COFFIN, Chief Judge.
In this case, New Hampshire state prisoners in protective custody mount a constitutional attack on the conditions of their imprisonment. The decision of the district court largely favored the prisoners, and only the state has appealed. We affirm in part and reverse in part.
There are about thirty-five protective custody prisoners at the state prison. The life of a protective custody prisoner is, for the most part, boring and filled with inconvenience; the tedium is relieved on occasion by moments of terror.1 Protective custody, as the name implies, is for prisoners who fear for their safety if they are housed with the general population. Some prisoners in protective custody have cooperated with the state in criminal prosecutions; others have been assaulted while in the general population; still others have run up large debts to other prisoners; some simply cannot get along with their fellow prisoners. Protective custody is not a punishment. It is voluntary, and protective custody prisoners may request a transfer to the general population at any time.
Protective custody prisoners are housed in the "Annex", a four-story building attached to the main cell block in the New Hampshire State Prison. The bottom floor of the Annex contains the prison library and a gang shower. The top three floors each have two outside corridors, 22 back-to-back cells, and a single stall shower. Eleven cells face each window-lined corridor. The corridors, or tiers, are about seventy feet long and eight feet wide. The cells are about six feet by eight feet, with steel walls, a steel bunk, and a toilet and sink. On the top floor, a steel mesh wall divides the corridors lengthwise. It is meant to keep prisoners from pelting the guards with food or more dangerous objects. It also has the effect of cutting off some light and air from the windows. Protective custody (PC) prisoners are kept on the top two floors of the Annex. The second floor houses prisoners under discipline, and "trustees". Except during the night shift, two guards keep watch in the Annex. One stays on the first floor at all times, guarding the entrances and answering the phone. The other escorts inmates to and from showers, visits, medical calls, and the like.
A PC prisoner's day begins with breakfast at 7:00 or 7:30 each morning. Breakfast is brought from the kitchen in a heated cart and carried to each cell on a tray. On each weekday, twenty men may take showers from 8:00 to 10:00 a.m. Every prisoner showers five times every two weeks, on Tuesdays and Thursdays one week and on Monday, Wednesday, and Friday the next. During the shower period, inmate workers clean the tiers and perform other chores. Lunch comes, again on trays, at about 10:30 a.m., roughly an hour before the general population eats. Unlike the general population, PC inmates seldom get second helpings. At about 11:30 a.m. the entire prison population is locked in, and a prisonwide head count takes place. Mail and medicine is also distributed at this time.
About twice a week, PC prisoners may go outdoors for recreation or to buy items from the prison canteen from 12:00 to 12:45 p.m. Once a week, at this time, they may go to the prison library for a little less than an hour. There they may do legal research and borrow a book, but no inmate is allowed to check out more than one book at a time. Sometimes a book will be sent to an inmate on request during the week. If the inmate is willing to pay the price, copies of cases can be ordered from the New Hampshire Supreme Court library.
Every weekday, PC inmates are let out of their cells for "tier time" from 12:45 to 2:45 p.m. They are confined to their corridors, but may exercise and visit with each other. At 2:45 they return to their cells, and inmate workers again clean the tiers. Supper is served at 3:30; the general population eats its supper an hour later. Again, second helpings are rare in the Annex. Except for special reasons, such as a visitor, PC prisoners remain locked up for the rest of the evening.
When PC inmates have visitors or wish to obtain medical or educational services, they go to the administrative section of the prison. This means that they must pass through the main cell block, escorted by a single guard. General population prisoners are often out of their cells at the time, so the trip presents a danger of both physical and verbal attack. The danger persists during visits, when PC and general population prisoners are intermingled.
In theory, many occupational, educational, and rehabilitative programs are open to PC prisoners, but as a practical matter PC inmates are afraid to join any program that is also open to the general population. Only a small number of programs are closed to the general population and thus available to PC inmates. A few inmates work in the Annex. Some get high school equivalency tutoring. Bible study class, drug and alcohol counseling, legal and medical services are also available. The general population has a much wider range of activities open to it, including woodcarving, pottery, yoga, vocational training, and many fairly sophisticated work opportunities. In addition, PC inmates who cannot work do not get "idle pay" - currently 38 cents a day - even though nonworking prisoners in the general population do get it. On the other hand, unlike the general population, PC inmates need not pay for tobacco, razor blades, and the like. Religious services are not held in the Annex, primarily because of lack of interest, but a minister and a priest are available for counseling or for administration of the sacraments.
The district court drew from the Eighth and Fourteenth Amendments closely related, if not identical, tests to evaluate the constitutionality of these practices. Nadeau v. Helgemoe, 423 F. Supp. 1250, 1262-65 (D. N.H. 1976). In considering the proscription of "cruel and unusual" punishment by the Eighth Amendment, the court recognized two traditional tests. The first is whether the punishment is so barbarous that it offends society's evolving sense of decency. See, e.g., Ingraham v. Wright, 430 U.S. 651, 97 S. Ct. 1401, 1410, 51 L. Ed. 2d 711 (1977); Estelle v. Gamble, 429 U.S. 97, 50 L. Ed. 2d 251, 97 S. Ct. 285 (1976); Trop v. Dulles, 356 U.S. 86, 101, 2 L. Ed. 2d 630, 78 S. Ct. 590 (1958) (opinion of Warren, C.J.). The second is whether the punishment is grossly disproportionate to the offense. See Weems v. United States, 217 U.S. 349, 30 S. Ct. 544, 54 L. Ed. 793 (1910). In addition, the court embraced what it called the third test under the Eighth Amendment. The essence of this test is that all punishments must have a legitimate penological purpose. See, e.g., Jordan v. Fitzharris, 257 F. Supp. 674 (N.D. Cal. 1966). Moreover, the court held:
"The fact that defendants have granted specific privileges and benefits to the general population gives rise to a presumption that those privileges and benefits serve a legitimate penological purpose; the wholesale denial to a few of the exact same privileges and benefits lifts the cloak of that presumption from defendants' acts." 423 F. Supp. at 1263-64.
Four penological objectives were recognized; deterrence, insulating society from offenders, rehabilitation, and prison security. See Pell v. Procunier, 417 U.S. 817, 822-23, 41 L. Ed. 2d 495, 94 S. Ct. 2800 (1974). The court declared that lack of funds was not an acceptable justification: "fiscal inability alone does not excuse the denial to some of privileges and benefits deemed necessary to achieve legitimate penological ends." 423 F. Supp. at 1264. The court also rejected administrative convenience as a proper reason for distinguishing between the two populations.
Parallel to this test runs a special version of the equal protection "rational basis" standard. In the absence of a fundamental interest or suspect classification, the court recognized that the state may draw any distinction between persons, so long as the classification is rationally related to some permissible state goal. See, e.g., New Orleans v. Dukes, 427 U.S. 297, 303, 49 L. Ed. 2d 511, 96 S. Ct. 2513 (1976). But the court concluded that in this context the only permissible state goals were the safety of the PC inmates and the four "legitimate penological objectives" it had already recognized. Again the court determined that "it is not sufficient for defendants to demonstrate that equal treatment [of general population and PC inmates] would be fiscally difficult or impossible." 423 F. Supp. at 1266.
Having stated the test it would apply, the district court then examined the plaintiffs' grievances with meticulous care. Some, the court found, were not proper grounds for relief. The security screen, the court noted, was installed in preparation for a shifting of disciplinary inmates to the fourth floor. The temporary discomfort suffered by PC inmates was not actionable because the screen was installed pursuant to a rational overall penological plan. Medical service in the Annex, the court found, was not worse than that offered the general population, and there was no evidence of inadequate care. The court also found no real demand for more group worship in the Annex, noting that the prison put no obstacle to worship or religious consultation in the path of PC prisoners. This disposed of the plaintiffs' freedom of religion claim.
The court also found that many aspects of the state prison's policies did not measure up to its penological purpose test. The practice of serving meals an hour earlier than the general population's meals failed the test, as did the serving of meals in the inmates' cells. The court ordered that one meal a day be communal and that lunch and supper be served no earlier than 11:00 a.m. and 4:00 p.m., respectively. Second helpings and daily showers were ordered. Limitations on tier time were struck down; PC inmates, the court held, must be given as much time out of their cells as other prisoners. The limited opportunities for outdoor exercise also failed the test; the court ordered that PC inmates have the same amount of outdoor recreation as the general population and in any event no less than an hour a day. The prison administration was also required to give the PC inmates work assignments equivalent to those available to the general population. These opportunities were to include work in such light industries as furniture and business machine repair, work that is available to the general population. The court stated:
"I do not mean that every program or job opportunity must be duplicated in the Annex. If the chance to learn a useful skill either through a job or a program is presented to the general population, a similar opportunity must be given plaintiffs." 423 F. Supp. at 1270.
In the absence of work, the court ordered PC inmates must be given idle pay, although the court recognized that the prison would then be free to discontinue the practice of providing tobacco and other articles free. All of these orders resulted from the application of both the "third" Eighth Amendment standard and the court's version of equal protection for prisoners. In addition, the court found that the practice of permitting general population and PC prisoners to share visiting quarters violated the Eighth Amendment. In the district court's view, this practice, coupled with the harrowing trip through the main cell block, exposed PC prisoners to danger without penological justification. The court therefore ordered that PC inmates be given a separate and secure time and place for visits, so as to minimize the danger to them. Finally, the court found the restrictions on library use infringed the plaintiffs' right of access to the courts.
We are not without sympathy for the approach pursued by the district court. On occasion we have applied a strict test in analyzing the process due a prisoner in a disciplinary hearing. See, e.g., Palmigiano v. Baxter, 487 F.2d 1280 (1st Cir. 1973), vacated, 418 U.S. 908, 94 S. Ct. 3200, 41 L. Ed. 2d 1155 (1974), opinion on remand, 510 F.2d 534 (1974), reversed, 425 U.S. 308, 96 S. Ct. 1551, 47 L. Ed. 2d 810 (1976). But we are constrained to say that at the present stage of development of the law relating to prisoners, the test used by the district court is not required by the Constitution. No majority opinion of the Supreme Court has ever adopted it; and the language in dissents and concurrences provides little support for it. For example, in Gregg v. Georgia, 428 U.S. 153, 49 L. Ed. 2d 859, 96 S. Ct. 2909 (1976), the plurality opinion noted that a penalty might be cruel and unusual if it is "so totally without penological justification that it results in the gratuitous infliction of suffering." Id. at 183. But this standard is a deferential one; the opinion later suggests that the legislative judgment with respect to penological effectiveness must prevail unless clearly wrong. Id. at 186. See also Rudolph v. Alabama, 375 U.S. 889, 891, 11 L. Ed. 2d 119, 84 S. Ct. 155 (1963) (Goldberg, J., dissenting) (penological effectiveness is a "relevant question" in death penalty cases); Estelle v. Gamble, supra at 103; Ingraham v. Wright, supra, 97 S. Ct. at 1412 ("After incarceration, only the 'unnecessary and wanton infliction of pain' . . . constitutes cruel and unusual punishment".) (dictum).2
Nor have the lower courts relied on the penological purpose test as a strong support for their remedial actions.3 A leading case, Jordan v. Fitzharris, supra, did mention that a penalty may be cruel and unusual if it goes beyond what is necessary to achieve its penal purpose. But the Jordan court did not stand squarely for the sort of analysis used below; it found that a more traditional standard had been violated, that being forced to sleep in the filth and vomit of past prisoners without light or ventilation was a barbaric punishment. Other courts have followed a similar pattern, touching briefly on the penological purpose test but resting their holdings of cruel and unusual punishment on indisputably barbaric and indecent conditions at the prison in question. See, e.g., Pugh v. Locke, 406 F. Supp. 318 (M.D. Ala. 1976) (filth, overcrowding, and rampant violence); Landman v. Royster, 333 F. Supp. 621 (E.D. Va. 1971) (prisoners left nude in unheated cells with windows open in winter); Hamilton v. Schiro, 338 F. Supp. 1016 (E.D. La. 1970) (no fire alarms or fire escapes in building so ill-heated that prisoners in overcrowded cells burn newspapers for warmth); Hancock v. Avery, 301 F. Supp. 786 (M.D. Tenn. 1969) (prisoners must sleep without blankets or clothes on concrete floor); Sinclair v. Henderson, 331 F. Supp. 1123 (E.D. La. 1971) (prisoners allowed out of cells for only 15 minutes a day to shower, wash clothes and exercise); see also Inmates, D.C. Jail v. Jackson, 416 F. Supp. 119 (D. D.C. 1976). Still other courts have accepted the test, but found no violation. See Williams v. Field, 416 F.2d 483 (9th Cir. 1969); Padgett v. Stein, 406 F. Supp. 287 (M.D. Pa 1975); Hoitt v. Vitek, 361 F. Supp. 1238 (D. N.H. 1973). Hoitt was affirmed by this circuit without reference to the penal goals test. See Hoitt v. Vitek, 497 F.2d 598 (1st Cir. 1974); Laaman v. Vitek, 502 F.2d 1158 (1st Cir. 1973).
None of these cases has elaborated on the meaning of the test. Nonetheless, the general understanding of the lower courts, like that of the Gregg plurality, is that the courts applying the test should give great weight to legislative and administrative judgments. See Dearman v. Woodson, 429 F.2d 1288, 1290 (10th Cir. 1970) (question is whether treatment of prisoners was "clear abuse or caprice"); Landman v. Royster, supra at 647 (forbidding "arbitrary and capricious action"); Beishir v. Swenson, 331 F. Supp. 1227, 1239 (W.D. Mo. 1971) ("conditions . . . [need only be] reasonably related to the security requirements of the institution"); Allen v. Nelson, 354 F. Supp. 505, 511-12 (N.D. Cal.), affirmed, 484 F.2d 960 (9th Cir. 1973). But see Craig v. Hocker, 405 F. Supp. 656 (D. Nev. 1975).
In contexts other than the "penological purpose" test, courts have considered the claim that nondisciplinary inmates, confined to high security segregated units for their own protection, have a right to the privileges afforded the general population. The prisoners seldom have won these cases. The Fourth Circuit has squarely rejected this claim twice. See Breeden v. Jackson, 457 F.2d 578 (1972); Sweet v. Department of Corrections, 529 F.2d 854 (1975) (en banc). Other courts have agreed. Smith v. Swenson, 333 F. Supp. 1253 (W.D. Mo. 1971); Joyner v. McClellan, 396 F. Supp. 912 (D. Md. 1975); Hundley v. Sielaff, 407 F. Supp. 543 (N.D. Ill. 1975); Krist v. Smith, 309 F. Supp. 497 (S.D. Ga. 1970) (dicta); affirmed, 439 F.2d 146 (5th Cir. 1971); cf. Taylor v. Strickland, 411 F. Supp. 1390 (D. S.C. 1976). See generally O'Brien v. Moriarty, 489 F.2d 941 (1st Cir. 1974) (leaving issue open).
The state's power to draw a distinction between PC prisoners and the general population is, as we discuss below, always subject to the constitutional requirement that the distinction be rational rather than arbitrary and capricious. See James v. Wallace, 382 F. Supp. 1177 (M.D. Ala. 1974) (complaint states a claim by charging that some prisoners are arbitrarily and capriciously denied an opportunity for rehabilitation); United States ex rel. Wakeley v. Pennsylvania, 247 F. Supp. 7, 14 (E.D. Pa. 1965) (claim that privileges are granted to most inmates but denied to plaintiff states claim under constitution). Plaintiffs point to Battle v. Anderson, 376 F. Supp. 402 (E.D. Okla. 1974) as authority favoring them. A short and unexplained portion of a very long order in that case does require that nondisciplinary segregated prisoners be given privileges as close to those of the general population as possible. Given the lack of rationale behind this order, Battle is not compelling authority. In any event it does not support the relief ordered below. Nothing in Battle suggested that conditions were to be equalized no matter the cost; relief was clearly limited by what the nature of protective custody allowed.
To summarize, we are not persuaded that courts' occasional references to the penological purposes served by various punishments establishes a constitutionally based test for reviewing prison practices. Many of the cases mustered in support of such a test are best understood as responding to genuinely barbarous prison conditions. Others merely state the test in dicta, or in dissents. Although a scattered case or two may order changes to equalize the lot of protective custody and general population prisoners, the great majority of courts refuse to intervene in prison affairs on this basis.
Even if we felt at liberty to require that what we consider to be substantial differences in treatment between PC and general population inmates be justified by a legitimate penological purpose, we would, despite the appeal of the proposition, hesitate to do so. For at least as of today, penology is far from an exact science or even a discipline enjoying wide doctrinal consensus. It is, rather, a field where both principles and experience are being reexamined, where experimentation in widely differing approaches is being advocated, and where vigorous debate and political controversy flourish. There is not only a high risk, but a likelihood that the legitimate penological purpose test, as applied below, would simply substitute the values and judgment of a court for the values and judgment of the legislature and prison administration. The test proposed by the court below would subject to intense scrutiny every difference between the treatment of PC inmates and that of the general population. In the end, it could prove counterproductive. The test protects only against the unequal treatment of general population and PC inmates. Presumably, then, the state could meet the test's requirements as easily by abolishing the privileges of the general population as by conferring those privileges on the PC prisoners.
Similarly, the court's refusal to consider administrative convenience or expense in determining the validity of prison policy reflects too austere a standard. True enough, the denial of a fundamental right or the use of a suspect category cannot be justified by reference to cost or convenience. See, e.g., Shapiro v. Thompson, 394 U.S. 618, 22 L. Ed. 2d 600, 89 S. Ct. 1322 (1969). But this case presents neither evil. Adopting the district court's test would not only subject a prison system to the exigencies of seeking unplanned, supplemental funding from a resistant legislature - a confrontation which should be reserved for the gravest of issues - but there is no probability that the net result would be a better prison system. Administrators and legislators would be liable to attack if they adopted an experimental program open to only a few inmates or a strategy of piecemeal improvement or replacement of facilities, for those who did not receive immediate improvement in their treatment would have a constitutional claim to equal treatment. We do not believe that anything in either the Fourteenth or Eighth Amendment requires so strict a test. See Breeden v. Jackson, supra at 578; Sweet v. Department of Corrections, supra.
The Eighth Amendment is of course not the only specific constitutional guarantee that applies in prison. Another guarantee of especial importance to prisoners is the right of access to the courts. See, e.g., Johnson v. Avery, 393 U.S. 483, 21 L. Ed. 2d 718, 89 S. Ct. 747 (1969). The district court found that the inmates in protective custody were unconstitutionally deprived of full access to the courts by the restrictions on library use. As we have seen, the inmates' access to law books is quite limited, and ordering copies of cases is expensive. The one prison attorney provided by the state is overworked and does not handle the full range of legal problems the inmates face. Books ordered during the week are not always delivered. The court therefore ordered a significant expansion of the PC inmates' library privileges.
Since the lower court's decision, the Supreme Court has considered the problem of state prisoners' access to law libraries. Bounds v. Smith, 430 U.S. 817, 97 S. Ct. 1491, 52 L. Ed. 2d 72 (1977). There the Court held
"that the fundamental constitutional right of access to the courts requires prison authorities to assist inmates in the preparation and filing of meaningful legal papers by providing prisoners with adequate law libraries or adequate assistance from persons trained in the law." Id. at 1498.
Bounds clearly vindicates the district court's broad concern over the PC inmates' access to the library. Whether Bounds fully supports the district court's conclusion that these inmates had been denied meaningful access is a somewhat closer question.
In Bounds, the plaintiffs had demanded a library at every one of North Carolina's 77 state prison facilities, but the lower court determined that meaningful access to the courts was assured by a state plan to establish a smaller number of libraries and to transport prisoners to these libraries for as much as a full day's research. Out of a state prison population near 10,000, only 350 inmates would be able to use the libraries each week. Smith v. Bounds, 538 F.2d 541, 542-43 (4th Cir. 1975). It was expected that some inmates might wait three or four weeks for a turn at the library. Although the issue was not directly presented to it, Bounds v. Smith, supra, 97 S. Ct. at 1494 n. 7, the Supreme Court seems to have agreed that this arrangement would not unduly restrict prisoner access to the library. In fact, the Court emphasized that providing libraries was only one way of meeting the state's obligation to assure inmates access to the courts; states were encouraged to experiment with other methods, such as student or inmate paraprofessionals and staff attorneys. The Court noted that the state plan for assuring meaningful access to the courts should be "evaluated as a whole". Id. at 1500. When viewed as a whole, New Hampshire's plan does not seem to give PC inmates significantly less access to the courts than North Carolina's. Although New Hampshire's PC inmates may have fewer opportunities for sustained research, the New Hampshire inmates never face a three or four week waiting period. In addition, New Hampshire provides its inmates with a staff attorney to handle some of their legal affairs.
If this were the end of analysis, we would be reluctant to drive a constitutional line between the two states' programs. But the Bounds Court identified another element relevant to this case: the cost of providing access. While the Court rejected the argument that cost might justify total denial of access to the courts, "this is not to say that economic factors may not be considered, for example in choosing the methods used to provide meaningful access." Id. at 1496. If the cost of providing access may on occasion be a shield in the hands of the state, we think it may also sometimes serve the prisoners as a sword. North Carolina could not have given its prisoners more library access without large expenditures, but it would cost New Hampshire little or nothing to give PC inmates greater freedom to use its library, which is down three or four flights of stairs from the PC cells.4
When library access is as limited as it is for these prisoners, and when it could be increased at little or no cost to the state, we think the inmates' constitutional rights justify an expanded library schedule. The real limitation on the access rights of the PC inmates is the similar rights of the general population, for the two groups cannot use the library at the same time. But the district court recognized this limit and framed its order accordingly.5 We find, therefore, no error in the district court's disposition of this issue.
The other relief granted by the court is not so easily reviewed, for we do not have enough basis to separate out the remedial orders that are supportable by more conventional analysis and those that are not. Because of this ambiguity, examples of which follow, we conclude that a remand is the best course.
As a matter of prison policy, the changes ordered by the district court strike us as wise. The plaintiffs' expert witness persuasively condemned the odd meal pattern, the procedure for seeing visitors, the severe restrictions on showers and outdoor exercise, and the limited rehabilitative and work opportunities. Although the district court stayed its order in large part, the prison has instituted a number of improvements in the life of PC inmates.6 In addition, the prison has prepared a plan for moving the PC inmates to a new location that will allow them to live a more normal prison life. Many of the changes the court ordered may last; the prison shows signs of continuing voluntarily the reforms instituted under court pressure. Nonetheless, to the extent reforms are not voluntarily undertaken, it is important that the court's order does not exceed the appropriate constitutional standards.
While we have rejected the "penological purposes" test that dominated the suit below, this does not immunize the state prison from judicial supervision of a more traditional stripe.7 Prison administrators do not have unlimited authority over their charges. Even when they do not create barbaric or indecent conditions of confinement, prison officials are subject to a limited judicial check; their actions cannot be entirely arbitrary and capricious. O'Brien v. Moriarty, 489 F.2d 941, 944 (1st Cir. 1974). But the judgments of prison officials concerning the necessity of particular practices must be given great weight. Id. The proceedings in the district court, dominated as they were by the "penological purposes" test, did not focus on possible administrative or fiscal justifications for the challenged prison practices. The court must now consider such justifications to see whether they constitute a rational basis or are wholly without substance.
Moreover, quite apart from a review for arbitrariness, courts are obligated to enforce the more specific constitutional guarantees. Among these guarantees is the Eighth Amendment as it has traditionally been interpreted. Here, too, the court's reliance on the "penological purpose" test makes it hard for us to apply a traditional Eighth Amendment analysis. The extent to which the district court's order rests on traditional analysis is uncertain. At one point in its opinion, the district court apparently rejects all claims founded on the traditional analysis, saying "I find no evidence of a mode of imprisonment that is so shocking or debased as to violate standards of common decency." 423 F. Supp. at 1261. Yet the court does go on to discuss traditional tests in relation to some of the prisoners' specific complaints. Sometimes the court mentions the traditional test, but expressly passes over it in favor of the novel test that we have already discussed. Id. at 1269 & 1272. At other times, the court makes no precise distinction between the tests. The court does not make clear, for example, whether its orders with respect to visiting and travel through the main cell block rest on traditional or novel Eighth Amendment analysis.8 We might eventually be able to resolve these ambiguities ourselves and, by combing the record, rule on the basis of the traditional Eighth Amendment approach. But this is a complex case, and the district court is far more familiar than we with the large record. Further, the parties may wish to introduce more evidence, particularly because changes in the treatment of PC prisoners have occurred since the record in this case was closed, and it would be unfair to deny the parties this opportunity.
A similar problem is posed by the court's order regarding exercise. The court concluded that limiting PC inmates to two hours a day "tier time" and something less than two hours a week outdoor exercise time posed a threat to plaintiffs' health over the long run. The court held, in reliance on several cases, that the exercise limit was therefore unconstitutional. See Rhem v. Malcolm, 371 F. Supp. 594 (S.D. N.Y.) (50 minutes a week unconstitutional), affirmed in pertinent part, 507 F.2d 333 (2d Cir. 1974); Spain v. Procunier, 408 F. Supp. 534 (N.D. Cal. 1976) (one hour a day indoors unconstitutional); Sinclair v. Henderson, supra (less than fifteen minutes a day unconstitutional). But see Hundley v. Sielaff, 407 F. Supp. 543 (N.D. Ill. 1975) (no regular exercise period constitutional); Jordan v. Arnold, 408 F. Supp. 869 (M.D. Pa. 1976) (two hours a week constitutional). This ruling, however, is not unequivocally founded on the traditional test; the district court relied in part on its conclusion that "defendants have not established a legitimate penological justification for locking plaintiffs in their cells twenty-two hours a day." 423 F. Supp. at 1269. Because the case must be remanded in any event, we think this aspect of the case would also benefit from reconsideration, free of the shadow cast by the penological purpose test. Our conclusion is strengthened by the state of the record on this issue. "Deliberate indifference to serious medical needs of prisoners" is cruel and unusual punishment. Estelle v. Gamble, supra at 104. The district court's finding that the exercise schedule endangered the inmates' health is thus a very important one, but we are not confident that the district court had sufficient evidence to make that finding, for the medical testimony on this point was scanty at best. Accordingly, on remand, if deemed appropriate, more evidence may be taken on the medical effects of the limited exercise available to PC inmates.
The district court's task on remand is twofold. The first step should be to determine the extent to which changed prison policies will continue in effect without the incentive provided by court order. If the prison administration is willing to leave in place many of the changes it has instituted, large parts of the suit will become moot. The second step is to subject those practices that are still in dispute to scrutiny under the Eighth and Fourteenth Amendment standards discussed above. If the practices violate these standards, they may again be enjoined. We emphasize that our rejection of the district court's novel test is not a signal to abandon any judicial role in prison affairs. Our major purpose in this opinion, compelled by the present state of the authorities, has been to indicate that while the court's review should continue to be rigorous, such review must take into account rational positions advanced by the prison authorities. But genuinely barbarous or shocking prison conditions must not be tolerated; nor are PC prisoners to be deprived of privileges enjoyed by others arbitrarily9 or without reason. With this understanding,
The judgment of the district court is affirmed in part and vacated and remanded in part. Each party shall bear its own costs.
* Hon. Jack R. Miller, of the U.S. Court of Customs and Patent Appeals, sitting by designation.
1. Although we use the present tense, we are describing conditions as they existed before the court-ordered changes resulting from this suit.
2. The reference to penological justification in Gregg, supra, and Coker v. Georgia, 433 U.S. 584, 97 S. Ct. 2861, 53 L. Ed. 2d 982, 45 U.S.L.W. 4961, 4963 (1977) in any event arises in a different context than is presented here. Whereas those cases discuss the relationship between a specific crime and the punishment for it, the inquiry here is into the validity of institutional practices not amounting to punishment, which are applied without regard to the specific offense which resulted in incarceration.
3. We put to one side cases dealing with the confinement of persons not yet brought to trial. E.g., Anderson v. Nosser, 438 F.2d 183 (5th Cir. 1971), cert. denied, 409 U.S. 848, 93 S. Ct. 53, 34 L. Ed. 2d 89 (1972); Wilson v. Beame, 380 F. Supp. 1232 (E.D. N.Y. 1974).
4. In response to the district court's decision, the state now permits PC inmates to use the library from 12:45 to 2:45 p.m. two days a week, and on other occasions by special request. So far as the record shows, this rescheduling has not added significantly to the cost of running the prison.
5. "H. Library and Law Library Defendants shall provide protective custody prisoners with as much access to the library and law library as is reasonably necessary for the preparation of legal papers, subject to their safety needs, necessary prison security requirements, and the rights of access to the courts of all other prisoners at the New Hampshire State Prison." 423 F. Supp. at 1275.
6. The guards now make greater efforts to avoid the main cell block when escorting PC prisoners to the administrative section. Indoor and outdoor recreation time has been greatly expanded. The noon meal is now communal, and inmates may take daily showers. The high school equivalency program is now accessible without passing through the main cell block. PC inmates receive idle pay when not working, and job opportunities have expanded slightly. Visiting time for PC inmates no longer coincides with visiting hours for the general population. The state also tells us that other changes are being made, but the extent to which they have been implemented is in dispute.
7. On appeal, but not below, the state argues that the Eleventh Amendment deprives the federal courts of jurisdiction to order relief that involves substantial expense. Because the Eleventh Amendment "partakes" of jurisdiction, Edelman v. Jordan, 415 U.S. 651, 678, 39 L. Ed. 2d 662, 94 S. Ct. 1347 (1974), we consider the state's belated claim, which is groundless in any event. The Eleventh Amendment bars federal courts from awarding damages against the states, but prospective relief, even costly prospective relief, is within the power of the federal courts. Edelman, supra; Milliken v. Bradley, 433 U.S. 267 , 97 S. Ct. 2749, 53 L. Ed. 2d 745, 45 U.S.L.W. 4873, 4879-80 (1977).
8. The court relied on Woodhous v. Virginia, 487 F.2d 889 (4th Cir. 1973), which held that prisoners have a right to be protected from the violence of other inmates. But the Woodhous court measured the officials' duty by a standard of "reasonable care". Id. at 890. The cost and inconvenience of supplying further protection are relevant to the reasonableness of the officials' policies, but the district court focused only on whether the policies could be justified under one of the four legitimate penal purposes it recognized. Thus the reasonableness of the policies has never been accurately weighed.
9. The district court identified, without relying on, a third source of judicial authority: state law. The district court, which is more familiar with state law than we, may consider on remand the extent to which state law will support relief in this case.