applicants are familiar with the criminal pro-
hibition on sexual assault, as everyone is presumed to know
the law”).
    Tenth Circuit: Schneider v. City of Grand Junction Police
Dep’t, 
717 F.3d 760
, 774 (10th Cir. 2013) (rejecting plaintiff’s
single-incident claim because “[s]pecific or extensive training
hardly seems necessary for a jailer to know that sexually as-
saulting inmates is inappropriate behavior’’ (quoting Barney
v. Pulsipher, 
143 F.3d 1299
, 1308 (10th Cir. 1998)); Barney v.
Pulsipher, 143 F.3d at 1308 
(“[T]his case does not fall within
the narrow range of circumstances justifying a finding of
deliberate indifference absent a pattern of violations … [be-
cause] we are not persuaded that a plainly obvious conse-
quence of a deficient training program would be the sexual
assault of inmates.”).
Nos. 18-1498, et al.                                            95

    Eleventh Circuit: Floyd v. Waiters, 
133 F.3d 786
, 796 (11th
Cir. 1998), vacated on other grounds by 
525 U.S. 802 
(1998), rein-
stated by 
171 F.3d 1264 
(11th Cir. 1999) (board of education
was “entitled to rely on the common sense of its” security
guards not to sexually harass and rape underage girls.).
    The only federal appellate case arguably to the contrary
may be Cash v. Cty. of Erie, 
654 F.3d 324
, 337–38 (2d Cir. 2011).
But Cash does not apply a single-incident theory analysis or
even cite to Canton’s single-incident hypothetical. Cash itself
states it is not a failure-to-train case: “the deliberate indiffer-
ence concern in this case … is not with a failure to train prison
guards to distinguish between permissible and impermissible
sexual contact with prisoners. Nor is it with providing suffi-
cient supervision to ensure that guards make correct choices
in this respect.”
 Id. at 336. 
Cash also based its conclusion of
liability on a generalized risk rather than a particularized in-
quiry (violating Bryan 
County, 520 U.S. at 410
–13), and it ig-
nored the requirement of similar prior violations (violating
Connick, 563 U.S. at 62
–63).
    A failure to supervise case, Cash holds: “[K]nowledge that
an established practice has proved insufficient to deter lesser
[sexual] misconduct can be found to serve notice that the
practice is also insufficient to deter more egregious miscon-
duct.” 654 F.3d at 337
. See Majority op. at p. 26 (quoting same).
But that holding offers no help. Here, unlike Cash, the district
court ruled that plaintiffs’ pattern evidence of lesser miscon-
duct failed to show notice of an obvious need for revised pol-
icies and training. Plaintiffs did not appeal that ruling. And
unlike the majority opinion here, Cash does not deem it obvi-
ous that without training male guards will rape female
96                                                     Nos. 18-1498, et al.

inmates. To the extent the majority opinion may read Cash dif-
ferently, district courts in the Second Circuit have not fol-
lowed suit.14
     B. Failure to Comply with Policies and Training
    A federal appellate court also has never extended the sin-
gle-incident exception when the employee’s compliance with
the municipality’s policy and training would have prevented
the injuries:
   Fifth Circuit: Pena v. City of Rio Grande City, 
879 F.3d 613
,
624 (5th Cir. 2018) (determining single incident theory should



14 Recall, to prove deliberate indifference under the single-incident theory

of liability, Canton requires employers “know to a moral certainty” that
their employees will face a “difficult choice of the sort that training or su-
pervision will make less difficult.” Walker v. City of New York, 
974 F.2d 293
,
297 (2d Cir. 1992). District courts in the Second Circuit after Cash have not
concluded that sexual assault qualifies as a “difficult decision.” See, e.g.,
R.A. v. City of New York, 
206 F. Supp. 3d 799
, 803 (E.D.N.Y. 2016) (“The
complaints against Defendant … show a conscious decision by Defend-
ant … to commit sexual assault, which does not present a ‘difficult choice’
[that] further training would prevent.” (quoting 
Walker, 974 F.2d at 297
(citations omitted)); Noonan v. City of New York, 
2015 WL 3948836
, at *4
(S.D.N.Y. June 26, 2015) (decision to commit a sexual assault cannot rea-
sonably be seen as posing the type of “difficult choice” contemplated by
the Second Circuit in Walker.); Doe v. City of New York, 
2013 WL 796014
, at
*4 (S.D.N.Y. Mar. 4, 2013) (“Plaintiff has not plausibly alleged that either
[officer]’s decision to rape her or [second officer]’s decision to aid and abet
that rape was a ‘difficult choice of the sort that training or supervision will
make less difficult.’” (quoting 
Walker, 974 F.2d at 298
)); Castilla v. City of
New York, 
2012 WL 5510910
, at *6 (S.D.N.Y. Nov. 14, 2012) (“It beggars
common sense to posit that [the officer] faced a difficult choice as to
whether or not to coerce sex from [plaintiff] and that training would have
alleviated that conundrum.”).
Nos. 18-1498, et al.                                           97

be limited to “cases in which the government actor was pro-
vided no training whatsoever”); Estate of Davis ex rel. McCully
v. City of N. Richland Hills, 
406 F.3d 375
, 377–78, 385–86 (5th
Cir. 2005) (concluding single-incident exception did not apply
when training occurred).
    Ninth Circuit: Jason v. Tanner, 
938 F.3d 191
, 199 (5th Cir.
2019) (rejecting single-incident theory because prison had
trained about sling-blade misuse, “albeit training that didn’t
prevent th[e] attack,” and “this was the first and only sling-
blade attack in a presumably otherwise incident-free pro-
gram”).
    Tenth Circuit: Waller v. City & Cty. of Denver, 
932 F.3d 1277
,
1288 (10th Cir. 2019) (rejecting single-incident theory because
“[t]his case does not involve technical knowledge or ambigu-
ous ‘gray areas’ in the law that would make it ‘highly predict-
able’ that a deputy sheriff … would need ‘additional specified
training’ to … put [him] on notice that [he] may not violently
assault a restrained detainee”); Keith v. Koerner, 
843 F.3d 833
,
839–40 (10th Cir. 2016) (rejecting failure-to-train liability be-
cause jailer received training about the impropriety of and po-
tential consequences for engaging in sexual misconduct);
Schneider, 717 F.3d at 774 
(10th Cir. 2013) (rejecting failure-to-
train liability where the offending officer “was, in fact, in-
structed against relationships with women he met on
duty. …”); Porro v. Barnes, 
624 F.3d 1322
, 1328–29 (10th Cir.
2010) (county jail’s policy of training jailers to use stun guns
and failure to enforce federal policy that banned use of stun
guns on immigration detainees did not demonstrate deliber-
ate indifference).
   On first glance, the only federal appellate court case on the
other side of this question might be Allen v. Muskogee, 
119 F.3d
98 
                                          Nos. 18-1498, et al.

837 (10th Cir. 1997), in which a divided panel reversed sum-
mary judgment granted to a city on an excessive force claim
and held that failure to train officers how to deal with armed
and mentally ill persons fits within the Canton single-incident
exception.
 Id. at 844. 
But in Allen the officer was trained to do
the wrong thing, not left untrained.
 Id. Unlike Allen
, no one
disputes that had Christensen followed his training or the
jail’s policies, the injuries to plaintiffs never would have oc-
curred.
    These cases instruct that single-incident liability should
not be extended to cases involving a rogue officer not com-
plying with uncomplicated and constitutionally sound poli-
cies and training. Christensen admitted at trial that the
County trained him on the illegality of sexual contact between
guards and inmates. Christensen also admitted he did not re-
quire more training to know his conduct was a crime. Even
plaintiff’s opinion witness Eiser conceded at trial that no
proof exists that better or more training could have dissuaded
Christensen from his predatory and assaultive behavior.
     C. No Need for Specialized Training
    The third requirement from Connick is that the degree of
training received by municipal employees is not relevant to
establishing liability under Canton’s single-incident hypothet-
ical. 563 U.S. at 68
. Per Connick, courts should be “concerned
with the substance of the training, not the particular instruc-
tional format.”
 Id. To rule 
to the contrary would “provide
plaintiffs or courts carte blanche to micromanage local govern-
ments.”
 Id. This would 
“‘engage the federal courts in an end-
less exercise of second-guessing municipal employee-training
Nos. 18-1498, et al.                                                        99

programs,’ thereby diminishing the autonomy of state and lo-
cal governments.” 
Connick, 536 U.S. at 75 
(Scalia, J. concur-
ring) (quoting 
Canton, 489 U.S. at 392
).
    Applying Connick, the inquiry is whether the Polk County
Jail trained its guards not to commit sexual assault, not the
amount or particulars of that training. 
Connick, 563 U.S. at 68
.
Nevertheless, the majority opinion is replete with conclusions
on the nature, quantity, and timing of the training: what lan-
guage was and was not included in Polk County’s written
policies; what topics were and were not discussed in training
sessions; when the training did and did not occur; and how
the training should have been done, in contrast to how it was
done.15 The majority opinion equates its conclusion of insuf-
ficient training with no training. But that decides the amount,

15 Majority op. at p. 2 (“Nor did the County provide any meaningful train-

ing on the topic [of sexual assault].”),
 id. (finding relevant 
that no further
training occurred after Jorgenson incident);
 

9 (“no training (in any
sense of the word) focused on the sexual harassment or assault”);
 id. at pp.
10–11 (quoting email that training would “hit the basics” but no require-
ment jail would be compliant with everything PREA calls for);
 

10
(finding relevant what topics Sergeant Schaefer did and did not train of-
ficers on);
 

11 (noting plaintiffs’ evidence on “inadequacy of Polk
County’s policies and training”);
 id. (recounting Eiser’s 
testimony on in-
sufficiency of training);
 

19 (“[T]he County’s training on preventing
and detecting the sexual harassment and abuse of inmates was all but non-
existent”);
 id. (“The training 
consisted almost exclusively of informing
guards of the easy and evident.”)
 id. (noting single 
session on PREA);
 

20 (stating jury could have found jail’s “sexual abuse prevention pro-
gram was entirely lacking”);
 id. (stating County 
“exacerbated the gap by
failing to use training as the means of making the policy prohibition a re-
ality”);
 

27 (identifying failure to institute a training session after
Jorgenson incident);
 

31 (stating County should have “put in place
some of Eiser’s proposed policies and training to prevent but especially
100                                                  Nos. 18-1498, et al.

type, and frequency of training. After 
Connick, 563 U.S. at 68
,
that is not the court’s inquiry.
    The County’s policies and training—that guards are not to
sexually assault inmates—admits of no nuance, separating it
from the deadly force training of Canton and the Brady prose-
cutorial obligations of Connick. The training here is imperative
and declarative: a jailer may not have sexual contact with an
inmate, and if the jailer does, the jailer will be fired and pros-
ecuted for a felony under Wisconsin law. That is not a gray
area, confounding correctional officials, nor did Christensen
so testify. In fact, he admitted exactly the opposite: he knew
what he was doing was criminal, and that no more training
would have altered that. Christensen, not the County, com-
mitted the legally and morally violative decisions here.
    Federal appellate courts to have addressed this issue have
held that specialized training is not required for an employee
to know that sexual assault is wrong:
    Eighth Circuit: McGuire v. Cooper, 
952 F.3d 918
, 
2020 WL
1069461 
at *3 (8th Cir. 2020) (concluding reasonable supervi-
sor would not know that a failure to specifically train a dep-
uty not to sexually assault a woman would cause that deputy
to engage in that behavior); 
Parrish, 594 F.3d at 999 
(“[W]e do
not believe that there is a patently obvious need to train an
officer not to sexually assault women.”); 
Andrews, 98 F.3d at
1077 
(“In light of the regular law enforcement duties of a po-




detect sexual abuse”);
 id. (“If Polk 
County had different policies or train-
ing, … these women or someone else may have felt able to report the
abuse … .”).
Nos. 18-1498, et al.                                           101

lice officer, we cannot conclude that there was a patently ob-
vious need for the city to specifically train officers not to rape
young women.”).
    Ninth Circuit: 
Flores, 758 F.3d at 1160 
(“We agree with our
sister circuits that ‘[i]n light of the regular law enforcement
duties of a police officer’ there is not ‘a patently obvious need
for the city [ ] specifically [to] train officers not to rape young
women.’” (quoting 
Andrews, 98 F.3d at 1077
)).
   Tenth Circuit: 
Schneider, 717 F.3d at 773
–74 (rejecting fail-
ure-to-train liability because “[s]pecific or extensive training
hardly seems necessary for a jailer to know that sexually as-
saulting inmates is inappropriate behavior” (quoting 
Barney,
143 F.3d at 1308
)); 
Barney, 143 F.3d at 1308 
(same).
   Eleventh Circuit: Sewell v. Town of Lake Hamilton, 
117 F.3d
488
, 490 (11th Cir. 1997) (reversing jury verdict against town
and rejecting failure-to-train liability because it is obvious that
a police officer should not “barter arrests for sexual favors”).
   The majority opinion’s expansion of § 1983 liability be-
comes even more apparent when one considers how many of
the decisions listed above would have come out the opposite
way under the majority opinion’s rule.
   D. Seventh Circuit
    For support the majority opinion cites to two decisions of
this court: Woodward v. Corr. Med. Servs. of Ill., Inc., 
368 F.3d
917 
(7th Cir. 2004), and Glisson v. Indiana Dep’t of Corr., 
849
F.3d 372 
(7th Cir. 2017) (en banc). Neither assists here.
   In Woodward, this court affirmed a judgment under Monell
when a contractor repeatedly failed to act in the face of known
violations. 368 F.3d at 930
. But unlike here, in Woodward the
102                                          Nos. 18-1498, et al.

contractor violated an express policy and willfully ignored its
own 
policies. 368 F.3d at 926
. Woodward was not a failure-to-
train case, and it did not grapple with the requirements of
Canton and Connick; indeed, it mentions neither case.
Woodward also found “a direct link” between the policy at is-
sue and the constitutional deprivation, an inmate’s 
suicide.
368 F.3d at 929
. That requirement comes from Bryan 
County.
520 U.S. at 404
. As noted above such a link is absent here.
    The majority opinion also cites Glisson as a pathway for
J.K.J. and M.J.J. to make their case to a jury. Glisson held that
an inmate healthcare provider could be liable under § 1983 for
failing to establish any protocol for the coordinated care of
inmates with chronic 
illnesses. 849 F.3d at 380
. But Glisson
involved a failure to enact a policy, not a failure to train em-
ployees.
 
Id. at 382. 
In Glisson, this court concluded that the
contractor had deliberately chosen not to have any policy as
to the coordination of care, even though the contractor had
actual knowledge that would result in deprivation of rights.

Id. at 382. 
That is not the case here. No one disputes the jail
had express zero-tolerance sexual assault policies and trained
its guards about those policies. And, actual knowledge of sex-
ual assaults is absent here. Even more, to align with Glisson,
sexual assaults by male guards would have to be as obvious
as not coordinating care for sick people, which was not shown
here.
VII. Conclusion
   A lone correctional officer covertly committed terrible sex-
ual assaults against two jail inmates. That employee is now
behind bars for 30 years and has millions of dollars of civil
judgments against him. At issue is whether his public em-
ployer is also liable for those crimes.
Nos. 18-1498, et al.                                        103

    Under the majority opinion, a single subordinate em-
ployee may secretly override municipal policy and create a
new policy under which that public employer is accountable.
That is vicarious liability, a collapse into respondeat superior
against which the Supreme Court has repeatedly warned for
60 years. By stepping out and recognizing fault and causation
on these facts, this decision departs from Supreme Court prec-
edents, imports a negligence standard into the law of deliber-
ate indifference, permits federal encroachment into an area of
traditional state authority, and splits with other federal cir-
cuits. On these facts and under the controlling law, the em-
ployee, not the employer, should be held responsible for these
plaintiffs’ injuries. Therefore, I respectfully dissent in part.