OPINION OF THE COURT

Jacqueline Heggenmiller1 and Tammy Davis appeal from a District Court order granting summary judgment on their claims brought under 42 U.S.C. § 1983 with respect to two prison administrators, John Terhune and Charlotte Blackwell (the “Administrative Defendants”). Appellants, both inmates at the Edna Mahan Correctional Facility (“EMCF”), alleged that a prison guard, Stewart Sella, sexually assaulted them over a period of time in 1997 through 1999 and that the Administrative Defendants were deliberately indifferent to a risk that such assaults would occur.2 Specifically, Heggenmiller and Davis contend the Administrative Defendants failed to have adequate privacy training and guard/inmate interaction policies and rules in place at the time of their assaults, despite (1) the existence of a consent decree requiring EMCF, like all New Jersey prisons, to have such training and policies in place and (2) the Administrative Defendants’ knowledge of prior instances of sexual misconduct by other guards at EMCF. We have jurisdiction pursuant to 28 U.S.C. § 1291 and will affirm.

I. STANDARD OF REVIEW

We review the District Court’s summary judgment order de novo, applying the standard that District Courts are to use. See Duffy v. Paper Magic Group, Inc., 265 F.3d 163 (3d Cir. 2001); Goodman v. Mead Johnson & Co., 534 F.2d 566, 573 (3d Cir. 1976). To affirm the District Court’s grant of summary judgment, we must be certain that there exists no genuine issue as to any material fact and that the Administrative Defendants are entitled to judgment as a matter of law even when the facts are viewed in the light most favorable to Appellants. Fed.R.Civ.P. 56(c); see also Anderson v. Liberty Lobby, Inc., 477 U.S. 242, 248, 106 S.Ct. 2505, 91 L.Ed.2d 202 (1986). The Administrative Defendants initially bear the burden to show the absence of a genuine issue of material fact, a burden that “ ‘may be discharged by “showing” — that is, pointing out to [this Court] — that there is an absence of evidence to support [Appellants’] case’ ” where, as here, Appellants “bear[ ] the ultimate burden of proof.” Singletary v. Pa. Dep’t of Corr., 266 F.3d 186, 193 (3d Cir. 2001) (quoting Celotex Corp. v. Catrett, 477 U.S. 317, 325, 106 S.Ct. 2548, 91 L.Ed.2d 265 (1986)). The burden then shifts to Appellants to “set forth specific facts showing that there is a genuine issue for trial.” Fed.R.Civ.P. 56(e); Anderson, 477 U.S. at 247-48. “There is no issue for trial unless there is sufficient evidence favoring the nonmoving party for a jury to return a verdict for that party.” Anderson, 477 U.S. at 249.

In Farmer v. Brennan, 511 U.S. 825, 114 S.Ct. 1970, 128 L.Ed.2d 811 (1994), which sets forth the Eighth Amendment deliberate indifference analysis for a pris*243on conditions case, the Supreme Court discussed the showing that a plaintiff must make to survive a summary judgment motion on a deliberate indifference claim: “[T]o survive summary judgment, [the plaintiff] must come forward with evidence from which it can be inferred that the defendant-officials were at the time the suit was filed, and are at the time of summary judgment, knowingly and unreasonably disregarding an objectively intolerable risk of harm and that they will continue to so.” Id. at 846. This court has expressly adopted this language in Farmer: “to defeat the summary judgment, [a plaintiff] must present enough evidence to support the inference that the defendants [are] ‘knowingly and unreasonably disregarding] an objectively intolerable risk of harm.’” Beers-Capitol v. Whetzel, 256 F.3d 120, 132 (3d Cir. 2001) (quoting Farmer, 511 U.S. at 846).

II. FACTS

Viewed in the light most favorable to Heggenmiller and Davis, the record evidence shows that Commissioner Terhune supervises EMCF Superintendent Blackwell, who supervises EMCF’s Director for Custody Operations, Dean Campbell, who in turn supervises a Captain Ochs. In December, 1998, Captain Ochs learned of Heggenmiller’s allegations of rape and other incidents of sexual assault perpetrated against her by Sella. Captain Ochs promptly reported the allegations up the chain of command, a formal investigation commenced, and Sella was ultimately fired and prosecuted. Later, in 2000, Director Campbell learned of Davis’ allegations of rape and other incidents of sexual assault perpetrated by Sella. Campbell promptly reported these allegations up the chain of command, a formal investigation commenced, and Sella was charged. In addition, a second guard, Regina Dozier — to whom Davis had first reported the assaults almost two years earlier, in 1998 — was investigated and fired for covering up Davis’ allegations.

In 2000, Commissioner Terhune became aware of the allegations against Sella through the Special Investigation Division of the New Jersey Department of Corrections, which conducted the investigations of Sella and Dozier. Neither Heggenmil-ler nor Davis personally reported their assaults to either Blackwell or Terhune, although it is not clear whether they or any other inmate would have had opportunity to do so. Deposition testimony from Terhune confirms that he was aware of other incidents of sexual activity among EMCF guards and inmates, all of which occurred prior to the assaults of Heggen-miller and Davis. That testimony, however, shows that Terhune recalled no specifics about any of those prior incidents.

At all relevant times, EMCF had in place written policies and training manuals for new guards that prohibited sexual contact with inmates. For example, the “Basic Course for State Corrections Officers” references a New Jersey state law making it illegal to have “criminal/sexual contact” with inmates. See N.J.A.C. 4A:2-2.3(a) (prohibiting undue familiarity with inmates); N.J.S.A. 2C:14-3b and 2C:14-2c(2) (prohibiting guard/inmate sexual relations). Instructional Unit 3.6 for the Basic Course for State Corrections Officers, entitled “Principles of Inmate Supervision and Discipline,” instructed guards to “[n]ever get on a personal level with an inmate and never obligate yourself to him/ her.” Instructional Unit 3.7, entitled “Staff-Inmate Relations,” instructed guards to avoid “fraternization” with inmates. Instructional Unit 3.9, entitled “Effects of Dehumanization in the Correctional Setting,” cited inmate “[s]hower[ing] while being observed by custody staff and other inmates” as an example of the type *244of guard behavior that could cause inmates to feel dehumanized.3 Additionally, at all relevant times, EMCF administrators, including Blackwell, verbally instructed guards that they would be terminated, prosecuted, or both in the event of sexual contact with inmates. See Affidavit of Charlotte Blackwell at ¶ 4 (informing new guards of “no contact” policy at orientation or when assigned to a post).

Additionally, pursuant to a consent decree applicable to all New Jersey prisons that had been in effect since 1991, EMCF was required to (1) make minimum privacy accommodations for female inmates and (2) institute formal training and policies for guards with respect to gender sensitivity and inmate privacy in areas such as locker rooms and showers. While EMCF acted with respect to the former, by installing shower pegs and curtains, it did not act with respect to the latter. Instead, EMCF chose to rely on its pre-existing training and policy, set forth above, in conjunction with its practice of firing and prosecuting guards who violated that policy.

Sella, the EMCF guard that raped and sexually assaulted inmates Heggenmiller and Davis, was not the first to be fired or prosecuted for improper contact with a female inmate. EMCF records and deposition testimony reveal as many as six “familiarity” or “contact” incidents between 1994 and 1998, all involving different guards, none of them Sella. At least five of these guards were fired and prosecuted for conduct ranging from sexual assault to consensual sexual contact with current and former inmates. Additionally, when viewed in the light most favorable to Appellants, deposition testimony from guards and their immediate supervisors, including Captain Ochs, shows the administrative hierarchy of EMCF knew of up to a total of ten “familiarity” or “contact” incidents over a period of time dating back to 1990. However, although all of these incidents involved improper interactions between male guards and female inmates, the record does not show that all of them were as serious as the rapes and assaults suffered by Heggenmiller and Davis. For example, at least some of the prior incidents involved consensual sexual contact among guards and females who were either former inmates or inmates that were soon to be released.4

*245III. ANALYSIS

The sexual assaults perpetrated against Heggenmiller and Davis (and the cover-up of Davis’ attempt to report the assaults) are not to be countenanced in any way. That said, the only claims before us in this appeal allege the supervisory liability of the Administrative Defendants — persons who indisputably had no direct part in either the assaults or the cover-up. Notwithstanding the offensiveness of the underlying crimes, the supervisory liability that Heggenmiller and Davis seek to impose is not without strict limits. See, e.g., C.H. ex rel. Z.H. v. Oliva, 226 F.3d 198, 201-02 (3d. Cir. 2000) (en banc) (“It is, of course, well established that a defendant in a civil rights case cannot be held responsible for a constitutional violation which he or she neither participated in nor approved.”). To establish supervisory liability, the misconduct of the guards in this case must be “affirmatively link[ed]” to the actions or inactions of the Administrative Defendants. Rizzo v. Goode, 423 U.S. 362, 371, 96 S.Ct. 598, 46 L.Ed.2d 561 (1976). This causal link is necessary because “[t]here is no vicarious, respondeat superior liability under [Section] 1983.” Oliva, 226 F.3d at 202.

A. Failure to Protect

It is well-settled that prison officials have a duty under the Eighth Amendment to “take reasonable measures to guarantee the safety of inmates.” Farmer, 511 U.S. at 832 (internal quotations omitted). Here, to establish an Eighth Amendment violation, Appellants must show that they faced a “substantial risk of harm” to which the Administrative Defendants acted with “deliberate indifference.” Id. at 828-29. The predicate substantial risk of harm must be objectively serious. Id. Deliberate indifference, in turn, requires a subjective showing that a prison official “knows of and disregards” that risk: “the official must be both aware of facts from which the inference could be drawn that a substantial risk of serious harm exists, and he must also draw the inference.” Id. at 837. We have thus previously explained that “(1) ‘the deprivation alleged must be, objectively, sufficiently serious;’ and (2) the ‘prison official must have a sufficiently culpable state of mind.’ ” Beers-Capitol, 256 F.3d at 125 (quoting Farmer, 511 U.S. at 834) (internal quotation marks and citations omitted).

The parties dispute whether the prior incidents between guards and female inmates at EMCF — not only their aggregate number, but also their relative degrees of harmfulness — placed the Administrative Defendants on notice that inmates like Heggenmiller and Davis faced a substantial risk of being raped and sexually assaulted by an EMCF guard. In the light most favorable to Appellants, the record suggests up to ten incidents; however, even in such light, only three can be said to approach the same level of objectively serious harm as those assaults suffered by Heggenmiller and Davis. Though sanctionable under EMCF policy, as evidenced by the firings, these other incidents are qualitatively dissimilar in nature when juxtaposed against rape and sexual assault. As such, even in the aggregate, we find that they could not have placed the Administrative Defendants on notice of a “substantial risk of harm.” Farmer, 511 U.S. at 828-29.

While “a defendant’s knowledge of a risk can be proved indirectly by circumstantial evidence,” Beers-Capitol, 256 F.3d at 131, including evidence that “the risk was obvious,” id., the following evidence demonstrates why such a showing is precluded in *246this case. First, none of the prior incidents is shown to have involved Sella, the guard who assaulted both Heggenmiller and Davis. Different guards committed the various past offenses. Second, the record shows that, as to these past incidents, the Administrative Defendants investigated, fired, and prosecuted at least five of the six guards involved in the incidents occurring between 1994 and 1998. This is evidence of reasonable action on the part of the Administrative Defendants that not only removed the possibility that those guards could engage in repeated inappropriate conduct with female inmates, but that also served as an unambiguous deterrent to EMCF’s other guards, including Sella. Third, even viewed in the light most favorable to Appellants, there is no evidence in the record that EMCF administrators either looked the other way or attempted to intervene on behalf of any guard shown to have violated the Administrative Defendants’ “no contact” rule.

As such, we do not believe that cases such as our decision in Hamilton v. Leavy, 117 F.3d 742 (3d Cir. 1997), apply. In Hamilton, we reversed a grant of summary judgment, concluding that a prisoner’s Eighth Amendment deliberate indifference claim against a prison official possessed sufficient circumstantial evidence to allow an inference that the prison officials were aware of a substantial risk. Hamilton, 117 F.3d at 747-48. The types of evidence supporting that conclusion in that case included evidence showing the official knew that the prisoner was more at risk than other prisoners because he had been an informant, and that the official had personally approved the prisoner for protective custody twice before. See id. at 747. Analogous circumstantial evidence is lacking here. Further, in Hamilton, we reversed the grant of summary judgment because there was “a genuine issue of material fact regarding whether the [defendant’s] response to the risk Hamilton faced was reasonable.” Id. at 748. Here, in contrast, there can be no dispute but that EMCF’s combination of training materials, its “no contact” policy, and its demonstrated practice of investigating, firing, and referring misconduct cases for prosecution constituted a reasonable response to any risk that was observable from the prior incidents at EMCF.

The record lacking any evidence of deliberate indifference on the part of the Administrative Defendants with respect to an objectively serious and substantial risk of harm, summary judgment on Appellants’ failure to protect claim was properly granted. See Beers-Capitol, 256 F.3d at 133 (even if a prima facie showing of deliberate indifference is made, a supervisor can rebut the showing by establishing either a lack of the requisite level of knowledge or that reasonable steps were taken to prevent the harm for occurring); see also Farmer, 511 U.S. at 844.

B. Failure to Train

Supervisory liability on an Eighth Amendment claim for failure to properly supervise or train is governed by this Court’s decision in Sample v. Diecks, 885 F.2d 1099 (3d Cir. 1989) (holding supervisory liability may attach if supervisor implemented deficient policies and was deliberately indifferent to the resulting risk); see also City of Canton v. Harris, 489 U.S. 378, 109 S.Ct. 1197, 103 L.Ed.2d 412 (1989). In Sample, we held that a prison official may be liable under the Eighth Amendment if (1) existing policy or practice creates an unreasonable risk of Eighth Amendment injury; (2) the supervisor was aware that the unreasonable risk was created; (3) the supervisor was indifferent to that risk; and (4) the injury resulted from the policy or practice. Sample, 885 F.2d *247at 1118; see also Beers-Capitol, 256 F.3d at 134. As the District Court correctly noted, while our decision in Sample observes that the requisite “unreasonable risk” is normally shown by evidence that “such harm has in fact occurred on numerous occasions,” Sample, 885 F.2d at 1118, “there are situations in which the risk of constitutionally cognizable harm is so great and so obvious that the risk and the failure of supervisory officials to respond will alone support findings of the existence of unreasonable risk, of knowledge of that unreasonable risk, and of indifference to it.” Id.

Here, however, even when construed in a light most favorable to Appellants, the evidence is insufficient under either approach. As discussed supra, evidence of numerous similar harms is nonexistent. At most three prior incidents reached the severity of the assaults suffered by Heggenmiller and Davis. The remainder of Appellants’ evidence is neither sufficiently similar nor sufficiently serious to support their claim under Sample. At the same time, the mere presence of a consent decree applicable to all New Jersey prisons, not just EMCF, cannot satisfy Sample’s alternative method for establishing an unreasonable risk of serious harm. The existence of the consent decree, which addresses guard assignments, privacy accommodations for both male and female inmates, and training and policies respecting inmate privacy concerns, cannot be said to have made it “so great and so obvious” to the Administrative Defendants that there existed a risk of assaults of the type Heggenmiller and Davis would suffer. Id.

Even if the combination of the pri- or incidents and the existence of the consent decree together were sufficient to satisfy the first two showings required by our decision in Sample, Appellants would still need to make a sufficient showing with respect to deliberate indifference and causation. Id. Such showings are not possible on this record. The Administrative Defendants’ vigorous enforcement of the “no contact” rule — evidenced by the firing and/or prosecution of five of the six guards responsible for the six documented incidents at EMCF between 1994 and 1998— precludes a finding that the Administrative Defendants demonstrated a “deliberate indifference,” since these were “reasonable steps to prevent the harm from occurring.” Beers-Capitol, 256 F.3d at 133. Moreover, the Administrative Defendants had promulgated policies forbidding sexual contact between correctional officers and inmates; these policies were communicated to all officers in their training, and were enforced by the regulations and criminal laws of the State of New Jersey. See, e.g. N.J.A.C. 4A:2-2.3(a) (prohibiting undue familiarity with inmates); N.J.S.A. 2C:14-3b and 2C:14-2c(2) (prohibiting guard/inmate sexual relations); see also Affidavit of Charlotte Blackwell at 114 (informing new guards of policy at orientation or when assigned to a post).

On these facts, the requisite showing of culpability necessary to sustain liability cannot exist. “A showing of simple or even heightened negligence will not suffice.” Board of Comm’rs of Bryan Cty. v. Brown, 520 U.S. 397, 407, 117 S.Ct. 1382, 137 L.Ed.2d 626 (1997). And while it is true that Canton does “not foreclose the possibility” of liability upon evidence of even a single violation of a federal right, Brown, 520 U.S. at 409, such possibility is limited to “a narrow range of circumstances” where a violation of federal rights is “a highly predictable” consequence of a failure to adequately train. Id. For the reasons we have discussed, that possibility is precluded by the evidence in this ease.

*248Finally, even assuming evidence sufficient to support the requisite levels of risk and indifference, Appellants would still need to show that the Administrative Defendants’ actions and inactions were “the moving force” behind the harms suffered by Heggenmiller and Davis. See Sample, 885 F.2d at 1117 (internal quotation omitted). “[I]t is not enough for a plaintiff to argue that the constitutionally cognizable injury would not have occurred if the superior had done more than he or she did.” Id. at 1118. Instead, Appellants’ evidence must demonstrate a close causal relationship between the Administrative Defendants’ conduct and Appellants’ injuries. See id.; see also City of Canton, 489 U.S. at 391. Here, the District Court correctly concluded that such a link cannot be drawn from the record evidence. This includes the testimony of Appellants’ expert, Alvin W. Cohn. For the reasons articulated by the District Court, that expert testimony is insufficient, even viewed in the light most favorable to Appellants, to show that the attacks “resulted from” the Administrative Defendants’ conduct. Sample, 885 F.2d at 1118. For all of these reasons, summary judgment on Appellants’ failure to train claim was also properly granted.

IV. CONCLUSION

Because Appellants were required to come forward with specific facts showing there was an issue for trial and did not do so, the August 27, 2003 order of the District Court’s granting summary judgment to the Administrative Defendants will be affirmed.-

dissenting.

I dissent in this case because I disagree with the majority’s view that, even giving the plaintiffs the benefit of all reasonable inferences, there are no material issues of fact for a jury to consider.

Plaintiff/Petitioners Heggenmiller5 and Davis were inmates at EMCF in 1998 when they were repeatedly sexually assaulted by a prison guard. In the preceding five years, at least five other female inmates were also sexually assaulted by various guards at EMCF. Deposition testimony offered by plaintiffs suggests up to five additional incidents — for a total of ten sexual assaults between 1994 and 1998.

In 1991, shortly before this series of sexual assaults began, the New Jersey Department of Corrections was ordered by the Czizmada Consent Decree (“Decree”) to assign officer positions at EMCF without reference to gender, with the exception of certain especially sensitive posts. It was simultaneously ordered to institute a program to train officers for gender neutral assignments and to teach them about privacy concerns and gender sensitivity. EMCF gender-neutralized its officer assignments, but it did not add training about gender sensitivity and privacy to its curriculum. Instead, it deemed the training it already had in place adequate to train the male guards newly introduced into the female inmate facility. In fact, the record suggests that EMCF did not alter or update its curriculum in any way in response to the Decree.6 Nor did it do so after any of the five sexual assaults that were uncovered in the five years before Heggenmiller and Davis were assaulted.

*249In my view, a jury could well conclude that the assaults on Heggenmiller and Davis were caused by the defendants’ failure to implement the provisions of the 1991 Czizmada Consent Decree at any time before plaintiffs were assaulted in late 1998.

The pattern of criminal sexual activity at EMCF, and the circumstances surrounding it, are important under the four-part test of Sample v. Diecks, 885 F.2d 1099 (3d Cir. 1989), that applies when a plaintiffs theory of liability is predicated on a defendant’s policymaking role, see Beers-Capitol v. Whetzel, 256 F.3d 120, 135 (3d Cir. 2001). Plaintiffs can meet the first prong, requiring them to show that defendants maintained a policy or practice that created an unreasonable risk of harm, by presenting “evidence that such harm has in fact occurred on numerous occasions.” Sample, 885 F.2d at 1118. Plaintiffs can also show, under the second prong, that defendants were aware of the unreasonable risk by presenting evidence that the alleged risk was “longstanding, pervasive, well-documented, or expressly noted by prison officials in the past, and [that] the circumstances suggest that the defendant-officials] being sued had been exposed to information concerning the risk and thus must have known about it.” Farmer v. Brennan, 511 U.S. 825, 842, 114 S.Ct. 1970, 128 L.Ed.2d 811 (1994); see also Baker v. Monroe Township, 50 F.3d 1186, 1193-94 (3d Cir. 1995) (reversing summary judgment in part because factfinder could infer that defendant knew how plaintiffs were being treated by officers under his supervision). It is clearly for a jury to decide whether the series of sexual assaults at EMCF constitutes sufficient evidence under the first two prongs of Sample.

The majority opines that not all of the prior incidents at EMCF “were as serious as the rapes and assaults suffered by Heg-genmiller and Davis,” and suggests that “at least some of the prior incidents involved consensual sexual contact among guards.” However, the record indisputably shows at least five prior sexual relationships between guards and inmates, two of which continued after the inmates were released, and even defendants concede that prison inmates cannot legally consent to sex with their prison guards.7 Moreover, as each of those five sexual assaults resulted in discharge and/or criminal convictions, they were almost certainly known to the defendants (indeed, defendants concede generally they were aware of prior sexual assaults).

Even if plaintiffs were able to show that defendants created an unreasonable risk of harm and were aware of that risk, the majority concludes that plaintiffs could not show deliberate indifference to that risk under the third prong of Sample. I would not foreclose such a factual finding. A reasonable jury could conclude that a failure to institute any gender sensitivity or privacy training — even after newly introducing male guards into a female prison, despite a series of sexual assaults by guards on prisoners, and notwithstanding a standing legal obligation to provide such training under the Consent Decree — was reckless. See Farmer, 511 U.S. at 836 (“acting or failing to act with deliberate indifference to a substantial risk of serious harm to a prisoner is the equivalent of recklessly disregarding that risk”).

*250The majority endorses defendants’ own apparent belief that the training in place for EMCF guards was sufficient. Guards had indeed been provided important training long before the Consent Decree. A training manual allegedly similar to that used in 1994 (when Stewart Sella, the guard who sexually assaulted Heggenmil-ler and Davis, was trained) informed recruits that sexual contact with inmates is illegal. The manual also discusses fraternization and undue familiarity, though that section of the manual focuses on avoiding preferential treatment and does not address sexual relations with inmates. Inmates’ lack of privacy (together with their lack of choice) and its dehumanizing effects are also discussed in the manual, though the recommended ways to minimize dehumanization make clear that this training was not intended to prevent sexual harassment.

Plaintiffs Heggenmiller and Davis argue, however, that EMCF should have provided its guards with specific training on privacy rights and gender sensitivity. They claim that defendants should have instituted written guidelines to govern prison guards’ conduct in female inmates’ living quarters. The record shows that guards have thus far relied, with varying success, on their “common sense” as to these matters. The Chief of Training at the Correctional .Officers Academy, Sally Scheidemantel, had no recollection of specific training that dealt with privacy accommodations and, though the record includes a training text on “Supervision of Inmates by Employees of the Opposite Sex” dated March 1983, she was skeptical that this text, in a format that would not have been used by the Police Training Commission, was used during her tenure, which began in 1993.

The majority gives defendants much credit for firing any guard reported for engaging in sexual relations with inmates. While a failure to discharge sexual assailants would be truly shocking, the converse is insufficient, standing alone, to immunize defendants from § 1983 liability. Whatever the abstract merits of the training curriculum, or the theoretical sufficiency of a “zero-tolerance” discharge policy, a jury could conclude, based on the evidence in this case, that defendants should have responded to the pattern of sexual assaults at the prison by taking additional “reasonably available measures to reduce or eliminate that risk.” Berg v. County of Allegheny, 219 F.3d 261, 275 (3d Cir. 2000). “If a [training] program does not prevent constitutional violations ... decisionmakers may eventually be put on notice that a new program is called for. Their continued adherence to an approach that they know or should know has failed to prevent [criminal] conduct by employees may establish ... the ‘deliberate indifference’ necessary to trigger ... liability.” Bd. of County Comm’rs v. Brown, 520 U.S. 397, 407, 117 S.Ct. 1382, 137 L.Ed.2d 626 (1997).

Finally, the majority determines, under the fourth prong of Sample, that defendants’ inaction did not cause Sella’s sexual assaults. The Supreme Court has acknowledged that “[predicting how a hypothetically well-trained officer would have acted under the circumstances may not be an easy task for the factfinder,” but it concluded that “judge and jury, doing their respective jobs; will be adequate to the task.” City of Canton v. Harris, 489 U.S. 378, 391, 109 S.Ct. 1197, 103 L.Ed.2d 412 (1989). The causation element in a § 1983 claim, like other difficult questions of fact, is normally reserved for a jury. See Rivas v. City of Passaic, 365 F.3d 181, 193 (3d Cir. 2004).

While it is impossible to know with absolute certainty whether Sella would have committed a sexual crime with better training, “the existence of a pattern of *251[criminal] conduct by inadequately trained employees may tend to show that the lack of proper training, rather than a one-time negligent administration of the program or factors peculiar to the officer involved in a particular incident, is the ‘moving force’ behind the plaintiffs injury.” Brown, 520 U.S. at 407-08. The fact that the pattern of sexual misconduct at EMCF arose after the Decree may tend to show that the introduction of large numbers of male officers into female inmates’ living quarters without proper training caused a series of sexual assaults to occur. Moreover, in this case, the record suggests that Sella’s relationships with Heggenmiller and Davis were “gradual” in the sense that he first stared and glared at them, then “flirted” with them, and only later engaged in more explicit sexual activity. A jury could very well be persuaded that formal privacy guidelines would have thwarted Sella’s improper conduct before it escalated to the level of sexual assault.

Of course, we cannot say with certainty that a reasonable jury would find defendants liable on the facts of this case. But precisely because we cannot know whether a reasonable jury would hold prison administrators Terhune and Blackwell liable for the sexual assaults that took place in their prison, I would reverse the order of summary judgment and allow Jacqueline Heggenmiller and Tammy Davis to go forward with their claims. I respectfully dissent.