Opinion for the Court filed by Circuit Judge TATEL.

TATEL, Circuit Judge:

Under Sandin v. Conner, segregative confinement in prison implicates a liberty interest protected by the Due Process Clause of the United States Constitution only if it “imposes atypical and significant hardship on the inmate in relation to the ordinary incidents of prison life.” 515 U.S. 472, 484, 115 S.Ct. 2293, 132 L.Ed.2d 418 (1995). In this case brought by a Lorton inmate claiming a liberty interest in avoiding such confinement, we must define “the ordinary incidents of prison life” — the comparative baseline for determining whether appellant’s segregation was an “atypical and significant hardship.” Considering Sandin’s language and objectives, we hold that due process is required when segregative confinement imposes an “atypical and significant hardship” on an inmate in relation to the most restrictive conditions that prison officials, exercising their administrative authority to ensure institutional safety and good order, routinely impose on inmates serving similar sentences. For appellant, these conditions include the usual conditions of administrative segregation at Lorton. They also include more restrictive conditions at other prisons if it is likely both that inmates serving sentences similar to appellant’s will actually be transferred to such prisons and that once transferred they will actually face such conditions. Because the district court did not apply this standard, we reverse its grant of summary judgment for appellee and remand for further consideration of appellant’s due process claim in light of this opinion.

Appellant Donald Hatch is a District of Columbia convict serving multiple sentences for armed robbery, kidnapping, sodomy, and rape. The events giving rise to *848 this suit occurred while Hatch was an inmate at the Lorton Correctional Complex. Because the district court granted summary judgment for the District, we describe the facts in the light most favorable to Hatch. See Fed.R.Ctv.P. 56(c); DeGraff v. District of Columbia, 120 F.3d 298, 299-300 (D.C.Cir.1997).

On January 5, 1994, while working as head clerk at Lorton’s law library, Hatch got into a fight with another prisoner over the use of a copy machine. Immediately after the, incident, the prison Housing Board, which “determined appropriate housing placement” to ensure prison safety and security, D.C. Mun. Regs. tit. 28, § 522.1 (1987), assigned Hatch to administrative segregation, a form of solitary confinement commonly used to separate disruptive prisoners. In addition, Hatch received a disciplinary report charging him with fighting, lack of cooperation, and creating a disturbance- — all “Class II” offenses under Lorton regulations. See id. §§ 503.1, 503.4, 503.5, 503.11.

On January 11, Hatch appeared before the prison Adjustment Board, which adjudicates charged offenses and imposes disciplinary penalties.. See id. §§ 508-5.15. Due to a mistake in the disciplinary report, the Adjustment Board dismissed all charges. The next day, the Housing Board met to consider Hatch’s confinement. Finding that Hatch posed a threat to the orderly operation of the prison, the Housing Board recommended that he remain in administrative segregation. Hatch had no notice of the Housing Board meeting, did not attend the meeting, and had no opportunity to testify or present evidence.

On January 20, the Adjustment Board, which had previously dismissed the charges against Hatch, met again to consider the same charges. The Adjustment Board denied Hatch’s requests to speak on his own behalf, to cross-examine adverse witnesses, and to call witnesses, including the writer of the disciplinary report. The Board acquitted him of creating a disturbance and lack of cooperation, but found him guilty of fighting. It sentenced him to fourteen days of adjustment segregation, another form of solitary confinement which, unlike administrative segregation, punishes individual inmates for specific, proven acts of misconduct.

On March 21, the Housing Board, as required by Lorton regulations, see id. § 527.1, conducted a sixty-day review of Hatch’s status. Determining that Hatch no longer presented a “management problem,” it recommended that he be returned to the prison’s general population. Supervising officials approved this recommendation in early April, but Hatch remained in segregation until August 11 — more than seven months after his initial placement in segregation. The District offers no' explanation for this delay.. Hatch claims that Lorton officials kept him in segregation because bed space was unavailable in the general population.

Although Hatch’s confinement consisted of two weeks of adjustment segregation and twenty-nine weeks of administrative segregation, the conditions of his confinement remained basically the -same throughout the seven months. Confined to his cell twenty-three and a half hours per day on weekdays and all forty-eight hours of the weekend, Hatch had no outdoor recreation and was not allowed to work or to visit the library, gym, health clinic, psychological services, mailroom, clothing and bedding exchange, or culinary unit. He had no access to a dentist despite four written requests to have a broken, decayed tooth extracted. He had no opportunity to wash his clothes or get a haircut. Whenever he left the cell block, he was transported in handcuffs and leg irons. Prison officials confiscated his legal papers and denied him access to legal telephone calls for ninety days.

On June 24, while still in administrative segregation, Hatch filed suit against the District of Columbia in the United States District Court, alleging that his confine *849 ment in adjustment and administrative segregation violated the Due Process Clause of the U.S. Constitution as well as D.C. regulations governing Lorton. The District moved to dismiss or, alternatively, for summary judgment. After requesting additional briefing on the conditions of Hatch’s confinement, the district court granted summary judgment for the District. See Hatch v. District of Columbia, No. 94-1393 (D.D.C. Oct. 11, 1996) (“Mem. Order”). Applying Sandin v. Conner and assuming Hatch’s description of his confinement to be true, the court determined that he “did not suffer an ‘atypical and significant hardship’” compared to “the typical restrictions imposed on prisoners in the general population.” Mem. Order at 5. It thus concluded that under Sandin, Hatch had no liberty interest in avoiding either adjustment or administrative segregation. See id. at 5-6.

Hatch appeals pro se, aided by court-appointed counsel who filed briefs and argued the case as amicus curiae. Our review is de novo. See Tao v. Freeh, 27 F.3d 635, 638 (D.C.Cir.1994).

Sandin v. Conner represents the culmination of a twenty-year effort by the Supreme Court to clarify when restrictions imposed by prison officials on lawfully incarcerated inmates constitute deprivations of “liberty” within the meaning of the Due Process Clause. Two basic principles have guided the Court’s effort. The first is that prison officials need “broad administrative and discretionary authority over the institutions they manage.” Hewitt v. Helms, 459 U.S. 460, 467, 103 S.Ct. 864, 74 L.Ed.2d 675 (1983). Recognizing the difficulty and complexity of operating safe and effective prisons, as well as the expertise of prison officials, the Supreme Court has repeatedly instructed federal courts “to afford appropriate deference and flexibility to state officials trying to manage a volatile environment.” Sandin, 515 U.S. at 482, 115 S.Ct. 2293 (citing cases); see also Hewitt, 459 U.S. at 470, 103 S.Ct. 864 (“[T]he safe and efficient operation of a prison on a day-to-day basis has traditionally been entrusted to the expertise of prison officials....”); Jones v. North Carolina Prisoners’ Labor Union, Inc., 433 U.S. 119, 125, 97 S.Ct. 2532, 53 L.Ed.2d 629 (1977) (requiring courts to “giv[e] appropriate deference to the decisions of prison administrators and appropriate recognition to the peculiar and restrictive circumstances of penal confinement”). Accordingly, the Supreme Court has refused to “ ‘subject to judicial review a wide spectrum of discretionary actions that traditionally have been the business of prison administrators rather than of the federal courts,’ ” Hewitt, 459 U.S. at 467, 103 S.Ct. 864 (quoting Meachum v. Fano, 427 U.S. 215, 225, 96 S.Ct. 2532, 49 L.Ed.2d 451 (1976)), making clear that the “ ‘withdrawal or limitation of many privileges and rights’ ” of prisoners is “ ‘justified by the considerations underlying our penal system,’ ” id. (quoting Price v. Johnston, 334 U.S. 266, 285, 68 S.Ct. 1049, 92 L.Ed. 1356 (1948)).

While recognizing the need to protect prison administrators’ discretion and flexibility, the Supreme Court has made equally clear a second, countervailing principle: “[T]hough his rights may be diminished by the needs and exigencies of the institutional environment, a prisoner is not wholly stripped of constitutional protections when he is imprisoned for crime.” Wolff v. McDonnell, 418 U.S. 539, 555, 94 S.Ct. 2963, 41 L.Ed.2d 935 (1974). “There is no iron curtain,” Wolff said, “drawn between the Constitution and the prisons of this country.” Id. at 555-56, 94 S.Ct. 2963. The constitutional protections retained by prisoners include those afforded by the Due Process Clause against arbitrary deprivations of “liberty.” Some protected liberty interests flow directly from the Due Process Clause itself. See, e.g., Washington v. Harper, 494 U.S. 210, 221-22, 110 S.Ct. 1028, 108 L.Ed.2d 178 (1990). Others are created by state laws regulat *850 ing the terms or conditions of a prisoner’s confinement. See, e.g., Board of Pardons v. Allen, 482 U.S. 369, 376, 107 S.Ct. 2415, 96 L.Ed.2d 303 (1987). State-created liberty interests — the focus of this case— have their origins in Wolff, where the Supreme Court held that a Nebraska prisoner had a constitutionally protected liberty interest in retaining good-time credits because Nebraska law “not only provided a statutory right to good time but also specifies that it is to be forfeited only for serious misbehavior.” 418 U.S. at 557, 94 S.Ct. 2963; see id. at 545-53, 94 S.Ct. 2963 (discussing Nebraska statutes and prison regulations). Noting that “the prisoner’s interest has real substance,” id. at 557, 94 S.Ct. 2963, Wolff concluded that “a person’s liberty is ... protected, even when the liberty itself is a statutory creation of the State” because “[t]he touchstone of due process is protection of the individual against arbitrary action of government.” Id. at 558, 94 S.Ct. 2963 (citing Dent v. West Virginia, 129 U.S. 114, 123, 9 S.Ct. 231, 32 L.Ed. 623 (1889)). Consistent with Wolff, we have recognized that D.C. prison regulations may give rise to constitutionally protected liberty interests. See, e.g., Ellis v. District of Columbia, 84 F.3d 1413, 1415 (D.C.Cir.1996).

The difficult question in a case such as this is how to reconcile the two principles at work in Sandin — that is, how do we define the range of state-created liberty interests protected by due process without unduly constricting management prerogatives of prison officials? Prior to Sandin, courts struck the balance by recognizing liberty interests where state laws or regulations contained explicit language circumscribing official authority to alter the conditions of a prisoner’s confinement. The key case was Hewitt v. Helms, supra, where a Pennsylvania inmate challenged the adequacy of proceedings that resulted in his confinement in administrative segregation after a prison riot. While observing that administrative segregation does not implicate “an interest independently protected by the Due Process Clause,” 459 U.S. at 468, 103 S.Ct. 864, Hewitt found that the prisoner had a protected liberty interest in avoiding such segregation because state law “require[d] that certain procedures ‘shall,’ ‘will,’ or ‘must’ be employed and that administrative segregation will not occur absent specified substantive predicates — viz., ‘the need for control,’ or ‘the threat of a serious disturbance,’ ” id. at 471-72, 103 S.Ct. 864 (quoting 37 Pa. Code § 95.103(b)(3) (1971)). “[T]he repeated use of explicitly mandatory language in connection with requiring specific substantive predicates,” Hewitt explained, “demands a conclusion that the State has created a protected liberty interest.” Id. at 472, 103 S.Ct. 864.

Twelve years later, Sandin abandoned Hewitt’s approach for two reasons. First, by “encouraging] prisoners to comb regulations in search of mandatory language on which to base entitlements to various state-conferred privileges,” the Court said, Hewitt’s methodology “creates disincentives for States to codify prison management procedures in the interest of uniform treatment.” Sandin, 515 U.S. at 481, 482, 115 S.Ct. 2293. Second, the Court said that “the Hewitt approach has led to the involvement of federal courts in the day-today management of prisons, often squandering judicial resources with little offsetting benefit to anyone.” Id. at 482, 103 S.Ct. 864. Citing cases where prisoners claimed liberty interests in, among other things, “receiving a tray lunch rather than a sack lunch,” id. at 483, 103 S.Ct. 864 (citing Burgin v. Nix, 899 F.2d 733, 735 (8th Cir.1990)), “receiving a paperback dictionary,” id. (citing Spruytte v. Walters, 753 F.2d 498, 506-08 (6th Cir.1985)), and “not being placed on [a] food loaf diet,” id. (citing United States v. Michigan, 680 F.Supp. 270, 277 (W.D.Mich.1988)), Sandin made clear that “the fine-tuning of the ordinary incidents of prison life” is a task for prison officials, not federal courts. Id.

Although Sandin rejected Hewitt’s methodology, it continued to “[f]ollow[] *851 Wolff [in] recognizing] that States may under certain circumstances create liberty interests which are protected by the Due Process Clause.” Id. at 483-84, 115 S.Ct. 2293. Critically, however, the Court refocused the test for identifying state-created liberty interests on what it considered “the real concerns undergirding the liberty protected by the Due Process Clause,” id. at 483, 115 S.Ct. 2293 — namely, whether the state had deprived the prisoner of “an interest of ‘real substance,’ ” id. at 480, 115 S.Ct. 2293 (quoting Wolff, 418 U.S. at 557, 94 S.Ct. 2963). Sandin declared that state-created liberty interests

will be generally limited to freedom from restraint which, while not exceeding the sentence in such an unexpected manner as to give rise to protection by the Due Process Clause of its own force, nonetheless imposes atypical and significant hardship on the inmate in relation to the ordinary incidents of prison life.

Id. at 484, 115 S.Ct. 2293 (citations omitted). Sandin thus “shift[ed] the focus of the liberty interest inquiry” from “the language of a particular regulation” to “the nature of the deprivation,” id. at 481, 115 S.Ct. 2293, or, as the Seventh Circuit put it, “from whether there was an entitlement [conferred by the state] to whether the entitlement was to some meaningful amount of liberty,” Wagner v. Hanks, 128 F.3d 1173, 1173 (7th Cir.1997).

Although clear in its intent, Sandin’s test for identifying liberty interests protected by the Due Process Clause has proven easier to articulate than to apply. See Brown v. Plant, 131 F.3d 163, 170 (D.C.Cir.1997) (identifying “a number of unsettled questions about how to apply Sandin”). The central difficulty in determining whether segregative confinement “imposes atypical and significant hardship on the inmate” is how to characterize the comparative baseline- — i.e., how to define “the ordinary incidents of prison life.” Two of our sister circuits have looked to conditions in the general prison population as the comparative baseline. See Beverati v. Smith, 120 F.3d 500, 504 (4th Cir.1997); Keenan v. Hall, 83 F.3d 1083, 1089 (9th Cir.1996). Two other circuits have looked to the typical conditions of administrative segregation. See Griffin v. Vaughn, 112 F.3d 703, 708 (3d Cir.1997); Brooks v. DiFasi, 112 F.3d 46, 49 (2d Cir.1997). Taking a different approach, the Seventh Circuit has defined the baseline as the conditions of non-disciplinary segregation in a state’s most restrictive prison. See Wagner, 128 F.3d at 1175. According to the Fifth Circuit, segregation never implicates a liberty interest unless it lengthens a prisoner’s sentence. See Carson v. Johnson, 112 F.3d 818, 821 (5th Cir.1997). The remaining circuits have applied San-din’s “atypical and significant hardship” test, but without characterizing the comparative baseline. See Bass v. Perrin, 170 F.3d 1312, 1318 (11th Cir.1999); Perkins v. Kansas Dep’t of Corrections, 165 F.3d 803, 809 (10th Cir.1999); Mackey v. Dyke, 111 F.3d 460, 463 (6th Cir.1997); Kennedy v. Blankenship, 100 F.3d 640, 642 (8th Cir.1996); Dominique v. Weld, 73 F.3d 1156, 1160 (1st Cir.1996).

We too faced this issue in Brown v. Plant, supra, another due process case brought by a Lorton prisoner challenging his placement in administrative segregation. But there we found it unnecessary to decide the “difficult and unsettled questions of constitutional law” implicated by Sandin. 131 F.3d at 165. Instead, we remanded the case to the district court to “decide, first, assuming that [the prisoner] had a liberty interest in avoiding administrative segregation, whether he received all the process that he was due.” Id. at 172. “If he did,” we said, “that will be the end of the matter.” Id. Consistent with Brown, the District claims that assuming Hatch had a liberty interest in avoiding segregative confinement, Lorton officials afforded him the process he was due. Based on the record before us, we disagree.

The parties in this case agree that if Hatch had a liberty interest in avoiding *852 administrative segregation, then Hemtt specifies the minimum procedures for placing him in such confinement. Those procedures include “some notice of the charges against him and an opportunity to present his views to the prison official charged with deciding whether to transfer him to administrative segregation.” Hewitt, 459 U.S. at 476, 103 S.Ct. 864; see Brown, 131 F.3d at 171. Although a hearing need not occur prior to confinement in administrative segregation, it “must occur within a reasonable time following an inmate’s transfer.” Hewitt, 459 U.S. at 476 n. 8, 103 S.Ct. 864. We said in Brown that these “requirements are not elaborate, but they are real, and must be strictly complied with.” 131 F.3d at 171.

Hatch alleges in his pro se complaint that he received no notice of the January 12, 1994 Housing Board hearing, that he was not allowed to attend the hearing, and that he had no opportunity to present witnesses or evidence. The District nowhere disputes these allegations, arguing instead that a subsequent exchange of letters between Hatch and Lorton officials afforded him due process under Hemtt. See 459 U.S. at 476, 103 S.Ct. 864 (noting that “[o]rdinarily a written statement by the inmate” will suffice to allow him to present his views). The record provides no support for the District’s claim. The first acknowledgment of Hatch’s letters by a prison official did not occur until February 28, over seven weeks after his initial placement in administrative segregation and over six weeks after the Housing Board hearing which Hatch did not attend — hardly “a reasonable time following [his] transfer.” Id. at 476 n. 8, 103 S.Ct. 864. Moreover, nothing in the record shows that prison officials even considered the claims Hatch raised in his letters. The facts of this case are thus unlike those in Hetoitt, where the Supreme Court found that a prisoner assigned to administrative segregation for misconduct had received due process because he “had an opportunity to present a statement to [prison officials]” at a hearing “five days after his transfer,” id. at 477, 103 S.Ct. 864, and because he “ ‘had the opportunity to have [his] version reported as part of the record,’ ” id. (quoting prisoner’s statement on misconduct report).

With respect to his placement in adjustment segregation, Hatch argues that assuming he had a liberty interest in avoiding such confinement, then he was entitled to the more elaborate protections specified in Wolff, which include the opportunity “to call witnesses and present documentary evidence in his defense.” 418 U.S. at 566, 94 S.Ct. 2963. Disagreeing with Hatch, the District claims that Wolff is inapplicable because that case involved an inmate’s loss of good-time credits, a deprivation more substantial than Hatch’s segregative confinement. We need not decide the applicability of Wolff, however, because we think it safe to say that whatever procedures are required for placing an inmate in disciplinary segregation (again, assuming a liberty interest in avoiding such confinement), they must at least encompass the Hewitt procedures that the District says are required for placing an inmate in administrative segregation. The record in this case shows that Lorton officials failed to meet those standards, i.e., they gave Hatch no “opportunity to present his views to the prison official charged with deciding whether to transfer him to ... segregation.” 459 U.S. at 476, 103 S.Ct. 864. According to Hatch’s complaint, at the January 20, 1994 Adjustment Board hearing, he “was not allowed to have any witnesses, ... was not allowed to have the writer of the [disciplinary] report present, to testify, [and] was [not] allowed to give any testimony on the record.” Amended Compl. at 1. The District challenges none of these allegations.

The District claims that the availability of habeas corpus in the D.C. courts satisfies Hewitt’s procedural requirements. But we doubt that resolution of a habeas claim would “occur within a reasonable time following an inmate’s transfer” to *853 segregation, as Hewitt requires. 459 U.S. at 476 n. 8, 103 S.Ct. 864. Moreover, given Sandin’s emphasis on preserving the administrative authority of prison officials, we are reluctant to shift primary responsibility for ensuring compliance with the Due Process Clause from Lorton administrators to D.C. judges.

Thus, because Hatch did not receive the process required by Hewitt, and because Hatch might have persuaded Lorton officials to reduce his time in segregation had he had a fair opportunity to present his views, we cannot resolve this case by taking the approach we followed in Brown.

The District suggests a second way we might decide this case without applying Sandin’s “atypical and significant hardship” test. According to the District, San-din’s test supplements Hewitt’s,, requiring Hatch to show not only that his segregative confinement was an “atypical and significant hardship,” but also that D.C. statutes or regulations had created an expectation that Lorton prisoners would not face such segregation absent certain substantive predicates. Claiming that Lorton regulations created no such expectation, the District argues that for this reason alone, Hatch had no protected liberty interest in avoiding segregative confinement.

We see no need to decide whether San-din’s test supplements or supplants Hewitt’s, for we disagree with the District that D.C. regulations governing Lorton contain no standards or guidelines limiting official discretion to place prisoners in segregative confinement. Those regulations make clear that before prison officials may place an inmate in administrative segregation, “there shall be a finding made that: (a) There is a clear and present threat to the safety of the resident; (b) The resident poses a clear and present threat to the safety of others; or (c) The resident poses a definite escape risk.” D.C. Mun. Regs, tit. 28, § 521.4; see also id. §§ 522.3, 531.2. The regulations also require Lorton officials to review an inmate’s placement in administrative segregation every thirty days, see id. § 527.1, and “[a]t each thirty-day review, it shall be the responsibility of the Board to determine whether the resident’s return to the general population at the time of that particular review still poses an escape risk or security risk to the resident or others,” id. § 527.2. The regulations authorize adjustment segregation only after an inmate has been found guilty of violating Lorton’s Code of Offenses, see id. §§ 505.1-505.3, 515.1, and they limit the term of adjustment segregation for inmates found guilty of Class II offenses to fourteen days, see id. § 505.2(c).

Like the Pennsylvania statute at issue in Hewitt, the D.C. regulations governing segregative confinement at Lorton thus contain the “repeated use of explicitly mandatory language in connection with requiring specific substantive predicates” that prior to Sandin would have “de-mandad] a conclusion that the State has created a protected liberty interest.” Hewitt, 459 U.S. at 472, 103 S.Ct. 864. Therefore, even assuming (as the District argues) that Hewitt’s test survives as an independent ground for denying the existence of protected liberty interests, we cannot avoid the key question at the heart of this case: Was Hatch’s seven-month confinement in adjustment and administrative segregation an “atypical and significant hardship ... in relation to the ordinary incidents of prison life”?

Ill

Answering this question requires us to define the comparative baseline— “ordinary incidents of prison life” — with specificity. Hatch argues that the proper baseline is the most restrictive form of confinement that Lorton officials may impose in their unfettered discretion. Claiming that Lorton officials have no discretionary authority to impose any form of confinement other than assignment to the general population, Hatch argues that comparing the conditions he faced in segregation to those faced by prisoners in the *854 general population shows that he suffered an “atypical and significant hardship.”

We faced this same issue in Neal v. District of Columbia, 131 F.3d 172 (D.C.Cir.1997), yet another case brought by a Lorton prisoner challenging his confinement in administrative segregation under the Due Process Clause. But in that case, we had no need to decide whether the proper test under Sandin “is to compare [the] circumstances [the inmate faced in segregation] to those of the general prison population” because we found that even assuming that to be the proper comparison, the inmate had not suffered an “atypical and significant hardship” within the meaning of Sandin. Id. at 175. Apart from the loss of work and other privileges, administrative segregation cost the inmate in Neal only “half of his out-of-cell time.” Id. In contrast, when Lorton officials transferred Hatch from the general population to segregative confinement, he lost not only his work privileges and his access to the gym, library, mailroom, health services, and other facilities, but also more than 95 percent of his out-of-cell time. Indeed, Hatch claims that while in the general population, he was confined “to being in his cell ... for only eleven (11) hours per day on weekdays, seven (7) hours per day on Friday, and Saturday nights, and the night before legal holidays.” Hatch Br. at 6 (filed pro se Aug. 7, 1995); cf. Roach Aff. ¶ 3 (affidavit of Lor-ton warden) (prisoners in general population “are locked down in their cells a total of at least nine (9) hours per day”). While in segregation, by comparison, he “was confined to a cell for twenty three and one half (23/6) hours per day” and all forty-eight hours of the weekend. Hatch Br. at 6,10. In addition, prisoners in the general population “are free to move from place to place within the prison complex by way of a movement pass or under Correctional Officer supervision,” “are allowed a minimum of one hour of recreation time daily,” “may engage in Group Programs, recreation and religious activities daily,” and “have daily access to the telephone between the hours of 6:00 A.M. and 12:00 Midnight.” Roach Aff. ¶ 3. Hatch “was forced to [wear] hand cuffs and leg irons whenever he left [the segregation cell block],” Hatch Br. at 9, “was not afforded any outside recreation at all,” id. at 10, was isolated from all other inmates when allowed out of his cell, see id., and received no legal telephone calls for ninety days, see id. at 11. We think these differences in confinement conditions foreclose the approach we took in Neal, requiring us now to decide whether, as Hatch argues, conditions in the general population form the proper baseline for Sandin’s “atypical and significant hardship” test. Cf. infra at 20 (explaining that the district court misread Sandin in concluding that Hatch suffered no “atypical and significant hardship” compared to conditions in the general population).

Hatch claims that his proposed baseline follows directly from the Supreme Court’s application of the “atypical and significant hardship” test in Sandin itself. Concluding that the thirty-day disciplinary segregation of a Hawaii prisoner “did not present the type of atypical, significant deprivation in which a State might conceivably create a liberty interest,” Sandin said:

The record shows that, at the time of [the inmate’s] punishment, disciplinary segregation, with insignificant exceptions, mirrored those conditions imposed upon inmates in administrative segregation and protective custody.... Thus, Conner’s confinement did not exceed similar, but totally discretionary, confinement in either duration or degree of restriction.... Based on a comparison between inmates inside and outside disciplinary segregation, the State’s actions in placing him there for 30 days did not work a major disruption in his environment.

515 U.S. at 486, 115 S.Ct. 2293 (footnotes omitted). Hatch reads this passage — in particular, the words “totally discretion *855 ary” — to mean that “the ordinary incidents of prison life” consist of the most restrictive confinement conditions that prison officials may impose in their unfettered discretion. According to Hatch, while this theory meant that conditions in administrative segregation or protective custody comprised the proper baseline in Sandin, here it means that conditions in the general population should serve as the baseline because that is the only form of confinement Lorton officials have unfettered discretion to impose.

We disagree with Hatch’s reading of Sandin. As Hatch recognizes, the phrase “similar, but totally discretionary, confinement” in the quoted passage refers to “administrative segregation and protective custody.” At the time of the events giving rise to Sandin, Hawaii prison officials did not have unfettered discretion to place inmates in administrative segregation or protective custody. State regulations authorized administrative segregation

(1) Whenever the facility administrator or a designated representative determines that an inmate or ward has committed or threatens to commit a serious infraction.

(2) Whenever the facility administrator or a designated representative, considering all the information available, incuding [sic] confidential or reliable heresay [sic] sources, determines that there is reasonable cause to believe that the inmate or ward is a threat to: (A) Life or limb; (B) The security or good government of the facility; (C) The community.

(3) Whenever any similarly justifiable reasons exists [sic].

Haw. Admin. Rule § 17-201-22 (1983). Hawaii regulations also provided:

Admission to protective custody may be made only where there is reason to believe that such action is necessary or the inmate or ward consents, in writing, to such confinement. Protective custody is continued only as long as necessary except where the inmate or ward needs long term protection and the facts requiring the confinement are documented.

Id. § 17-201-23. These regulations did not authorize prison officials to impose administrative 'segregation or protective custody for no reason at all. Because the Sandin Court was fully aware of these regulations, see 515 U.S. at 476 n. 2, 115 S.Ct. 2293 (citing Haw. Admin. Rule §§ 17-201-22, 17-201-23), we believe its use of the words “totally discretionary” cannot mean that what prison officials may do in their unfettered discretion is the touchstone for elucidating “the ordinary incidents of prison life.”

To be sure, Sandin nowhere directly explains why it used administrative segregation as the comparative baseline. But given the objectives Sandin sought to further, see supra at 5-6, we think the reason is not that such confinement is literally “totally discretionary,” but rather that prison officials routinely impose such confinement for non-punitive reasons related to effective prison management. Support for this interpretation comes from what the Court said in Hewitt about administrative segregation:

It is plain that the transfer of an inmate to less amenable and more restrictive quarters for nonpunitive reasons is well within the terms of confinement ordinarily contemplated by a prison sentence. The phrase “administrative segregation,” as used by the state authorities here, appears to be something of a catchall: it may be used to protect the prisoner’s safety, to protect other inmates from a particular prisoner, to break up potentially disruptive groups of inmates, or simply to await later classification or transfer. See 37 Pa.Code §§ 95.104 and 95.106.... Accordingly, administrative segregation is the sort of confinement that inmates should reasonably anticipate receiving at some point in their incarceration.

*856 459 U.S. at 468, 103 S.Ct. 864. Like the Pennsylvania regulations in Hewitt, the Hawaii regulations in Sandin and the D.C. regulations in this case make clear that administrative segregation functions as a “catchall,” a flexible management tool for ensuring safety and good orfler in prison. See Haw. Admin. Rule §§ 17-201-22 to - 24;' D.C. Mun. Regs. tit. 28, § 521. Given Sandin’s insistence on affording “appropriate deference and flexibility to state officials trying to manage a volatile environment,” 515 U.S. at 482, 115 S.Ct. 2293, it makes- -sense that the Court would treat administrative segregation as an “ordinary incident of prison life.” Such a baseline for identifying constitutionally protected liberty interests ensures that “the day-today management of prisons” will remain in the hands of prison administrators, not federal judges. Id.

Reading Sandin to require that we look to conditions in administrative segregation as the proper baseline does not end our analysis. Sandin took two additional factors into account. First, it observed that the prisoner’s confinement “did not exceed similar ... confinement in either duration or degree of restriction.” 515 U.S. at 486, 115 S.Ct. 2293 (emphasis added); see id. (“[T]he State’s action in placing him there for 30 days did not work a major disruption in his environment.”). When we compare Hatch’s confinement to administrative segregation, we must therefore look not only to the nature of the deprivation (e.g., loss of privileges, loss of out-of-cell time) but also to -its length in evaluating its “atypicality” and “significance.” Second, Sandin noted that the prisoner’s thirty-day disciplinary segregation “was within the range of confinement to be normally expected for one serving an indeterminate term of 30 years to life.” Id. at 487, 115 S.Ct. 2293. We read this to mean that “atypicality” also depends in part on the length of the sentence the prisoner is serving. See id. at 485, 115 S.Ct. 2293 (disciplinary segregation was not “a dramatic departure from the basic conditions of Conner’s indeterminate sentence”); id. at 486 n. 9, 115 S.Ct. 2293 (“[T]he conditions suffered were expected within the contour of the actual sentence imposed.”). We have previously interpreted Sandin just this way. In Franklin v. District of Columbia, we said that courts must consider not only “the discipline involved” but also “the nature of the prisoner’s term of incarceration” in determining “whether a prisoner’s ‘liberty’ is threatened.” 163 F.3d 625, 634 (D.C.Cir.1998).

To sum up, we interpret Sandin to mean that a deprivation in prison implicates a liberty interest protected by the Due Process Clause only when it imposes an “atypical and significant hardship” on an inmate in relation to the most restrictive confinement conditions that prison officials, exercising their administrative authority to ensure institutional safety and good ’order, routinely impose on inmates serving similar sentences. We think this standard captures what Sandin means by the phrase “ordinary incidents of prison life.” While the “incidents of prison life” encompass more or less restrictive forms of confinement depending on prison management imperatives, the term “ordinary” limits the comparative baseline to confinement conditions that prison officials routinely impose. We also think our interpretation of the test' is faithful to the principles animating Sandin: It ensures that prison officials have broad administrative authority to “fine-tun[e] the [conditions] of prison life,” 515 U.S. at 483, 115 S.Ct. 2293, while preserving-a zone of liberty interests with “ ‘real substance’ ” protected by the Due Process Clause, id. at 480, 115 S.Ct. 2293.

We turn finally to the parties’ competing claims regarding the significance of inter-prison inmate transfers for Sandin’s baseline. According to Hatch, the baseline must be defined by reference to conditions at Lorton only. We agree with the District, however, that the possibility of transfer is one of the “ordinary incidents of prison life” for most prisoners in the eoun- *857 try, including those at Lorton. See D.C.Code Ann. § 24-425 (1981) (giving Attorney General broad discretion to transfer Lorton inmates to any federal prison); cf. Meachum v. Fano, 427 U.S. 215, 225, 96 S.Ct. 2532, 49 L.Ed.2d 451 (1976) (holding that transfer to prison with more onerous conditions does not deprive a prisoner of constitutionally protected liberty “as long as prison officials have discretion to transfer him for whatever reason or for no reason at all”). At the same time, we disagree with the District that the possibility of transfer means that the baseline must consist of the most restrictive conditions routinely imposed on inmates in any prison nationwide, including conditions at the federal penitentiary at Marion, Illinois, an especially restrictive prison where all inmates are locked down almost the entire day.

Sandin defined the “ordinary incidents of prison life” in terms of the “basic conditions” of a prisoner’s sentence, 515 U.S. at 485, 115 S.Ct. 2293, the conditions “normally expected” for a prisoner serving a given term, id. at 487, 115 S.Ct. 2293. What matters, therefore, is not simply the possibility of transfer but also its likelihood. The mere fact that the Attorney General has discretion to transfer a Lorton inmate to prisons like Marion does not make such transfers “ordinary.” Properly constructed, Sandin’s baseline requires not mere inquiry into- the most restrictive conditions prison officials have legal authority to impose for administrative reasons, but a factual determination of the most restrictive conditions prison officials “ordinarily” or “routinely” impose.

We thus think that to the extent Hatch might face more burdensome conditions at other prisons, those conditions become part of the baseline only if it is likely both that inmates serving sentences similar to Hatch’s actually will be transferred to such prisons and that once transferred they actually will face such conditions. If, as the District claims, conditions for all inmates at Marion are more burdensome than the most restrictive conditions at Lor-ton that prison officials routinely impose in their administrative discretion, then conditions at Marion would form the proper baseline’under Sandin if the District can show that transfers to Marion are “normally expected” for Lorton inmates serving sentences similar to Hatch’s. Sandin, 515 U.S. at 487, 115 S.Ct. 2293. Not only does the record contain no information about the frequency of inmate transfers from Lorton to Marion, but the District’s lawyer, asked at oral argument “how many D.C. prisoners go to Marion,” said, “I don’t have a number, but at least one.” She then conceded that “[p]erhaps that one prisoner alone would not support our argument.”

• This brings us to the disposition of this case. The district court compared the conditions of Hatch’s segregative confinement (as he described them) with conditions faced by prisoners in the general population. See Mem. Order at 3-4. Finding these differences no greater than the differences in Sandin between that prisoner’s disciplinary segregation and his confinement in the general population, it then concluded that Hatch suffered no “atypical and significant hardship.” See id. at 5.

To'be sure, Sandin observed in dictum that “the conditions at Halawa involve significant amounts of'‘lockdown time’ even for inmates in the general population.” 515 U.S. at 486, 115 S.Ct. 2293. But as our earlier discussion indicates, see supra at 14-15, Sandin’s holding turned on a comparison of the prisoner’s confinement to administrative segregation: “[A]t the time of Conner’s punishment, disciplinary segregation, with insignificant exceptions, mirrored those conditions imposed upon inmates in administrative segregation and protective custody.” Id. Indeed, the Court noted that Hawaii inmates in administrative segregation receive only “one extra phone call and one extra visiting privilege” than inmates in disciplinary seg *858 regation. Id. at 476 n. 2, 115 S.Ct. 2293. The question the district court should have asked, therefore, is this: Were the differences between the conditions of Hatch’s segregative confinement and the conditions routinely imposed on Lorton inmates serving similar sentences, including the usual conditions of administrative segregation, sufficiently greater than “one extra phone call and one extra visiting privilege” so as to constitute an “atypical and significant hardship”?

We thus reverse the district court’s grant of summary judgment for the District and remand for further fact-finding consistent with this opinion. In evaluating whether Hatch had a liberty interest in avoiding adjustment segregation, the district court should begin by determining the usual conditions of administrative segregation at Lorton. It should treat those conditions as the baseline for evaluating whether Hatch’s two-week adjustment segregation was an “atypical and significant hardship.” If using that comparison the court finds that his adjustment segregation was “atypical and significant,” it should then take into account the possibility that Hatch will be transferred to other prisons. The district court should redefine the comparative baseline by reference to more restrictive conditions at other prisons if it finds that it is likely both that inmates serving sentences similar to Hatch’s will actually be transferred to such prisons and that once transferred they will actually face such conditions. The term “likely,” as we use it here, means not that the combination of events must be more probable than not, but that there must be a substantial chance of its occurrence.

As to whether Hatch had a liberty interest in avoiding administrative segregation, the fact that routine conditions of administrative segregation form the proper baseline under Sandin does not foreclose Hatch’s claim for two reasons. First, Hatch alleges that although twenty-nine weeks of his segregation were nominally “administrative,” he actually spent his entire confinement in conditions of adjustment segregation. As long as this allegation remains undisputed, the district court should undertake the same comparative analysis outlined above. Second, even if the conditions Hatch faced were no more restrictive than ordinary conditions of administrative segregation, the district court should determine whether its duration— twenty-nine weeks, including twenty weeks after the Housing Board found that he no longer posed a management problem — was “atypical” compared to the length of administrative segregation routinely imposed on similarly situated prisoners. See Brooks, 112 F.3d at 49 (“[T]he mere fact that [state] prison regulations permit extended administrative segregation does not tell how frequently or for what durations such segregation is [actually] imposed.”).

So ordered.