The Nebraska State Penitentiary (NSP) is forced to house many of its prisoners two men to a cell. This practice, known as "double celling," requires two men to share a 74-square-foot cell. The District Court1 held that, while this practice is not a per se constitutional violation, randomly placing two prisoners together under conditions as they exist at the NSP violates the plaintiffs' right to be free from cruel and unusual punishment. The defendants, the director of the Nebraska State Prison System and the Warden of the NSP, appeal that decision. The plaintiffs cross-appeal the portion of the order applicable to long-term inmates and the District Court's holding that the defendants are entitled to qualified immunity, shielding them from damages. We affirm the District Court's order. We also affirm the District Court's2 subsequent orders instructing the defendants to adopt a remedial plan, and granting the plaintiffs attorneys' fees.

The plaintiffs in this case are the class of inmates housed or to be housed in the four main housing units of the Nebraska State Penitentiary.3 They brought this case under 42 U.S.C. § 1983, challenging the conditions of their confinement in two respects. First, they contend that the practice of double celling violates the Eighth Amendment under the circumstances of this case. Second, they contend that the policy of holding both inmates responsible for contraband found in a double cell violates the Due Process Clause of the Fourteenth Amendment. An 18-day evidentiary hearing was held. The District Court rejected the due-process claim, but held that, while the practice of double celling inmates did not itself violate the Constitution, the manner in which the defendants were conducting that practice did.

The District Court made extensive findings of fact in its thorough opinion. Jensen v. Gunter, 807 F. Supp. 1463 (D.Neb.1992). We will recount those findings here only to the extent necessary for our review. The NSP, opened in 1981, is a maximum security prison, housing the State's most violent offenders. It consists of six housing units. Units one through four, the main housing units, are at issue in this case. The cells are approximately 74 square feet in size, and were intended to house one inmate. Because of the large prison population, that limitation has never been possible. The population at NSP hovers at about 150% of capacity. To accommodate the large number of prisoners, the NSP must double cell the inmates.

All adult male inmates in the Nebraska prison system are screened at the Lincoln Correctional Center prior to being assigned to an institution. Included in this screening is an assessment of personal risk factors such as potential for suicide, propensity for violence, victim potential, escape risk, security risk, and risk of drug and alcohol abuse. At the time this suit was filed, this assessment was used only in determining to which institution an inmate would be sent. It was not used to predict compatibility of inmates, or to help make cell assignments.

If prison officials are familiar with an inmate, they rely on that familiarity when making cell assignments. All other incoming inmates are assigned cells and cellmates on the basis of availability; in other words, randomly. The only exceptions occur if an inmate identifies another inmate with whom he has a serious problem, or if an inmate is clearly hostile toward another inmate or group of inmates. By way of example, a white supremacist will not be celled with a black inmate.

The guards in the control centers cannot see into the cells. Cell doors are solid, with a small window at eye level and a small vent. Monitoring is accomplished primarily through an intercom system. The speakers and receivers for this system are in the hallway outside of the cells, requiring inmates to shout through the vent in the door in order to get the attention of the guards on duty.

The statistics introduced during the 18-day bench trial in this case portray the NSP as an increasingly violent place. The number of inmates found guilty of violent offenses such as assault, aggravated assault, fights, and threatened bodily harm from 1983 to 1991 has increased at an alarming rate. Likewise, the number of prisoners requesting protective custody because they were afraid to remain in the general population increased "dramatically" in a relatively short period of time. The District Court also noted ample anecdotal evidence of violence in the prison. It particularly relied on the testimony of inmates Jensen and Hart. Inmate Jensen recounted how he was beaten by his cellmate while yelling for help. Inmate Hart described his numerous assaults on inmates who had the misfortune of being assigned to his cell.

Moreover, many acts of violence at the NSP go unreported and undocumented for three reasons. First, if an inmate reports violence by another inmate, the reporting inmate will be labeled a snitch by other inmates. Inmates do not want to be known as snitches, thus they often do not report violence. Also, if an inmate reports violence in which he is involved, both he and the other inmate will receive misconduct reports, and may be disciplined. Finally, if an inmate reports a violent incident, but there is neither a witness nor physical evidence of the reported violence, neither inmate is disciplined. These combined factors cause the statistics to understate the actual level of violence at the NSP.

When we initially reviewed this case, we found it necessary to remand it to the District Court for further findings in light of Farmer v. Brennan, 511 U.S. 825, 114 S. Ct. 1970, 128 L. Ed. 2d 811 (1994). Remand was required because the District Court, following our pre-Farmer precedents, found only that the defendants knew or should have known that the plaintiffs faced a pervasive risk of harm. Farmer requires a finding of actual knowledge on the part of prison officials in order to support an Eighth Amendment violation. Id. at ----, 114 S.Ct. at 1981.

On remand, the District Court met that requirement. El Tabech v. Gunter, 922 F. Supp. 244 (D.Neb.1996) (El Tabech III ). It found that the defendants were aware of the level of violence at the NSP, and that the violence spilled over to the double cells. Id. at 257-61. It went on to find that newly arriving inmates are randomly assigned to cells, Id. at 248-49, and that the defendants are aware of that fact. Id. at 252-54. Thus, the District Court reaffirmed its original position that the plaintiffs had proved an Eighth Amendment violation.

It is crucial at this point to understand thoroughly the contours of the District Court's holding. All of the talk by the parties, and to some degree by the District Court, of double celling and overcrowding at the NSP has the potential of overshadowing the real issues. The District Court did not hold that either double celling or overcrowding at the NSP violated the plaintiffs' Eighth Amendment rights. In fact, it specifically rejected those claims.

In its initial opinion finding liability, the District Court noted that "[t]his case is not an overcrowding case in the sense that plaintiffs are asserting that the penitentiary houses more inmates than it can manage" or for whom it can provide services. Jensen, 807 F. Supp. at 1469. In addition, it stated that it was using the term "overcrowding" to "refer to the fact that the number of inmates exceeds the design capacity of the facilities; it does not imply any judgment about that fact." Id. at 1468 n. 3. Thus, the issue in this case is not whether the NSP is overcrowded to a constitutionally significant degree. Notably, the plaintiffs never made any such claim.

The plaintiffs did, however, claim that the practice of double celling at the NSP was an Eighth Amendment violation. This claim is based on the notion that double celling can be a constitutional violation when it leads "to deprivations of essential food, medical care, or sanitation," or when it causes an "increase [in] violence among inmates or create[s] other conditions intolerable for prison confinement." Rhodes v. Chapman, 452 U.S. 337, 348, 101 S. Ct. 2392, 2400, 69 L. Ed. 2d 59 (1981); Cody v. Hillard, 830 F.2d 912, 914 (8th Cir.1987) (en banc), cert. denied, 485 U.S. 906, 108 S. Ct. 1078, 99 L. Ed. 2d 237 (1988). The District Court, however, held the plaintiffs' evidence that "double celling has taxed the penitentiary beyond its limits to provide essential human services, resources, and adequate physical structures" to be "lacking." Jensen, 807 F. Supp. at 1481. Likewise, the plaintiffs were unable to present sufficient evidence to establish that "double celling is the cause of an increase in violence institution wide." Id. at 1482. The plaintiffs do not appeal this decision. Thus, this case is not a "double celling" case in the conventional sense.

What, then, are the issues in this case? After the detailed factual development outlined above, the District Court found that the inmates in the double cells in the four main housing units faced a pervasive risk of harm in the form of violence or threatened violence from cellmates. Id. at 1483. The defendants, moreover, had been deliberately indifferent to that risk by randomly assigning incoming inmates to cells without assessing whether the new cellmates would be compatible. Id. at 1484. However, where those inmates who have been at the NSP for some time were concerned, prison officials consider their "propensities for violence" when making cell assignments, at least on an informal basis. Ibid. Stated differently, we see the District Court's order as holding that, while the practice of double celling at the NSP is not constitutionally suspect, the manner in which that practice was being carried out prior to this lawsuit violated the Eighth Amendment by exposing prisoners to a substantial risk of harm that is avoidable by simply considering whether incoming inmates will be compatible with their cellmates. Therefore, the issues that we must address are whether the plaintiffs are exposed to a substantial risk of physical harm, and, if so, whether the defendants have been deliberately indifferent to that risk by randomly double celling incoming inmates.

Before proceeding to a review of the merits of the District Court's order, it is helpful to explain how this characterization of the District Court's holding disposes of part of the plaintiffs' cross-appeal. The parties are of the opinion that the District Court found a constitutional violation only with respect to the newly arriving inmates. The plaintiffs argue that such a holding is erroneous because a violent incoming inmate could be celled with a nonviolent existing inmate, thus violating the right of the existing inmate to be free from assaults at the hands of the new inmate. They read the District Court's holding as ignoring this inevitability by finding a violation of only the incoming inmates' Eighth Amendment rights. If the District Court had made such an illogical finding, we would probably agree with the plaintiffs and reverse. However, it did not.

The District Court found that "a pervasive risk of harm exist[ed] in the four main housing units." Jensen, 807 F. Supp. at 1483. It did not limit this finding to newly arriving inmates. Then the District Court held that randomly assigning "newly arriving inmates into double cells under the volatile conditions that exist in the four main housing units is not a reasonable response to the pervasive risk of harm." Id. at 1484. Once again, it did not limit its holding to the newly arriving inmates. We read the District Court's opinion as finding a substantial risk of serious harm to all inmates in the form of violence from cellmates, to which risk the defendants have shown deliberate indifference by randomly assigning incoming inmates to cells. The District Court described it as "an Eighth Amendment violation respecting random double-celling of newly arrived inmates." El Tabech v. Gunter, 869 F. Supp. 1446, 1467 (D.Neb.1994) (El Tabech II ). That constitutional violation applies to every inmate who faces the possibility of being randomly celled either as an incoming inmate or with an incoming inmate, which is to say any inmate in the four main housing units. It is only the remedy, an injunction against randomly assigning cells to incoming inmates, that applies to incoming inmates only. The plaintiffs need not appeal this holding, because it is exactly what they want.

We can dispose one of the arguments raised by the defendants in similar fashion. The Violent Crime Control and Law Enforcement Act of 1994 forbids federal courts from holding "prison or jail overcrowding unconstitutional under the Eighth Amendment except to the extent that an individual plaintiff" proves the violation. 18 U.S.C. § 3626(a)(1). This legislation, the defendants argue, precludes class-action prison lawsuits challenging prison overcrowding. The case before us, they continue, is a class-action suit challenging overcrowding.

We need not decide whether the defendants' reading of the statute, that it precludes class-action prison suits, is correct because the statute does not apply to this case in any event. By its very terms, the statute applies to suits challenging "prison or jail crowding." This case, as we have seen, and as the District Court held, El Tabech II, 869 F. Supp. at 1450, is not a simple crowding case. It is a failure-to-protect case, focusing not on crowding but on the manner of assignment of new inmates to cells. Thus, regardless of what the Act means for class-action overcrowding cases, an issue we do not decide, it does not apply here. Furthermore, if the statute did apply, the relief granted here would not violate it. Individual plaintiffs have proved a violation of the Eighth Amendment, which is what the statute requires.

We also reject the other half of the plaintiffs' cross-appeal, challenging the District Court's grant of qualified immunity. The District Court correctly held that the precedents are diverse on the issue of the constitutional necessity of classification systems. See, e.g., McGill v. Duckworth, 944 F.2d 344 (7th Cir.1991), cert. denied, 503 U.S. 907, 112 S. Ct. 1265, 117 L. Ed. 2d 493 (1992); Walsh v. Mellas, 837 F.2d 789 (7th Cir.), cert. denied, 486 U.S. 1061, 108 S. Ct. 2832, 100 L. Ed. 2d 933 (1988). This diversity precludes a holding that reasonable prison officials would have known that they were violating the plaintiffs' clearly established rights, Harlow v. Fitzgerald, 457 U.S. 800, 818, 102 S. Ct. 2727, 2738, 73 L. Ed. 2d 396 (1982), by randomly assigning incoming inmates to double cells. Kennedy v. Schafer, 71 F.3d 292, 294 (8th Cir.1995), cert. denied, --- U.S. ----, 116 S. Ct. 2548, 135 L. Ed. 2d 1068 (1996).

The Eighth Amendment proscribes the infliction of "cruel and unusual punishments." The Supreme Court counsels that this amendment imposes upon prison officials the duty to "provide humane conditions of confinement." Farmer, 511 U.S. at ----, 114 S. Ct. at 1976. That duty, in part, requires those officials to take reasonable measures to " 'protect prisoners from violence at the hands of other prisoners.' " Ibid. (quoting Cortes-Quinones v. Jimenez-Nettleship, 842 F.2d 556, 558 (1st Cir.), cert. denied, 488 U.S. 823, 109 S. Ct. 68, 102 L. Ed. 2d 45 (1988)). The Eighth Amendment imposes this duty because being subjected to violent assaults is not "part of the penalty that criminal offenders pay for their offenses...." Rhodes, 452 U.S. at 347, 101 S.Ct. at 2399.

In order to prevail in failure-to-protect cases, inmates must make two essential showings. First, they must demonstrate that they are "incarcerated under conditions posing a substantial risk of serious harm." Farmer, 511 U.S. at ----, 114 S. Ct. at 1977. This objective requirement ensures that the deprivation is sufficiently serious to amount to a deprivation of constitutional dimension.

The second requirement inquires into the subjective state of mind of the prison official who is being sued. It mandates that the plaintiff inmates show that the official "knows of and disregards an excessive risk to inmate health or safety; the official must both be aware of facts from which the inference could be drawn that a substantial risk of harm exists, and he must also draw the inference." Id. at ----, 114 S.Ct. at 1979. This subjective requirement ensures that "only the unnecessary and wanton infliction of pain implicates the Eighth Amendment." Wilson v. Seiter, 501 U.S. 294, 297, 111 S. Ct. 2321, 2323, 115 L. Ed. 2d 271 (1991).

Chief Judge Posner has provided an apt description of what prisoners must prove in deliberate-indifference cases.

[T]o be guilty of deliberate indifference [prison officials] must know they are creating a substantial risk of bodily harm. If they place a prisoner in a cell that has a cobra, but they do not know that a cobra is there (or even that there is a high probability that there is a cobra there), they are not guilty of deliberate indifference even if they should have known about the risk, that is, even if they were negligent--even grossly negligent or even reckless in the tort sense--in failing to know. Bowers v. DeVito, 686 F.2d 616, 618 (7th Cir.1982). But if they know that there is a cobra there or at least that there is a high probability of a cobra there, and do nothing, that is deliberate indifference.

Billman v. Indiana Dept. of Corrections, 56 F.3d 785, 788 (7th Cir. 1995). In this case, the "cobra" in the cell is the potentially violent cellmate a prisoner faces each time he enters his cell at the NSP. The District Court found that this risk existed and that prison officials knew it existed.

Once that much is accomplished, prison officials still have a defense. They may be "found free from liability if they responded reasonably to the risk, even if the harm ultimately was not averted." Farmer, 511 U.S. at ---- - ----, 114 S. Ct. at 1982-83. This defense is available because the "duty under the Eighth Amendment is to ensure 'reasonable safety,' " Id. at ----, 114 S.Ct. at 1983 (quoting Helling, 509 U.S. at 33, 113 S.Ct. at 2481), a standard that is mindful of the very difficult task of warehousing the most dangerous people our society has to offer in a safe environment. Ibid. Thus, "[w]hether one puts it in terms of duty or deliberate indifference, prison officials who act reasonably cannot be found liable under the Cruel and Unusual Punishments Clause." Ibid.

Each step of this inquiry is fact-intensive. See Reece v. Groose, 60 F.3d 487, 490 (8th Cir.1995). We review the District Court's factual conclusions for clear error. Fed.R.Civ.P. 52(a).

A most contentious point in this case is whether the plaintiffs faced a substantial risk of serious harm, the first step in the inquiry outlined above. Assault at the hands of fellow inmates, as previously noted, is a "serious harm." The question, then, is whether the District Court's finding that inmates at the NSP faced a substantial risk of assault at the hands of cellmates is clearly erroneous. The District Court relied on several intertwined pieces of evidence to reach this conclusion.

First, the Court examined statistics compiled by the prison system reflecting the number of inmates found guilty of violent offenses by a disciplinary committee since the NSP opened. These statistics reveal that violent incidents have increased from a low of 77 in 1983 to a high of 359 in 1987. Through June of 1991, the last set of numbers available to the District Court, there had been 179 findings of guilt. In total, 1,774 guilty findings were made during these years. Also indicative of the level of violence is that there has been a dramatic increase in the number of inmates requesting protective custody in recent years. Notably, these increases occurred while the number of prisoners being double celled increased.

The defendants challenge the District Court's reliance on these statistics. They explain the increase in guilty verdicts on reporting changes and changes in the definition of assault, and the increase in protective-custody requests on numerous factors other than violence. They also challenge the District Court's conclusion that these statistics reflect a positive correlation between double celling and increasing violence. It is true that these statistics, as is the case with most statistics, are subject to more than one interpretation. That fact is not enough, however, for us to hold that the District Court's finding is clearly erroneous.

The District Court also found that the violence institution wide carried over into the double cells. In order to do so, the District Court relied on anecdotal evidence of violence in the form of testimony from prisoners. It also found that inside the double cells tensions are increased by the cell size, lack of privacy, the ineffective surveillance system, deterrents to reporting, the contraband rule, and the excessive amount of time spent on lockdown status.

This evidence is ample support for the District Court's conclusion that inmates in the double cells face a substantial risk of assault at the hands of their cellmates. This record compares favorably with those in prior cases holding that a jury question existed regarding whether there was a pervasive risk of harm. See, e.g., Butler v. Dowd, 979 F.2d 661, 674-75 (8th Cir.1992), cert. denied, 508 U.S. 930, 113 S. Ct. 2395, 124 L. Ed. 2d 297 (1993).

We now address the subject of our earlier remand, whether the defendants were deliberately indifferent to the risk that the plaintiffs faced. Farmer counsels that this question, like the first step in our analysis, is a question of fact. It is "subject to demonstration in the usual ways, including inference from circumstantial evidence, ... and a factfinder may conclude that a prison official knew of a substantial risk from the very fact that the risk was obvious." 511 U.S. at ----, 114 S. Ct. at 1981 (citation omitted). Furthermore, the Supreme Court in Farmer set forth certain types of evidence that can be useful in making this determination. When evidence is introduced

showing that a substantial risk of inmate attacks was longstanding, pervasive, well-documented, or expressly noted by prison officials in the past, and the circumstances suggest that the defendant-official being sued had been exposed to information concerning the risk and thus 'must have known' about it, then such evidence could be sufficient to permit a trier of fact to find that the defendant-official had actual knowledge of the risk.

Id. at ---- - ----, 114 S. Ct. at 1981-82.

The District Court undertook just such an analysis in this case. Initially, the District Court found that the statistics detailing the level of violence at the NSP, the same statistics utilized by the District Court to find a substantial risk of serious harm, were provided to the defendants. Thus, each defendant was actually informed "of virtually every verified incident of violence" that occurred at the NSP. El Tabech III, 922 F. Supp. at 257-58. Notably, the defendants admitted as much in their testimony. Id. at 258.

The District Court also found, with justification, that the defendants knew that the reported assaults understated the actual level of violence at the NSP. Once again, the defendants admitted as much in their testimony. Ibid. They were also aware of the physical construction of the cells that made monitoring the interior of the cells impossible, and the prison rules that acted as a disincentive to reporting assaults.

The District Court also relied on statements by prison officials, including the defendants, to support its finding. One official, a former director of Nebraska prisons, told the Nebraska legislature that "a widely accepted principle of prison management" is that "a prison cell should not be used for more than one prisoner." Id. at 260. To do so erodes the ability to manage the prison, and the potential for violence increases. Ibid. Likewise, one of the defendants, Harold Clarke, when he became director, told the Nebraska legislature that "predatory behavior is increasing and weaker inmates are forced to seek protective custody, which has increased by 70 per cent. in less than two years." Id. at 261. Finally, defendant Frank Hopkins testified that he knew that the increase in misconduct reports was "substantial." Ibid.

While the defendants would have us read the evidence differently from the District Court, we cannot say that the District Court's painstaking analysis, an analysis in complete harmony with Supreme Court precedent, resulted in a clearly erroneous conclusion.

The final issue is the reasonableness of the defendants' actions in light of their knowledge of the risk faced by the plaintiffs. The defendants do not argue that random cell assignments are reasonable. Rather, they challenge the District Court's factual conclusion that cell assignments for newly arriving inmates were, indeed, random.

All adult male prisoners in the Nebraska prison system undergo a detailed evaluation called a classification study. El Tabech III, 922 F. Supp. at 248. This evaluation, conducted at the Diagnostic and Evaluation Center, considers factors such as the crime for which the inmate was convicted, the inmate's criminal history, his medical history and psychological status, and any particular needs or problems the inmate may have. Ibid. A condensed version of the study called a scoring instrument is created from the classification study. Ibid. These studies, as well as observations of the inmate while he is housed at the Diagnostic and Evaluation Center, are used to determine to which institution the inmate will be sent.

Either of these resources could be used to help predict whether inmates who are slated to become cellmates will be compatible. However, the District Court found that neither resource is so utilized. El Tabech III, 922 F. Supp. at 248-49. Rather, the District Court found that cell assignments were made based on "space availability." Jensen, 807 F. Supp. at 1477. Space availability is just another way to say randomly. Ibid.

The District Court came to these conclusions after hearing testimony from numerous prison officials who are intimately familiar with the cell-assignment procedure. Virtually every witness, prison officials all, testified that the primary, if not sole, factors used in determining where an incoming inmate will be celled were "available bunks" and racial balance. El Tabech III, 922 F. Supp. at 249. The housing unit managers, who actually made the cell assignments, would not have seen either classification resource prior to making cell assignments. Nor did they know the inmate's size, age, or length of sentence, all important factors in predicting compatibility. Id. at 249-50.

It is true, as the defendants point out, that this initial cell assignment is not final in a technical sense. Soon after the inmate's arrival at the NSP he is subjected to an initial classification meeting with housing personnel. The classification instruments are available for this meeting, and the inmate is free, perhaps even encouraged, to inform housing personnel of any particular problems he may have with his cell assignment. Moreover, if prison officials know the prisoner, his cell assignment can be changed on that basis. Id. at 250.

The problem with the defendants' argument, as the District Court pointed out, is two-fold. First, the classification meeting took place quite often well after the cell assignment. Thus, the inmate would have to spend at least some time with a cellmate prior to the meeting. Ibid. Second, and most convincing, the District Court viewed the initial classification meeting more as an orientation session than as a cell-assignment method. Id. at 251. Rarely, if ever, had the meeting resulted in a change in cell assignment. Thus, for all practical purposes, the initial cell assignment was final. We have reviewed the record quite thoroughly and cannot find clear error in any of these findings or in the ultimate finding to which they led: that incoming inmates are assigned on a random basis.

It is even less difficult to affirm the District Court's finding that the defendants actually knew that incoming inmates were randomly assigned to cells. As the District Court pointed out, the defendants were intimately familiar with the workings of the NSP. They held several positions within the NSP as they worked their way up the Department's hierarchy. As they readily admitted, they were "responsible for the overall operation" of the NSP, and had "direct ongoing contact with housing unit procedures." Id. at 253-54. They also were familiar with the regulations outlining the initial classification procedure, including the fact that the regulation made no provision for cell assignments and inmate-compatibility determinations. This evidence was sufficient to allow the District Court to find that the defendants knew that cell assignments were random.

After liability was determined in this case, the District Court fashioned a remedy in the form of an injunction. That injunction imposes upon the defendants a duty to use the classification instruments available to them to try to predict whether incoming inmates and their cellmates will be compatible. El Tabech v. Gunter, No. CV87-L-377, slip op. at 26 (D.Neb.1994) (El Tabech I ). Citing language from Farmer v. Brennan, supra, the defendants claim that this injunction should not have been issued.

In Farmer, the Supreme Court wrote that in order to "establish eligibility for an injunction, the inmate must demonstrate the continuance of [prison officials' disregard of a risk of harm] during the remainder of the litigation and into the future." Farmer, 511 U.S. at ----, 114 S. Ct. at 1983. Parties may rely on "developments that postdate the pleadings and pretrial motions" in order to determine whether an injunction is appropriate. Ibid. Furthermore, prison officials who are violating prisoners' rights when a lawsuit is filed can "prevent the issuance of an injunction by proving, during the litigation, that they [are] no longer unreasonably disregarding an objectively intolerable risk of harm and that they [will] not revert to their obduracy upon cessation of the litigation." Ibid. n. 9.

The plaintiffs introduced sufficient evidence, in the first instance, to justify the District Court's finding that a constitutional violation had occurred, as we have discussed. Rather than immediately remedying that violation, the defendants chose to take a premature appeal to this Court, which was dismissed. When, after months of waiting, the defendants filed a remedial plan, it was inadequate and had to be modified. El Tabech I, slip op. at 18-26. That is sufficient evidence to support the District Court's finding that the violation would continue into the future.

We recognize that the defendants were merely proceeding as they felt they had a right to proceed by appealing to this Court. However, an immediate appeal was not their only option, even if such an appeal would have been proper. For example, they could easily have ceased random cell assignments, and then appealed the finding on liability. If we had eventually reversed the liability finding, they could have reverted to their prior cell-assignment system if they had so desired. Thus, we cannot find error in the District Court's holding that an injunction should issue.

We also disagree with the defendants' assertion that the District Court improperly imposed the burden of proving the above-described elements on them. When the District Court wrote that the defendants had not "met their burden of proving that an injunction is no longer necessary," El Tabech I, slip op. at 19, it was in reference to the language in Farmer setting forth how prison officials can avoid an injunction, not the language describing what the plaintiff would have to prove to merit an injunction. The District Court correctly placed the burden of proof. Farmer, 511 U.S. at ---- n. 9, 114 S. Ct. at 1983 n. 9 ("prison officials ... could prevent issuance of an injunction by proving ... that they were no longer unreasonably disregarding an objectively unreasonable risk of harm and that they would not revert to their obduracy upon cessation of the litigation").

The defendants make no attempt to argue that they had ceased violating the plaintiffs' rights prior to the injunction, nor could they. Indeed, they fought imposition of a change in cell-assignment methods every step of the way. The District Court correctly issued the injunction.

After finding liability on the part of the defendants and imposing the injunction, the District Court awarded the plaintiffs attorneys' fees. The Court did so under the provisions of 42 U.S.C § 1988, which allows for the award of attorneys' fees to prevailing parties in Section 1983 cases. The defendants make three challenges to this award. We will address each in turn.

In Hutto v. Finney, 437 U.S. 678, 98 S. Ct. 2565, 57 L. Ed. 2d 522 (1978), the Supreme Court held that the Eleventh Amendment does not bar an award of attorneys' fees ancillary to prospective relief, even though the fees would be paid from the state treasury. Id. at 693-98, 98 S.Ct. at 2574-77. The defendants argue that this holding was overturned sub silentio by Seminole Tribe of Florida v. Florida, --- U.S. ----, 116 S. Ct. 1114, 134 L. Ed. 2d 252 (1996). There, the Supreme Court reiterated its rule that Congress could abrogate the sovereign immunity that states enjoy by virtue of the Eleventh Amendment, but only if its "intention [is] unmistakably clear in the language of the statute," Id. at ----, 116 S.Ct. at 1123 (quoting Dellmuth v. Muth, 491 U.S. 223, 227-28, 109 S. Ct. 2397, 2400, 105 L. Ed. 2d 181 (1989)), and Congress acts "pursuant to a valid exercise of power." Ibid. (quoting Green v. Mansour, 474 U.S. 64, 68, 106 S. Ct. 423, 425-26, 88 L. Ed. 2d 371 (1985)). Section 1988, the defendants argue, has no "unmistakably clear" language, and, thus cannot abrogate sovereign immunity. (The fees in this case would be paid from the state's coffers, thus implicating the Eleventh Amendment).

This very argument was made, and rejected by the Supreme Court, in Missouri v. Jenkins, 491 U.S. 274, 109 S. Ct. 2463, 105 L. Ed. 2d 229 (1989). There, the State of Missouri argued that "the principle enunciated in Hutto has been undermined by subsequent decisions of [the Supreme] Court that require Congress to 'express its intention to abrogate the Eleventh Amendment in unmistakable language in the statute itself.' " Id. at 279, 109 S.Ct. at 2467 (quoting Atascadero State Hospital v. Scanlon, 473 U.S. 234, 243, 105 S. Ct. 3142, 3148, 87 L. Ed. 2d 171 (1985)); Welch v. Texas Dept. of Highways and Public Transportation, 483 U.S. 468, 107 S. Ct. 2941, 97 L. Ed. 2d 389 (1987). The "flaw in this argument," an argument identical to the one made by the defendants in this case, with one more citation added to the list, "lies in its misreading of the holding of Hutto." Jenkins, 491 U.S. at 279, 109 S. Ct. at 2467.

In Jenkins, the Supreme Court made it quite clear that "application of § 1988 to the States did not depend on congressional abrogation of the States' immunity." Ibid. Rather, Hutto held that Section 1988 "imposes attorney's fees 'as a part of costs.' Costs have traditionally been awarded without regard for the States' Eleventh Amendment immunity." Hutto, 437 U.S. at 695, 98 S.Ct. at 2576. Indeed, following Hutto and Jenkins "it must be accepted as settled that an award of attorney's fees ancillary to prospective relief is not subject to the strictures of the Eleventh Amendment." Jenkins, 491 U.S. at 279, 109 S. Ct. at 2467. In short, Section 1988 attorneys' fees do not depend on abrogation of sovereign immunity, and Seminole Tribe does not affect the fee award in this case.

The defendants' second challenge to the fee award involves the Prison Litigation Reform Act, Pub.L. No. 104-134, 110 Stat. 1321, to be codified at 18 U.S.C. § 3626 and 42 U.S.C. § 1997. This legislation, enacted well after both the liability and attorneys' fee determinations in this case, alters how prison cases are to be prosecuted in various ways. One provision of the Act, Section 803(7)(d), applies to attorneys' fees. In order to affect this case, Section 803(7)(d) must have retroactive application. We hold that it does not.

The Supreme Court recently announced a procedure for determining when a statute is to be applied to actions that occurred prior to enactment of the statute. Landgraf v. USI Film Products, 511 U.S. 244, 114 S. Ct. 1483, 128 L. Ed. 2d 229 (1994). Initially, a court should determine whether "Congress has expressly prescribed the statute's proper reach." Id. at ----, 114 S.Ct. at 1505. If so, the dictates of the statute should be followed, barring some constitutional prohibition. Id. at ---- - ----, 114 S. Ct. at 1497-98. Absent an express command, "the court must determine whether the statute would have retroactive effect, i.e., whether it would impair rights a party possessed when he acted, increase a party's liability for past conduct, or impose new duties with respect to transactions already completed." Id. at ----, 114 S.Ct. at 1505. If so, then the "traditional presumption" against retroactivity precludes application "absent clear congressional intent favoring such a result." Ibid.

Nothing in this portion of the Act expressly prescribes its reach. The Act was not in effect when the plaintiffs' attorneys accepted this appointment, when liability and fee determinations were made, or even when we remanded this case to the District Court in light of Farmer. When the plaintiffs' attorneys were exerting what the District Court quite fairly described as "herculean" efforts on their behalf, they expected to have their fee determined under Section 1988. If we apply the Act, those expectations will be foiled. Thus, application of the Act in this case would have the retroactive effect of disappointing reasonable reliance on prior law. That leaves us with the "traditional presumption" against retroactive application.

The defendants base their argument for retroactivity on Bradley v. Richmond School Bd., 416 U.S. 696, 94 S. Ct. 2006, 40 L. Ed. 2d 476 (1974). In Bradley, the District Court awarded attorney's fees and costs to parents who had prevailed in a school desegregation case on general equitable principles. Bradley, 416 U.S. at 706, 94 S.Ct. at 2013-14. While the appeal of the case was pending, Congress passed a statute that allowed courts to award fees to prevailing parties in school desegregation cases. Id. at 709, 94 S.Ct. at 2015. The Supreme Court held that the statute applied to the case at hand because courts are "to apply the law in effect at the time [they] render [their] decisions, unless doing so would result in manifest injustice or there is statutory direction or legislative history to the contrary." Id. at 711, 94 S.Ct. at 2016.

Two clear distinctions between Bradley and this case defeat the defendants' argument. First, in Bradley, there was no "manifest injustice" in allowing the fee statute to apply because the lower courts had already awarded fees on general equitable principles. As the Supreme Court wrote in Landgraf, it would be difficult to imagine a "stronger equitable case for an attorney's fee award" than a school desegregation case. Landgraf, 511 U.S. at ----, 114 S. Ct. at 1503. Given the availability of fees under an alternative theory, the new fee statute did not impose an "unforeseeable obligation" on the school board. Bradley, 416 U.S. at 721, 94 S.Ct. at 2021. Thus, being ordered to pay attorneys' fees was no great surprise, even though the legal theory under which those fees were to be imposed changed.

Conversely, in this case the plaintiffs and their attorneys have proceeded from the outset under the assumption that Section 1988 would apply to this case. They have litigated for literally years under that assumption. The Act was passed after our remand to the District Court, and after the District Court's findings on remand. Indeed, it did not become law until shortly before we were prepared to decide this case once and for all. It would be "manifestly unjust" to upset those reasonable expectations and impose new guidelines at this late date.

Furthermore, there is evidence of congressional intent contrary to retroactive application of this portion of the Act. Section 802, the section of the Act dealing with prospective relief, specifically provides that it "shall apply with respect to all prospective relief whether such relief was originally granted or approved before, on, or after the date of enactment of this title." § 802(b)(1). Section 803, conversely, is silent on retroactive application. Congress saw fit to tell us which part of the Act was to be retroactively applied, Section 802. The exclusion of Section 803 and its fee provisions from that clear statement is inconsistent with the defendants' argument for retroactivity.

Finally, we reach the issue of the correctness of the District Court's attorneys' fee award. Section 1988 provides for the payment of attorneys' fees to prevailing parties at the discretion of the court in Section 1983 cases. A fee award will not be altered absent an abuse of discretion. Butler, 979 F.2d at 676.

The plaintiffs in this Section 1983 case are clearly prevailing parties. The District Court, employing the procedure described by the Supreme Court in Hensley v. Eckerhart, 461 U.S. 424, 103 S. Ct. 1933, 76 L. Ed. 2d 40 (1983), awarded them $178,865.10 in fees and expenses. This method, known as the "lodestar" method, id. at 433, 103 S.Ct. at 1939, focuses on "the significance of the overall relief obtained by the plaintiff in relation to the hours actually expended on the litigation." Id. at 435, 103 S.Ct. at 1940. The district court multiplies the number of hours reasonably expended by the relevant market rate for legal services, then reduces the amount for partial success, if necessary.

The defendants challenge two elements used in the calculation of that award. First, they argue that the number of hours reasonably expended should have been lowered because of inadequate records on the part of the plaintiffs' attorneys. Second, they argue that the District Court should have allowed a greater reduction for limited success.

The gravamen of the defendants' argument regarding the number of hours reasonably expended is that the plaintiffs' lawyers' records were inadequate to inform the court of the nature and reasonableness of the services rendered. The District Court considered this argument in its thorough and detailed opinion on fees. It noted that "the documentation submitted by the plaintiffs' counsel was voluminous, detailed, and in most cases, fully in compliance with our local rules of practice." El Tabech II, 869 F. Supp. at 1460. However, "there were certain instances where the documentation was simply not sufficient to make an intelligent determination as to whether the hours expended were in fact reasonable." Ibid. Therefore, the District Court imposed an across-the-board reduction in hours of 10%.

Though this deduction is significant, the defendants would have us add another 50% deduction. As we have consistently held, "[t]he trial court is in a much better position than this court to view the evidence and to evaluate the testimony and work product of the attorney." Vosburg v. Solem, 845 F.2d 763, 770 (8th Cir.), cert. denied, 488 U.S. 928, 109 S. Ct. 313, 102 L. Ed. 2d 332 (1988). Here, the District Court clearly considered, and to a degree accepted, the arguments of the defendants. However, it held that the severe deduction advocated by the defendants was far too great given the amount of detail and explanation provided by the plaintiffs' attorneys. We see no abuse of discretion in that holding.

Likewise, the District Court rejected the defendants' request to reduce the fee award by 75% for partial success, choosing instead a 15% reduction. El Tabech II, 869 F. Supp. at 1464. The defendants request a 75% reduction in the award because the plaintiffs failed altogether on their Fourteenth Amendment claim related to the contraband rule, and enjoyed only partial success on their Eighth Amendment claims. The Eighth Amendment claim that double celling was causing a deprivation of essential services was rejected.

As the District Court noted, the defendants' argument ignores the fact that the plaintiffs prevailed on their "main claim" that the prison was unsafe. Id. at 1467-68. They obtained injunctive relief for that constitutional violation designed to make the prison safer for all inmates. It is of little significance that the plaintiffs' claims regarding services and the contraband rule were rejected, particularly given that those claims, and the evidence introduced in their support, helped significantly in proving that the NSP was unsafe. Those conditions, though short of a constitutional violation, contributed to the tense and hostile environment in which the plaintiffs existed. Finally, we reiterate that the constitutional violation affects any inmate who may be celled with a newly-arriving inmate, not just newly-arriving inmates. The relief obtained is major, and we see no abuse of discretion on the part of the District Court. The fee award is affirmed.

The order of the District Court is affirmed in all respects. We join the District Court in commending the parties for the quality and thoroughness of their work before the Court. Their briefs have been literate and their arguments well-reasoned. We also commend the District Court for its detailed and exhaustive work on this difficult and fact-intensive case. Its meticulous opinions have been most helpful to us.

Affirmed.