dissenting.
The questions presented in this case require us to resolve two conflicting lines of precedent. On the one hand, as the Court stresses, this Court has said that “ ‘all racial classifications reviewable under the Equal Protection Clause must be strictly scrutinized.’” Gratz v. Bollinger, 539 U. S. 244, 270 (2003) (quoting Adarand Constructors, Inc. v. Peña, 515 U. S. 200, 224 (1995); emphasis added). On the other, this Court has no less categorically said that “the [relaxed] standard of review we adopted in Turner [v. Safley, 482 U. S. 78 (1987),] applies to all circumstances in which the needs of prison administration implicate constitutional rights." Washington v. Harper, 494 U. S. 210, 224 (1990) (emphasis added).
Emphasizing the former line of cases, the majority resolves the conflict in favor , of strict scrutiny. I disagree. The Constitution has always demanded less within the prison walls. Time and again, even when faced with constitutional rights no less “fundamental” than the right to be free from state-sponsored racial discrimination, we have deferred to the reasonable judgments of officials experienced in running this Nation’s prisons. There is good reason for such deference in this case. California oversees roughly 160,000 inmates in prisons that have been a breeding ground for some of the most violent prison gangs in America — all of them organized along racial lines. In that atmosphere, California racially segregates a portion of its inmates, in a part of its prisons, for brief periods of up to 60 days, until the State can arrange permanent housing. The majority is concerned with sparing inmates the indignity and stigma of racial discrimination. Ante, at 507-508. California is concerned with their safety and saving their lives. I respectfully dissent.
*525I
To understand this case, one must understand just how limited the policy at issue is. That requires more factual background than the Court’s opinion provides. Petitioner Garrison Johnson is a black inmate in the California Department of Corrections (CDC), currently serving his sentence for murder, robbery, and assault with a deadly weapon. App. 255a-256a, 259a. Johnson began serving his sentence in June 1987 at the California Institution for Men in Chino, California. Id., at 79a, 264a. Since that time he has been transferred to a number of other facilities within the CDC. Id., at 79a-82a.
When an inmate like Johnson is admitted into the California prison system or transferred between the - CDC’s institutions, he is housed initially for a brief period — usually no more than 60 days — in one of California’s prison reception centers for men. Id., at 303a-305a. CDC, Department Operations Manual §61010.3 (2004) (hereinafter CDC Operations Manual), available at http://www.corr.ca.gov/ RegulationsPolicies/PDF/DOM/00_dept_ops_maunal.pdf (all Internet materials as visited Feb. 18, 2005, and available in Clerk of Court’s case file). In 2003, the centers processed more than 40,000 newly admitted inmates, almost 72,000 inmates returned from parole, over 14,000 inmates admitted for other reasons, and some portion of the 254,000 inmates who were transferred from one prison to another. CDC, Movement of Prison Population 3 (2003).
At the reception center, prison officials have limited information about an inmate, “particularly if he has never been housed in any CDC facility.” App. 303a. The inmate therefore is classified so that prison officials can place the inmate in appropriate permanent housing. During this process, the CDC evaluates the inmate’s “physical, mental and emotional health.” Ibid. The CDC also reviews the inmate’s criminal *526history and record in jail to assess his security needs and classification level. Id., at 304a. Finally, the CBC investigates whether the inmate has any enemies in prison. Ibid. This process determines the inmate’s ultimate housing placement and has nothing to do with race.
While the process is underway, the CDC houses the inmate in a one-person cell, a two-person cell, or a dormitory. Id., at 305a. The few single cells available at reception centers are reserved for inmates who present special security problems, including those convicted of especially heinous crimes or those in need of protective custody. See, e.g., CDC Operations Manual §61010.11.3. At the other end of the spectrum, lower risk inmates are assigned to dormitories. App. 189a-190a. Placement in either a single cell or a dormitory has nothing to do with race, except that prison officials attempt to maintain a racial balance within each dormitory. Id., at 250a. Inmates placed in single cells or dormitories lead fully integrated lives: The CBC does not distinguish based on race at any of its facilities when it comes to jobs, meals, yard and recreation time, or vocational and educational assignments. Ibid.
Yet some prisoners, like Johnson, neither require confinement in a single cell nor may be safely housed in a dormitory. The CBC houses these prisoners in double cells during the 60-day period. In pairing cellmates, race is indisputably the predominant factor. Id., at 305a, 309a. California’s reason is simple: Its prisons are dominated by violent gangs. Brief for Respondents 1-5. And as the largest gangs’ names indicate — the Aryan Brotherhood, the Black Guerrilla Family, the Mexican Mafia, the Nazi Low Riders, and La Nuestra Familia — they are organized along racial lines. See Part II-B, infra.
According to the State, housing inmates in double cells without regard to race threatens not only prison discipline, but also the physical safety of inmates and staff. App. 305a-306a, 310a-311a. That is because double cells are es*527pecially dangerous. The risk of racial violence in public areas of prisons is high, and the tightly confined, private conditions of cells hazard even more violence. Prison staff cannot see into the cells without going up to them, and inmates can cover the windows to prevent the staff from seeing inside the cells. Id., at 306a. The risk of violence caused by this privacy is grave, for inmates are confined to their cells for much of the day. Ibid.; id., at 187a-188a.
Nevertheless, while race is the predominant factor in pairing cellmates, it is hardly the only one. After dividing this subset of inmates based on race, the CDC further divides them based on geographic or national origin. As an example, Hispanics from northern and southern California are not housed together in reception centers because they often belong to rival gangs — La Nuestra Familia and the Mexican Mafia, respectively. Id., at 185a. Likewise, Chinese and Japanese inmates are not housed together, nor are Cambodians, Filipinos, Laotians, or Vietnamese. Id., at 189a. In addition to geographic and national origin, prison officials consider a host of other factors, including inmates’ age, mental health, medical needs, criminal history, and gang affiliation. Id., at 304a, 309a. For instance, when Johnson was admitted in 1987, he was a member of the Crips, a black street gang. Id., at 93a. He was therefore ineligible to be housed with nonblack inmates. Id., at 183a; Brief for Respondents 12, n. 9.
Moreover, while prison officials consider race in assigning inmates to double cells, the record shows that inmates are not necessarily housed with other inmates of the same race during that 60-day period. When a Hispanic inmate affiliated with the Crips asked to be housed at the reception center with a black inmate, for example, prison administrators granted his request. App. 183a-184a, 199a. Such requests are routinely granted after the 60-day period, when prison officials complete the classification process and transfer an *528inmate from the reception center to a permanent placement at that prison or another one.1 Id., at 311a-312a.
II
Traditionally, federal courts rarely involved themselves in the administration of state prisons, “adopting] a broad hands-off attitude toward problems of prison administration.”2 Procunier v. Martinez, 416 U. S. 396, 404 (1974). For most of this Nation’s history, only law-abiding citizens could claim the cover of the Constitution: Upon conviction and incarceration, defendants forfeited their constitutional rights and possessed instead only those rights that the State chose to extend them. See, e. g., Shaw v. Murphy, 532 U. S. 223, 228 (2001); Ruffin v. Commonwealth, 62 Va. 790, 796 (1871). In recent decades, however, this Court has decided *529that incarceration does not divest prisoners of all constitutional protections. See, e. g., Wolff v. McDonnell, 418 U. S. 539, 555-556 (1974) (the right to due process); Cruz v. Beto, 405 U. S. 319, 322 (1972) (per curiam) (the right to free exercise of religion).3
At the same time, this Court quickly recognized that the extension of the Constitution’s demands behind prison walls had to accommodate the needs of prison administration. This Court reached that accommodation in Turner v. Safley, 482 U. S. 78 (1987), which “adopted a unitary, deferential standard for reviewing prisoners’ constitutional claims,” Shaw, supra, at 229. That standard should govern Johm son’s claims, as it has governed a host of other claims challenging conditions of confinement, even when restricting the rights at issue would otherwise have occasioned strict scrutiny. Under the Turner standard, the CDC’s policy passes constitutional muster because it is reasonably related to legitimate penological interests.
A
Well before Turner, this Court recognized that experienced prison administrators, and not judges, are in the best position to supervise the daily operations of prisons across this country. See, e. g., Jones v. North Carolina Prisoners' Labor Union, Inc., 433 U. S. 119, 125 (1977) (courts must give “appropriate deference to the decisions of prison administrators”); Procunier, supra, at 405 (“[C]ourts are ill equipped to deal with the increasingly urgent problems of prison ad*530ministration and reform”). Turner made clear that a deferential standard of review would apply across the board to inmates’ constitutional challenges to prison policies.
At issue in Turner was the constitutionality of a pair of Missouri prison regulations limiting inmate-to-inmate correspondence and inmate marriages. The Court’s analysis proceeded in two steps. First, the Court recognized that prisoners are not entirely without constitutional rights. As proof, it listed certain constitutional rights retained by prisoners, including the right to be “protected against invidious racial discrimination . . . , Lee v. Washington, 390 U. S. 333 (1968).” Turner, 482 U. S., at 84. Second, the Court concluded that for prison administrators rather than courts to “ ‘make the difficult judgments concerning institutional operations,”’ id., at 89 (quoting Jones, supra, at 128), courts should uphold prison regulations that impinge on those constitutional rights if they reasonably relate to legitimate pe-nological interests, 482 U. S., at 89. Nowhere did the Court suggest that Lee's right to be free from racial discrimination was immune from Turner’s deferential standard of review. To the contrary, “[w]e made quite clear that the standard of review we adopted in Turner applies to all circumstances in which the needs of prison administration implicate constitutional rights.” Harper, 494 U. S., at 224 (emphasis added).
Consistent with that understanding, this Court has applied Turner’s standard to a host of constitutional claims by prisoners, regardless of the standard of review that would apply outside prison walls.4 And this Court has adhered to *531Turner despite being, urged to adopt different standards of review based on the constitutional provision at issue. See Harper, supra, at 224 (Turner's standard of review “applies] in all cases in which a prisoner asserts that a prison regulation violates the Constitution, not just those in which the. prisoner invokes the First Amendment” (emphasis added)); O’Lone v. Estate of Shabazz, 482 U. S. 342, 353 (1987) (“We take this opportunity to reaffirm our refusal, even where claims are made under the First Amendment, to substitute our judgment on ... difficult and sensitive matters of institutional administration for the determinations of those charged with the formidable task of running a prison” (internal quotation marks and citation omitted; emphasis added)). Our steadfast adherence makes sense: If Turner is our accommodation of the Constitution’s demands to those of prison administration, see supra, at 530, we should apply it uniformly to prisoners’ challenges to their conditions of confinement.
After all, Johnson’s claims, even more than other claims to which we have applied Turner’s test, implicate Turner’s rationale. In fact, in a passage that bears repeating, the Turner Court explained precisely why deference to the judgments of California’s prison officials is necessary:
“Subjecting the day-to-day judgments of prison officials to an inflexible strict scrutiny analysis would seriously hamper their ability to anticipate security problems and to adopt innovative solutions to the intractable problems of prison administration. The rule would also distort the decisionmaking process, for every administrative judgment would be subject to the possibility that some court somewhere would conclude that it had a less restrictive way of solving the problem at hand. Courts inevitably would become the primary arbiters of what constitutes the best solution to every administrative *532problem, thereby unnecessarily perpetuating the involvement of the federal courts in affairs of prison administration.” 482 U. S., at 89 (internal quotation marks and alteration omitted).
The majority’s failure to heed that advice is inexplicable, especially since Turner itself recognized the “growing problem with prison gangs.” Id., at 91. In fact, there is no more “intractable problem” inside America’s prisons than racial violence, which is driven by race-based prison gangs. See, e. g., Dawson v. Delaware, 503 U. S. 159, 172-173, and n. 1 (1992) (Thomas, J., dissenting); Stefanow v. McFadden, 103 F. 3d 1466, 1472 (CA9 1996) (“Anyone familiar with prisons understands the seriousness of the problems caused by prison gangs that are fueled by actively virulent racism and religious bigotry”).
B
The majority decides this case without addressing the problems that racial violence poses for wardens, guards, and inmates throughout the federal and state prison systems. But that is the core of California’s justification for its policy: It maintains that, if it does not racially separate new cellmates thrown together in close confines during their initial admission or transfer, violence will erupt.
The dangers California seeks to prevent are real. See Brief for National Association of Black Law Enforcement Officers, Inc., as Amicus Curiae 12. Controlling prison gangs is the central challenge facing correctional officers and administrators. Carlson, Prison Interventions: Evolving Strategies to Control Security Threat Groups, 5 Corrections Mgmt. Q. 10 (Winter 2001) (hereinafter Carlson). The worst gangs are highly regimented and sophisticated organizations that commit crimes ranging from drug trafficking to theft and murder. Id., at 12; Cal. Dept. of Justice, Division of Law Enforcement, Organized Crime in California Annual Report to the California Legislature 2003, p. 15, available *533at http://caag.state.ca.us/publications/org_crime.pdf. In fact, street gangs are often just an extension of prison gangs, their “ ‘foot soldiers’ ” on the outside. Ibid.; Widens, Structure, Content and the Exigencies of War: American Prison Law After Twenty-Five Years 1962-1987, 37 Am. U. L. Rev. 41, 55-56 (1987). And with gang membership on the rise, the percentage of prisoners affiliated with prison gangs more than doubled in the 1990’s.5
The problem of prison gangs is not unique to California,6 but California has a history like no other. There are at least five major gangs in this country — the Aryan Brotherhood, the Black Guerrilla Family, the Mexican Mafia, La Nuestra Familia, and the Texas Syndicate — all of which originated in California’s prisons.7 Unsurprisingly, then, California has the largest number of gang-related inmates of any correctional system in the country, including the Federal Government. Carlson 16.
As their very names suggest, prison gangs like the Aryan Brotherhood and the Black Guerrilla Family organize themselves along racial lines, and these gangs perpetuate hate and violence. Irwin 182, 184. Interracial murders and as*534saults among inmates perpetrated by these gangs are common.8 And, again, that brutality is particularly severe in California’s prisons. See, e. g., Walker v. Gomez, 370 F. 3d 969, 971 (CA9 2004) (describing “history of significant racial tension and violence” at Calipatria State Prison); id., at 979-980 (Rymer, J., dissenting) (same); App. 297a-299a (describing 2-year span at Pelican Bay Prison, during which there were no fewer than nine major riots that left at least one inmate dead and many more wounded).
C
It is against this backdrop of pervasive racial violence that California racially segregates inmates in the reception centers’ double cells, for brief periods of up to 60 days, until such time as the State can assign permanent housing. Viewed in that context and in light of the four factors enunciated in Turner, California’s policy is constitutional: The CDC’s policy is reasonably related to a legitimate penological interest; alternative means of exercising the restricted right remain open to inmates; racially integrating double cells might negatively impact prison inmates, staff, and administrators; and there are no obvious, easy alternatives to the CDC’s policy.
1
First, the policy is reasonably related to a legitimate penological interest. Turner, supra, at 89. The protection of inmates and staff is undeniably a legitimate penological interest. See Bell v. Wolfish, 441 U. S. 520, 546-547 (1979). *535The evidence shows, and Johnson has never contested, that the objective of California’s policy is reducing violence among the inmates and against the staff. No cells are designated for, nor are special privileges afforded to, any racial group. App. 188a, 305a. Because prison administrators use race as a factor in making initial housing assignments “solely on the basis of [its] potential implications for prison security,” the CDC’s cell assignment practice is neutral. Thornburgh v. Abbott, 490 U. S. 401, 415 (1989); Turner, 482 U. S., at 90.
California’s policy bears a valid, rational connection to this interest. The racial component to prison violence is impossible for prison administrators to ignore. Johnson himself testified that he is afraid of violence — based solely on the color of his skin.9 In combating that violence, an inmate’s arrival or transfer into a new prison setting is a critical time for inmate and staff alike. The policy protects an inmate from other prisoners, and they from him, while prison officials gather more information, including his gang affiliation, about his compatibility with other inmates. App. 249a. This connection between racial violence and the policy makes it far from “arbitrary or irrational.” Turner, supra, at 89-90.
Indeed, Johnson concedes that it would be perfectly constitutional for California to take account of race “as part of an overall analysis of proclivity to violence based upon a series of facts existing in that prison.” Tr. of Oral Arg. 15. But that is precisely what California does. It takes into account a host of factors in addition to race: geographic or national *536origin, age, physical size, mental health, medical needs, criminal history, and, of course, gang affiliation. Supra, at 527. California does not simply assign inmates to double cells in the reception centers based on race — it also separates intra-racially (for example, northern from southern Hispanies or violent from nonviolent offenders).
2
Second, alternative means of exercising the restricted right remain open to inmates like Johnson. Turner, supra, at 90. The CDC submits, and Johnson does not contest, that all other facets of prison life are fully integrated: work, vocational, and educational assignments; dining halls; and exercise yards and recreational facilities. App. 250a. And after a brief detention period at the reception center, inmates may select their own cellmates regardless of race in the absence of overriding security concerns. Id., at 311a-312a. Simply put, Johnson has spent, and will continue to spend, the vast bulk of his sentence free from any limitation on the race of his cellmate.
3
Third, Johnson fails to establish that the accommodation he seeks — i e., assigning inmates to double cells without regard to race — would not significantly impact prison personnel, other inmates, and the allocation of prison resources. Harper, 494 U. S., at 226-227; Turner, supra, at 90. Prison staff cannot see into the double cells without going up to them, and inmates can cover the windows so that staff cannot see inside the cells at all. App. 306a. Because of the limited number of staff to oversee the many cells, it “would be very difficult to assist inmates if the staff were needed in several places at one time.” Ibid. Coordinated gang attacks against nongang cellmates could leave prison officials unable to respond effectively. In any event, diverting prison resources to monitor cells disrupts services elsewhere.
*537Then, too, fights in the cells are likely to spill over to the exercise yards and common areas. Ibid.; see also id., at 187a. As Turner made clear: “When accommodation of an asserted right will have a significant ‘ripple effect’ on fellow inmates or on prison staff, courts should be particularly deferential to the informed discretion of corrections officials.” 482 U. S., at 90; see also White v. Morris, 832 F. Supp. 1129, 1130 (SD Ohio 1993) (racially integrated double-celling contributed to a race riot in which 10 people were murdered). California prison officials are united in the view that racially integrating double cells in the reception centers would lead to serious violence.10 This is precisely the sort of testimony that the Court found persuasive in Turner itself. 482 U. S., at 92.
4
Finally, Johnson has not shown that there are “obvious, easy alternatives” to the CDC’s policy. Id., at 90. Johnson contends that, for newly admitted inmates, prison officials need only look to the information available in the presen-tence report that must accompany a convict to prison. See Cal. Penal Code Ann. § 1203(c) (West 2004); Cal. Rules of Ct., Crim., Rule 4.411(d) (West Supp. 2004). But prison officials already do this to the extent that they can. Indeed, gang affiliation, not race, is the first factor in determining initial housing assignments. App. 315a. Race becomes the predominant factor only because gang affiliation is often not known, especially with regard to newly admitted inmates. As the Court of Appeals pointed out: “There is little chance *538that inmates will be forthcoming about their past violent episodes or criminal gang activity so as to provide an accurate and dependable picture of the inmate.” 321 F. 3d 791, 806 (CA9 2003); see also App. 185a, 189a. Even if the CDC had the manpower and resources to prescreen the more than 40,000 new inmates it receives yearly, leafing through pre-sentence reports would not tell prison officials what they need to know. See ante, at 521-523 (Stevens, J., dissenting).
Johnson presents a closer case with regard to the segregation of prisoners whom the CDC transfers between facilities. As I understand it, California has less need to segregate prisoners about whom it already knows a great deal (since they have undergone the initial classification process and been housed for some period of time). However, this does not inevitably mean that racially integrating transferred inmates, while obvious and easy, is a true alternative. For instance, an inmate may have affiliated with a gang since the CDC’s last official assessment, or his past lack of racial violence may have been due to the absence of close confinement with members of other races. The CDC’s policy , does not appear to arise from laziness or neglect; California is a leader in institutional intelligence gathering. See Carlson 16 (“The CDC devotes 75 intelligence staff to gathering and verifying inmate-related information,” both in prisons and on the streets). In short, applying the policy to transfers is not “arbitrary or irrational,” requiring that we set aside the considered contrary judgment of prison administrators. Turner, supra, at 89-90.
Ill
The majority claims that strict scrutiny is the applicable standard of review based on this Court’s precedents and its general skepticism of racial classifications. It is wrong on both scores.
A
Only once before, in Lee v. Washington, 390 U. S. 333 (1968) (per curiam), has this Court considered the constitu*539tionality of racial classifications in prisons. The majority claims that Lee applied “a heightened standard of review.” Ante, at 506. But Lee did not address the applicable standard of review. And even if it bore on the standard of review, Lee would support the State here.
In Lee, a three-judge District Court ordered Alabama to desegregate its prisons under Brown v. Board of Education, 347 U. S. 483 (1954). Washington v. Lee, 263 F. Supp. 327, 331-332 (MD Ala. 1966). In so doing, the District Court rejected any notion that “consideration^] of prison security or discipline” justified the “complete and permanent segregation of the races in all the Alabama penal facilities.” Id., at 331. However, the District Court noted “that in some isolated instances prison security and discipline necessitates segregation of the races for a limited period.” Ibid, (footnote omitted). It provided only one example — “the ‘tank’ used in . . . large municipal jails where intoxicated persons are placed upon their initial incarceration and kept until they become sober,” id., at 331, n. 6 — and the court left unmentioned why it would have-been necessary to separate drunk whites from blacks on a Birmingham Saturday night.
This Court, in a per curiam, one-paragraph opinion, affirmed the District Court’s order. It found “unexceptionable” not only the District Court’s general rule that wholesale segregation of penal facilities was unconstitutional, but also the District Court’s “allowance for the necessities of prison security and discipline.” Lee, 390 U. S., at 334. Indeed, Justices Black, Harlan, and Stewart concurred
“to make explicit something that is left to be gathered only by implication from the Court’s opinion. This is that prison authorities have the right, acting in good faith and in particularized circumstances, to take into account racial tensions in maintaining security, discipline, and good order in prisons and jails.” Ibid.
*540Those Justices were “unwilling to assume” that such an “explicit pronouncement [would] evinc[e] any dilution of this Court’s firm commitment to the Fourteenth Amendment’s prohibition of racial discrimination.” Ibid.
Lee said nothing about the applicable standard of review, for there was no need. Surely Alabama’s wholesale segregation of its prisons was unconstitutional even under the more deferential standard of review, that applies within prisons. This Court’s brief, per curiam opinion in Lee simply cannot bear the weight or interpretation the majority places on it. See U. S. Bancorp Mortgage Co. v. Bonner Mall Partnership, 513 U. S. 18, 24 (1994) (noting “our customary skepticism toward per curiam dispositions that lack the reasoned consideration of a full opinion”); Edelman v. Jordan, 415 U. S. 651, 670-671 (1974).
Yet even if Lee had announced a heightened standard of review for prison policies that pertain to race, Lee also carved out an exception to the standard that California’s policy would certainly satisfy. As the Lee concurrence explained without objection, the Court’s exception for “the necessities of prison security and discipline” meant that “prison authorities háve the right, acting in good faith and in particularized circumstances, to take into account racial tensions in maintaining security, discipline, and good order in prisons and jails.” 390 U. S., at 334 (opinion of Black, Harlan, and Stewart, JJ., concurring) (emphasis added).
California’s policy — which is a far cry from the wholesale segregation at issue in Lee — would fall squarely within Lee’s exception. Johnson has never argued that California’s policy is motivated by anything other than a desire to protect inmates and staff. And the “particularized” nature of the policy is evident: It applies only to new inmates and transfers, only in a handful of prisons, only to double cells, and only then for a period of no more than two months. In the name of following a test that Lee did not create, the majority *541opts for a more demanding standard of. review than Lee's language even arguably supports.
The majority heavily relies on this Court’s statement that “ ‘all racial classifications [imposed by government]... must be analyzed by a reviewing court under strict scrutiny.’” Ante, at 505 (emphasis deleted) (quoting Adarand Constrctors, Inc., 515 U. S., at 227). Adamnd has nothing to do with this case. Adamnd's statement that “all racial classifications” are subject to strict scrutiny addréssed the contention that classifications favoring rather than disfavoring blacks are exempt. Id., at 226-227; accord, Grutter v. Bollinger, 539 U. S. 306, 353 (2003) (Thomas, J., concurring in part and dissenting in part). None of these statements overruled, sub silentio, Turner and its progeny, especially since the Court has repeatedly held that constitutional demands are diminished in the unique context of prisons. See, e. g., Harper, 494 U. S., at 224; Abbott, 490 U. S., at 407; Turner, 482 U. S., at 85; see also Webster v. Fall, 266 U. S. 507, 511 (1925) (“Questions which merely lurk in the record, neither brought to the attention of the court nor ruled upon, are not to be considered as having been so decided as to constitute precedents”).
B
The majority offers various other reasons for applying strict scrutiny. None is persuasive. The majority’s main reason is that “Turner’s, reasonable-relationship test [applies] only to rights that are ‘inconsistent with proper incarceration.’” Ante, at 510 (quoting Overton v. Bazzetta, 539 U. S. 126, 131 (2003)). According to the majority, the question is thus whether a right “need necessarily be compromised for the sake of proper prison administration.” Ante, at 510. This inconsistency-with-proper-prison-administration test begs the question at the heart of this case. For a court to know whether any particular right is inconsistent with proper prison administration, it must have some implicit notion of what a proper prison ought to look like and how it *542ought to be administered. Overton, supra, at 139 (Thomas, J., concurring in judgment). But the very issue in this case is whether such second-guessing is permissible.
The majority’s test eviscerates Turner. Inquiring whether a given right is consistent with “proper prison administration” calls for precisely the sort of judgments that Turner said courts were ill equipped to make. In none of the cases in which the Court deferred to the judgments of prison officials under Turner did it examine whether “proper” prison security and discipline permitted greater speech or associational rights (Abbott, supra; Shaw, 532 U. S. 223; and Overton, supra); expanded access to the courts (Lewis v. Casey, 518 U. S. 343 (1996)); broader freedom from bodily restraint (Harper, supra); or additional free exercise rights (O’Lone, 482 U. S. 342). The Court has steadfastly refused to undertake the threshold standard-of-review inquiry that Turner settled, and that the majority today resurrects. And with good reason: As Turner pointed out, these judgments are better left in the first instance to the officials who run our Nation’s prisons, not to the judges who run its courts.
In place of the Court’s usual deference, the majority gives conclusive force to its own guesswork about “proper” prison administration. It hypothesizes that California’s policy might incite, rather than diminish, racial hostility.11 Ante, *543at 506-508. The majority's speculations are implausible. New arrivals have a strong interest in promptly convincing other inmates of their willingness to use violent force. See Brief for National Association of Black Law Enforcement Officers, Inc., as Amicus Curiae 13-14 (citing commentary and congressional findings); cf. United States v. Santiago, 46 F. 3d 885, 888 (CA9 1995) (describing one Hispanic inmate's murder of another in order to join the Mexican Mafia); United States v. Silverstein, 732 F. 2d 1338, 1341 (CA7 1984) (prospective members of the Aryan Brotherhood must “make bones,” or commit a murder, to be eligible for membership). In any event, the majority’s guesswork falls far short of the compelling showing needed to overcome the deference we owe to prison administrators.
The majority contends that the Court “[has] put the burden on state actors to demonstrate that their race-based policies are justified,” ante, at 506, n. 1, and “[has] refused to defer to state officials’ judgments on race in other areas where those officials traditionally exercise substantial discretion,” ante, at 512. Yet two Terms ago, in upholding the University of Michigan Law School’s affirmative-action program, this Court deferred to the judgment by the law school’s faculty and administrators on their need for diversity in the student body. See Grutter, supra, at 328 (“The Law School’s educational judgment that... diversity is essential to its educational mission is one to which we defer”). Deference would seem all the more warranted in the prison context, for whatever the Court knows of administering educational institutions, it knows much less about administering penal ones. The potential consequences of second-guessing the judgments of prison administrators are also much more severe. See White v. Morris, 832 F. Supp. 1129, 1130 (SD Ohio 1993) (racially integrated double-celling that resulted *544from federal consent decree was a factor in the worst prison riot in Ohio history). More importantly, as I have explained, the Court has recognized that the typically exacting review it applies to restrictions on fundamental rights must be relaxed in the unique context of prisons. See, e. g., Harper, 494 U. S., at 224; Abbott, 490 U. S., at 407; Turner, 482 U. S., at 85. The majority cannot fall back on the Constitution’s usual demands, because those demands have always been lessened inside the prison walls. See supra, at 529.
The majority also mentions that California’s policy may be the only one of its kind, as virtually all other States and the Federal Government manage their prison systems without racially segregating inmates. Ante, at 508-509. This is both irrelevant and doubtful. It is irrelevant because the number of States that have followed California’s lead matters not to the applicable standard of review (the only issue the Court today decides), but to whether California satisfies whatever standard applies, a question the majority leaves to be addressed on remand. In other words, the uniqueness of California’s policy might show whether the policy is reasonable or narrowly tailored — but deciding whether to apply Turner or strict scrutiny in the first instance must depend on something else, like the majority’s inconsistency-with-proper-prison-administration test. The commonness of California’s housing policy is further irrelevant because strict scrutiny now applies to all claims of racial discrimination in prisons, regardless of whether the policies being challenged are unusual.
The majority’s assertion is doubtful, because at least two other States apply similar policies to newly admitted inmates. Both Oklahoma and Texas, like California, assign newly admitted inmates to racially segregated cells in their prison reception centers.12 The similarity is not surprising: *545States like California and Texas have historically had the most severe problems with prison gangs. However, even States with less severe problems maintain that policies like California’s are necessary to deal with race-related prison violence. See Brief for States of Utah, Alabama, AJaska, Delaware, Idaho, Nevada, New Hampshire and North Dakota as Amici Curiae 16. Relatedly, 10.3% of all wardens at maximum security facilities in the United States report that their inmates are assigned to racially segregated cells — apr parently on a permanent basis. Henderson, Cullen, Carroll, & Feinberg, Race, Rights, and Order in Prison: A National Survey of Wardens on the Racial Integration of Prison Cells, 80 Prison J. 295, 304 (Sept. 2000). In the same survey, 4.3% of the wardens report that their States have an official policy against racially integrating male inmates in cells. Id., at 302. Presumably, for the remainder of prisons in which inmates are assigned to racially segregated cells, that policy is the result of discretionary decisions by wardens rather than of official state directives. Ibid. In any event, the ongoing debate about the best way to reduce racial violence in prisons should not be resolved by judicial decree: It is the job “of prison administrators ... and not the courts, to make the difficult judgments concerning institutional operations.” Jones, 433 U. S., at 128.
The majority also observes that we have already carved out an exception to Turner for Eighth Amendment claims of cruel and unusual punishment in prison. See Hope v. Pelzer, *546536 U. S. 730, 738 (2002). In that context, we have held that “[a] prison official’s ‘deliberate indifference’ to a substantial risk of serious harm to an inmate violates the Eighth Amendment.” Farmer v. Brennan, 511 U. S. 825, 828 (1994). Setting aside whether claims challenging inmates’ conditions of confinement should be cognizable under the Eighth Amendment at all, see Hudson v. McMillian, 503 U. S. 1, 18-19 (1992) (Thomas, J., dissenting), the “deliberate indifference” standard does not bolster the majority’s argument. If anything, that standard is more deferential to the judgments of prison administrators than Turner's reasonable-relationship test: It subjects prison officials to liability only when they are subjectively aware of the risk to the inmate, and they fail to take reasonable measures to abate the risk. Farmer, supra, at 847. It certainly does not demonstrate the wisdom of an exception that imposes a heightened standard of review , on the actions of prison officials.
Moreover, the majority’s decision subjects prison officials to competing and perhaps conflicting demands. In this case, California prison officials have uniformly averred that random double-celling poses a substantial risk of serious harm to the celled inmates. App. 245a-246a, 251a. If California assigned inmates to double cells without regard to race, knowing full well that violence might result, that would seem the very definition of deliberate indifference. See Robinson v. Prunty, 249 F. 3d 862, 864-865 (CA9 2001) (prisoner alleged an Eighth Amendment violation because administrators had failed to consider race when releasing inmates into the yards); Jensen v. Clarke, 94 F. 3d 1191, 1201, 1204 (CA8 1996) (court held that random double-celling by prison officials constituted deliberate indifference, and affirmed an injunction and attorney’s fees awarded against the officials). Nor would a victimized inmate need to prove that prison officials had anticipated any particular attack; it would be sufficient that prison officials had ignored a dangerous condi*547tion that was chronic and ongoing — like interracial housing in closely confined quarters within prisons dominated by racial gangs. Farmer, supra, at 843-844. Under Farmer, prison officials could have been ordered to take account of the very thing to which they may now have to turn a blind eye: inmates’ race.
Finally, the majority presents a parade of horribles designed to show that applying the Turner standard would grant prison officials unbounded discretion to segregate inmates throughout prisons. See ante, at 513-514. But we have never treated Turner as a blank check to prison officials. Quite to the contrary, this Court has long had “confidence that ... a reasonableness standard is not toothless.” Abbott, 490 U. S., at 414 (internal quotation marks omitted). California prison officials segregate only double cells, because only those cells are particularly difficult to monitor— unlike “dining halls, yards, and general housing areas.” Ante, at 514. Were California’s policy not so narrow, the State might well have race-neutral means at its disposal capable of accommodating prisoners’ rights without sacrificing their safety. See Turner, 482 U. S., at 90-91. The majority does not say why Turner’s standard ably polices all other constitutional infirmities, just not racial discrimination. In any event, it is not the refusal to apply — for the first time ever — a strict standard of review in the prison context that is “fundamentally at odds” with our constitutional jurisprudence. Ante, at 506, n. 1. Instead, it is the majority’s refusal — for the first time ever — to defer to the expert judgment of prison officials.
IV
Even under strict scrutiny analysis, “it is possible, even likely, that prison officials could show that the current policy meets the test.” 336 F. 3d 1117, 1121 (CA9 2003) (Ferguson, J., joined by Pregerson, Nelson, and Reinhardt, JJ., dissenting from denial of rehearing en banc). As Johnson concedes, all States have a compelling interest in *548maintaining order and internal security within their prisons. See Reply Brief for Petitioner 18; see also Procunier, 416 U. S., at 404. Thus the question on remand will be whether the CDC’s policy is narrowly tailored to serve California’s compelling interest.13 The other dissent notes the absence of evidence on that question, see ante, at 518-621 (opinion of Stevens, J.), but that is hardly California’s fault.
. From the outset, Johnson himself has alleged, in terms taken from Turner, that the CDC’s policy is “not related to a legitimate penological interest.” Johnson v. California, 207 F. 3d 650, 655 (CA9 2000) (per curiam) (discussing Johnson’s Third Amended Complaint). In reinstating Johnson’s equal protection claim following the District Court’s dismissal, the Court of Appeals repeated Johnson’s allegation, without indicating that strict scrutiny should apply on remand before the District Court.14 Ibid. And on remand, again Johnson alleged only that the CDC’s policy “is not reasonably related to the legitimate penological interests of the CBC.” App. 51a (Fourth Amended Complaint ¶ 23).
After the District Court granted qualified immunity to some of the defendants, Johnson once again appealed. In his brief before the Court of Appeals, Johnson assumed that *549both Lee and Turner applied, without arguing that there was any tension between them; indeed, nowhere in his brief did Johnson even mention the words “strict scrutiny.” Brief for Appellant in No. 01-56436 (CA9), pp. 20, 26, 2001 WL 34091249. Perhaps as a result, the Court of Appeals did not discuss strict scrutiny in its second' decision, the one currently before this Court. The Court of Appeals did find tension between Lee and Turner; however, it resolved this tension in Turner’s favor. 321 F. 3d, at 799. Yet the Court of Appeals accepted Lee’s test at face value: Prison officials may only make racial classifications “ ‘in good faith and in particularized circumstances.’” 321 F. 3d, at 797. The Court of Appeals, like Johnson, did not equate Lee’s test with strict scrutiny, and in fact it mentioned strict scrutiny only when it quoted the portion of Turner that rejects strict scrutiny as the proper standard of review in the prison context. 321 F. 3d, at 798. Even Johnson did not make the leap equating Lee with strict scrutiny when he requested that the Court of Appeals rehear his case. Appellant’s Petition for Panel Rehearing with Suggestion for Rehearing En Banc in No. 01-56436 (CA9), pp. 4-5. That leap was first made by the judges who dissented from the Court of Appeals’ denial of rehearing en banc. 336 F. 3d, at 1118 (Ferguson, J., joined by Pregerson, Nelson, and Reinhardt, JJ., dissenting from denial of rehearing en banc).
Thus, California is now, after the close of discovery, subject to a more stringent standard than it had any reason to anticipate from Johnson’s pleadings, the Court of Appeals’ initial decision, or even the Court of Appeals’ decision below. In such circumstances, California should be allowed to present evidence of narrow tailoring, evidence it was never obligated to present in either appearance before the District Court. See Lucas v. South Carolina Coastal Council, 505 U. S. 1003, 1031-1032 (1992) (remanding for consideration under the correct legal standard); id., at 1033 (Kennedy, J., concurring in judgment) (“Although we establish a frame*550work for remand,... we do not decide the ultimate [constitutional] question [because] [t]he facts necessary to the determination have not been developed in the record”).
* * *
Petitioner Garrison Johnson challenges not permanent, but temporary, segregation of only a portion of California’s prisons. Of the 17 years Johnson has been incarcerated, California has assigned him a cellmate of the same race for no more than a year (and probably more like four months); Johnson has had black cellmates during the other 16 years, but by his own choice. Nothing in the record demonstrates that if Johnson (or any other prisoner) requested to be housed with a person of a different race, it would be denied (though Johnson’s gang affiliation with the Crips might stand in his way). Moreover, Johnson concedes that California’s prisons are racially violent places, and that he lives in fear of being attacked because of his race. Perhaps on remand the CDC’s policy will survive strict scrutiny, but in the event that it does not, Johnson may well have won a Pyrrhic victory.
Johnson has never requested — not during his initial admittance, nor his subsequent transfers, nor his present incarceration — that he be housed with a person of a different race. App. 106a, 112a-113a, 175a. According to Johnson, he considered the policy a barrier to any such request; however, Johnson has also testified that he never filed a grievance with prison officials about the segregation policy. Id., at 112a-113a, 124a-125a. Neither the parties nor the majority discusses whether Johnson has exhausted his action under Rev. Stat. § 1979, 42 U. S. C. § 1983, as required by the Prison Litigation Reform Act of 1995, 110 Stat. 1321-66, as amended, 42 U. S. C. § 1997e(a). See Booth v. Churner, 532 U. S. 731, 734 (2001). The majority thus assumes that statutorily mandated exhaustion is not jurisdictional, and that California has waived the issue by failing to raise it. See, e. g., Richardson v. Goord, 347 F. 3d 431, 433-434 (CA2 2003); Perez v. Wisconsin Dept. of Corrections, 182 F. 3d 532, 536 (CA7 1999).
The majority refers to my approach as a “hands-off” one because I would accord deference to the judgments of the State’s prison officials. See ante, at 506, n. 1. Its label is historically inaccurate. The “hands-off” approach was that taken prior to the 1960’s by federal courts, which generally declined to consider the merits of prisoners’ claims. See, e. g., J. Fliter, Prisoners’ Rights: The Supreme Court and Evolving Standards of Decency 64-65 (2001); M. Feeley & E. Rubin, Judicial Policy Making and the Modern State 30-34 (2000); S. Krantz & L. Branham, Cases and Materials on the Law of Sentencing, Corrections and Prisoners’ Rights 264-265 (4th ed. 1991).
A prisoner may not entirely surrender his constitutional rights at the prison gates, Bell v. Wolfish, 441 U. S. 520, 545 (1979); Jones v. North Carolina Prisoners’ Labor Union, Inc., 433 U. S. 119, 129 (1977), but certainly he leaves some of his liberties behind him. When a prisoner makes a constitutional claim, the initial question should be whether the prisoner possesses the right at issue at all, or whether instead the prisoner has been divested of the right as a condition of his conviction and confinement. See Overton v. Bazzetta, 539 U. S. 126, 140 (2003) (Thomas, J., concurring in judgment); Coffin v. Reichard, 143 F. 2d 443, 445 (CA6 1944).
See, e. g., Overton, supra, at 132 (the right to association under the First and Fourteenth Amendments); Shaw v. Murphy, 532 U. S. 223, 228-229 (2001) (the right to communicate with fellow inmates under the First Amendment); Lewis v. Casey, 518 U. S. 343, 361 (1996) (the right of access to the courts under the Due Process and Equal Protection Clauses); Washington v. Harper, 494 U. S. 210, 223-225 (1990) (the right to refuse forced medication under the Due Process Clause); Thornburgh v. Abbott, 490 U. S. 401, 413-414 (1989) (the right to receive correspondence under the *531First Amendment); O’Lone v. Estate of Shabazz, 482 U. S. 342, 349-350 (1987) (the right to free exercise of religion under the First Amendment).
See National Gang Crime Research Center, A National Assessment of Gangs and Security Threat Groups (STGs) in Adult Correctional Institutions: Results of the 1999 Adult Corrections Survey, p. 5, http://www. ngcrc.com/ngcrc/page7.htm.
See, e. g., Fraise v. Terhune, 283 F. 3d 506, 512-513 (CA3 2002) (describing violence caused by a single black prison gang, the Five Percent Nation, in various New Jersey correctional facilities); Conroy v. Dingle, No. Civ. 01-1626 (RHK/RLE), 2002 WL 31357055, *1-*2 (D. Minn., Oct. 11, 2002) (describing rival racial gangs at Minnesota’s Moose Lake facility, a medium security prison).
See D. Orlando-Morningstar, Prison Gangs, Special Needs Offenders Bulletin, Federal Judicial Center 4 (Oct. 1997); see also J. Irwin, Prisons 'in Turmoil 189 (1980) (hereinafter Irwin) (describing the establishment and rise of gangs inside the California prison system, first the Mexican Mafia, followed by La Nuestra Familia, the Aryan Brotherhood, and the Black Guerrilla Family); United States v. Shryock, 342 F. 3d 948, 961 (CA9 2003) (detailing rise of Mexican Mafia inside the California prison system).
See, e. g., id., at 962-969 (describing a host of murders and attempted murders by a handful of Mexican Mafia members); United States v. Silverstein, 732 F. 2d 1338, 1341-1342 (CA7 1984) (describing murder of a black inmate by members of the Aryan Brotherhood); State v. Kell, 61 P. 3d 1019, 1024-1025 (Utah 2002) (describing fatal stabbing of a black inmate by two white supremacists); State v. Farmer, 126 Ariz. 569, 570-571, 617 P. 2d 521, 522-523 (1980) (en banc) (describing murder of a black inmate by members and recruits of the Aryan Brotherhood).
Specifically, Johnson testified:
“I was incarcerated at Calipatria before the major riot broke out there with Mexican and black inmates.... If I would have stayed there, I would have been involved in that because you have four facilities there and each facility went on a major riot and a lot of people got hurt and injured just based on your skin color. I’m black, and if I was there I would have been hurt.” App. 102a (emphasis added).
See id., at 245a-246a (Cambra declaration) (“If race were to be disregarded entirely, however, I am certain, based upon my experience with CDC prisoners, that... there will be fights in the cells and the problems will emanate onto the prison yards”); id., at 250a-251a (Schulteis declaration) (“At CSP-Lancaster, if we were to disregard the initial housing placement [according to race], then I am certain there would be serious violence among inmates. I have worked in five different CDC institutions and this would be true for all of them”),
The majority’s sole empirical support for its speculation is a study of Texas prison desegregation that found the rate of violence higher in racially segregated double cells. Ante, at 507-508 (citing Trulson & Mar-quart, The Caged Melting Pot: Toward an Understanding of the Consequences of Desegregation in Prisons, 36 Law & Soc. Rev. 743, 774 (2002)). However, the study’s authors specifically note that Texas — like California — does not integrate its “initial diagnostic facilities” or its “transfer facilities.” See id., at 753, n. 13. Thus the study says nothing about the violence likely to result from integrating cells when inmates are thrown together for brief periods during admittance or transfer. What the study does say is that, once Texas has had the time to gather inmate-related information and make more permanent housing assignments, racially inte*543grated cells may be the preferred option. But California leaves open that door: Inmates are generally free to room with whomever they like on a permanent basis.
See Oklahoma Dept, of Corrections, Policies and Procedures, Operations Memorandum No. OP-030102, Inmate Housing (Sept. 16, 2004) (“Upon arrival at the assessment and reception center ... [f]or reasons of *545safety and security, newly received inmates are not generally assigned randomly to racially integrated cells”) (available at http://www.doc.state. ok.us/docs/policies.htm); Texas Dept, of Criminal Justice, Security Memorandum No. SM-01.28, Assignment to General Population Two-Person Cells (June 15, 2002) (“Upon arrival at a reception and diagnostic center . . . [f]or reasons of safety and security, newly-received offenders are not generally assigned randomly to racially integrated cells due to the fact that the specific information needed to assess an offender’s criminal and victimization history is not available until after diagnostic processing has been completed”).
On the majority's account, deference to the judgments of prison officials in the application of strict scrutiny is presumably warranted to account for “the special circumstances [that prisons] present,” ante, at 515. See Grutter v. Bollinger, 539 U. S. 306, 328 (2003). Although I disagree that deference is normally appropriate when scrutinizing racial classifications, there is some logic to the majority’s qualification in this case because the Constitution’s demands have always been diminished in the prison context. See, e. g., Harper, 494 U. S., at 224; Abbott, 490 U. S., at 407; Turner v. Safley, 482 U. S. 78, 85 (1987).
The Court of Appeals cited both Turner and Lee v. Washington, 390 U. S. 333 (1968) (per curiam), for the proposition that certain constitutional protections, among them the protection against state-sponsored racial discrimination, extend to the prison setting. However, the Court of Appeals did not discuss the applicable standard of review, nor did it attempt to resolve the tension between Turner and Lee that the majority finds.