Randolph, Circuit Judge:

In this class action by prisoners and former prisoners of the District of Columbia, brought under 42 U.S.C. Section(s) 1983, the district court ruled that the Due Process Clause of the Fifth Amendment required the District of Columbia Board of Parole to hold parole hearings far enough in advance of prisoners' parole eligibility dates so that prisoners may be released on that date if the Board's decisions are in favor of parole; to provide certain material to prisoners to assist them in understanding the Board's parole decisions in their cases; and, with respect to parole revocations, to offer prompt "preliminary hearings" after parole warrants are executed and to provide parole revocation hearings no more than 90 days thereafter. This is an appeal from the permanent injunction issued to enforce the court's ruling.

I.

We will deal first with the district court's judgment that some, but not all, prisoners have a due process liberty interest in parole, an interest protected by the procedural rules imposed in the court's order concerning the Board's parole eligibility determinations.

A.

Because the Constitution itself does not create any liberty interest in parole (see Greenholtz v. Nebraska Penal Inmates, 442 U.S. 1, 7 (1979)), such an interest must emanate from state law, or in this case, District of Columbia law. The statutory law of parole, contained in the District of Columbia Code, is as follows:

Whenever it shall appear to the Board of Parole that there is a reasonable probability that a prisoner will live and remain at liberty without violating the law, that his release is not incompatible with the welfare of society, and that he has served the minimum sentence . . . the Board may authorize his release on parole.

D.C. Code Section(s) 24-204(a). We held in Price v. Barry, 53 F.3d 369 (D.C. Cir. 1995) (per curiam), that this statute created "no `expectancy of release' entitling a prisoner to due process protections." Id. at 371. Our reasoning was straightforward and rested on the language of Section(s) 24-204(a): even if a prisoner established everything the statute required, the Board of Parole still had discretion to deny parole. Yet state laws may only give rise "to a constitutionally protected liberty interest if they contain substantive limitations on official discretion, embodied in mandatory statutory or regulatory language." Id. at 370.

While plaintiffs thus cannot establish a liberty interest stemming from the District's parole statute, they say one may be derived from parole regulations the Board adopted in 1987 pursuant to its authority "to establish rules and regulations for its procedure." D.C. Code Section(s) 24-201a (1987), superseded by D.C. Code Section(s) 24-204.1-.3. Like the statute, the regulations state that the Board may "release a prisoner on parole in its discretion" after the prisoner has served one-third of his sentence, provided that the prisoner substantially complied with prison rules, there was a reasonable probability the prisoner would not violate the law upon release, and releasing the prisoner would not be "incompatible with the welfare of society." D.C. Mun. Regs. tit. 28, Section(s) 200.1. The regulations establish a scoring system to guide the Board's parole decisions. Section(s) 204.1. Each parole candidate is assigned a "Salient Factor Score" to assist in determining the risk of releasing the prisoner. As the regulations put it, the "SFS" serves as "one of the factors used in calculating parole eligibility pursuant to the provisions of this section." Section(s) 204.2. To calculate the SFS, the Board assigns a numerical value for each of six categories:

* Prior convictions and adjudications (ranging from 0-3),

* Prior commitments of more than thirty days (0-2),

* Age at the time of the commission of the current offense (0-2),

* Recent commitment-free period (0-1),

* Status of the prisoner at the time of commission of the current offense (0-1), and

* History of heroin or opiate dependence (0-1).

Section(s) 204.4; App. 2-1, at 2-31 to -32. These categories and the determinants of the numerical values are described in detail in 204.5-.16. The numerical values assigned to these six categories are added to determine the SFS, which can range from 0-10. App. 2-1, at 2-31 to -32. Prisoners with an SFS of 9-10 are regarded as low risk; those with scores of 6-8 are regarded as fair risk; those with scores of 4-5 are regarded as moderate risk; and those with scores of 0-3 are regarded as high risk. Section(s) 204.17; App. 2-1, at 2-32.

The Board modifies a prisoner's risk category by adding or subtracting points for pre- and post-incarceration factors. Points are added if:

* The prisoner's current conviction involved violence against a person, the use of a dangerous weapon, or drug distribution; or if the prisoner has two or more previous convictions for these types of crimes; or

* The prisoner has committed serious disciplinary infractions.

Section 204.18(a)-(h); App. 2-1, at 2-32 to -34. A point is subtracted if the inmate has demonstrated sustained achievement in prison programs, industries or work assignments. Section(s) 204.18(i); App. 2-1, at 2-33 to -34. Application of these pre- and post-incarceration factors to the prisoner's risk category yields the "total point score," which can range from 0-5. Section(s) 204.19-.20; App. 2-1, at 2-34. In initial parole hearings, the regulations state that adults with total point scores of 0-2 and youth offenders with a total point score of 0 "shall be granted" parole; adults with total points scores of 3-5 and youth offenders with total point scores of 1-5 "shall be denied" parole. Section(s) 204.19-.20; App. 2-1, at 2-34, 2-36. In later parole hearings, the Board begins with the total point score from the previous hearing and either adds or subtracts one point depending upon whether the inmate's institutional adjustment was negative or positive. Adult and youth offenders with point scores of 0-3 "shall be granted" parole; adult and youth offenders with point scores of 4-5 "shall be denied" parole. Section(s) 204.21; App. 2-2, at 2-37.

The regulations permit the Board to deviate from the outcome suggested by the total point score "in unusual circumstances." Section(s) 204.22. Appendix 2-1 lists six reasons for denying parole despite a low total point score:

* Repeated failure under parole supervision,

* Current offense involves on-going criminal behavior,

* Lengthy history of criminally related alcohol abuse,

* History of repetitive sophisticated criminal behavior,

* Unusually extensive and serious prior record (at least five felony convictions), and

* Unusual cruelty to victims.

App. 2-1, at 2-34 to -35. The list in Appendix 2-1 also contains categories for "Other" and "Other change in circumstances." Id. at 2-35. The Board supplemented these appendices with an "Addendum to Board Order" which laid out four additional factors that could justify deviating from the numerical guidelines.

The district court, impressed by the mandatory language of Section(s) 204.19-.21, held that inmates with low total point scores had an "expectation of release" and thus a liberty interest in parole. As the court saw it, the command of the regulations — prisoners with low total point scores "shall be released" — so limited the Board's discretion that the "total point score effectively determine[d] the parole decision." Although the Board could disregard the parole determination indicated by its guidelines, the Board could do so only for unusual circumstances. The court therefore concluded that the Constitution required the Board to conduct parole hearings of prisoners with low total point scores sufficiently in advance of their parole eligibility dates that they could be released as soon as they became eligible and to provide all prisoners denied parole with more information explaining the reasons for the denial.

B.

By the time of the district court's decision, the Supreme Court had settled on an approach to the Due Process Clause that made the existence, or lack thereof, of a liberty interest in parole turn on the language of the regulations governing parole. While there "is no constitutional or inherent right of a convicted person to be conditionally released before the expiration of a valid sentence" (Greenholtz, 442 U.S. at 7), a state's parole regulations might require release after a parole board "determines (in its broad discretion) that the necessary prerequisites exist" (Board of Pardons v. Allen, 482 U.S. 369, 376 (1987)), in which event the state has created an expectation of release rising to the level of a liberty interest within the meaning of the Due Process Clause. Greenholtz, Allen, and Hewitt v. Helms, 459 U.S. 460 (1983), came to stand for the proposition that a state's "use of `explicitly mandatory language,' in connection with the establishment of `specified substantive predicates' to limit discretion, forces a conclusion that the state has created a liberty interest." Kentucky Dep't of Corrections v. Thompson, 490 U.S. 454, 463 (1989).

Where the Supreme Court stands on this subject is no longer certain. Sandin v. Conner, 115 S.Ct. 2293 (1995), a case dealing with prison disciplinary proceedings, roundly criticized the methodology just described. The Sandin majority thought the Court had gone astray, particularly in Hewitt v. Helms, when it made liberty interests depend on the "somewhat mechanical dichotomy" between state regulations that were mandatory and those that were discretionary. Id. at 2298. The five Justices in the Sandin majority, and two in dissent, agreed that this approach had created undesirable disincentives for states to cabin the discretion of prison officials: states that set down strict guidelines for prison officials exposed themselves to constitutional claims; those that set down no rules to control prison officials immunized themselves from such claims. Id. at 2299; id. at 2303 (Ginsburg, J., joined by Stevens, J., dissenting). The Court also believed that "the Hewitt approach" had insinuated federal courts into the management of state prisons. "The time has come," the Court said, "to return to the due process principles . . . established" before the mandatory-discretionary dichotomy took hold. Id. at 2300. While state laws may still create liberty interests protected by the Due Process Clause, henceforth "these interests will generally be 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.

The Sandin test relates to claims dealing with the day-to-day management of prisons. It seems ill-fitted to parole eligibility determinations. Parole is, in the words of Sandin, surely a "freedom from restraint" but the restraint itself will always be an "ordinary incident of prison life." Id. In other words, if a prisoner is denied parole — if, in terms of Sandin, the prisoner is restrained — the prisoner will never suffer an "atypical" or "significant hardship" as compared to other prisoners. He will continue to serve his sentence under the same conditions as his fellow inmates. There is no room for an argument that the denial of parole always imposes extraordinary hardship by extending the length of incarceration, and therefore gives rise to a liberty interest protected by the Due Process Clause. That is simply a recasting of the argument — rejected in Greenholtz, 442 U.S. at 7-11, and unaffected by Sandin — that a liberty interest in parole stems directly from the Constitution without regard to state law. And yet given Greenholtz and Allen, an inferior court could not accept an argument that, no matter what state law provides, a prisoner's interest in parole can never amount to a liberty interest protected by the Due Process Clause. Where does this leave us? Sandin did not overrule Greenholtz or Allen or any other Supreme Court decision. 115 S.Ct. at 2300 n. 5. To be sure, it abandoned the reasoning embodied in those opinions, at least insofar as applied to prisoners challenging the conditions of their confinement or the administration of the prison. In this situation, we think the only course open to us is to comply with the rule expressed in Rodriguez de Quijas v. Shearson/American Express, Inc., 490 U.S. 477 (1989): "If a precedent of this Court has direct application in a case, yet appears to rest on reasons rejected in some other line of decisions, the Court of Appeals should follow the case which directly controls, leaving to this Court the prerogative of overruling its own decisions." Id. at 484; see also American Trucking Ass'ns v. Smith, 496 U.S. 167, 180 (1990); United States v. $639,558 in United States Currency, 955 F.2d 712, 718 (D.C. Cir. 1992). Until the Court instructs us otherwise, we must follow Greenholtz and Allen because, unlike Sandin, they are directly on point. Both cases deal with a prisoner's liberty interest in parole; Sandin does not. And so we return to the language of the regulations.

Our independent review of the regulations leads us to doubt whether the district court correctly estimated the degree to which the numerical guidelines control the Board's judgment. The first section of the chapter on parole recognizes the Board's authority "to release a prisoner on parole in its discretion." D.C. Mun. Regs. tit. 28, Section(s) 200.1 (emphasis added). The calculations done under the regulations are intended to "enable the Board to exercise its discretion when, and only when, release is not incompatible with the safety of the community." 204.1 (emphasis added). The regulations characterize the Salient Factor Score as "an actuarial parole prognosis aid to assess the degree of risk posed by a parolee" (Section(s) 204.3) and as "one of the factors used in calculating parole eligibility pursuant to the provisions of this section" (Section(s) 204.2). Without more, these provisions call into question the district court's conclusion that the numerical scores provided by the regulations "effectively determine the parole decision."

Furthermore, the regulations explicitly authorize the Board to disregard the numerical guidelines simply by referring to "the specific aggravating or mitigating factors as stated in Appendices 2-1 and 2-2." Section(s) 204.1. Among such factors listed in these Appendices are "Other" and "Other change in circumstances." App. 2-1, at 2-35; App. 2-2, at 2-38. "Other" is scarcely constraining language. It suggests the following interpretation: under the regulations, a prisoner with a low total point score shall be granted parole unless the Board, in the exercise of its discretion, believes there is some other reason for not granting him parole. The case of the lead plaintiff, Michael Ellis, is revealing in this regard. Ellis had been sentenced to 18-75 years for abducting women at gun point on three separate occasions and brutally raping them.

Before January 1988, Ellis was a "serious management problem" who had to be repeatedly disciplined for misconduct. In the two years immediately preceding his initial parole hearing, Ellis avoided any disciplinary infractions and completed his General Equivalency Diploma, a Drug Alcohol Abuse Treatment Program, and several other self-help programs.

Because Ellis had only a single prior criminal conviction, was only 21 years old at the time of the rapes, had not been imprisoned or otherwise restricted immediately prior to his conviction for those offenses, and was not addicted to heroin or opiates, Ellis received a Salient Factor Score of eight, indicating that he was a fair parole risk. Pursuant to the regulations, his risk level was adjusted up for the violence of his crimes, but was adjusted back down for his model conduct during the most recent two years of his imprisonment. In the end, he received a total point score of 1. According to the reasoning of the district court, this score entitled him to parole, subject to a high level of supervision. Section(s) 204.19(b). Instead, the Board refused to follow the indicated result and denied parole. In making this decision, the Board relied in part on one of the countervailing factors specifically listed in Appendix 2-1 and in the Board's Addendum — the fact that Ellis's offenses involved unusual cruelty to his victims. The Board was also concerned that a recent psychological assessment had described Ellis as "impulsive and rebellious with low tolerance for frustration and prone to act out with physical aggression." This last factor was not listed in Appendix 2-1 or in the Board's Addendum. But the Board apparently did not think this precluded it from relying on the unlisted factor. Instead, the Board invoked the residual category of "Other" as the basis for denying parole.

An opinion of the District of Columbia Court of Appeals, handed down after the district court decision in this case, ends all doubt about the meaning of the Board's regulations. McRae v. Hyman, 667 A.2d 1356, 1357 (D.C. 1995), presented the same issue now before us: whether the District of Columbia parole regulations create "a constitutionally protected liberty interest." Id. at 1358. In order to decide that issue the Court of Appeals recognized, as have we, that it first had to construe the regulations with an eye to determining the extent to which the Board's statutory discretion to grant or deny parole had been circumscribed. Id. at 1358-59. As to the scoring system, the court viewed this as merely a guide, not a "rigid formula," and not a constraint on the discretion conferred upon the Board by the parole statute. Id. at 1360-61. Under Section(s) 204.22 of the regulations, if the Board wished to disregard the results of the scoring system it merely had to say so in writing. Id. at 1361. The "Board is not required to either grant or deny parole based upon the score attained." Id. The Court of Appeals therefore "ma[d]e explicit what was implied in" two of its recent parole decisions — namely, that the statute and the regulations vest substantial discretion in the Board to grant or deny parole and that the regulations lack the sort of "mandatory character" needed to support a liberty interest. Id. at 1357, 1367.

Although we are not bound by the D.C. Court of Appeals's interpretation of the Constitution, we must respect its construction of D.C. law. See Mills v. Rogers, 457 U.S. 291, 300-04 (1982); cf. Clemons v. Mississippi, 494 U.S. 738, 747 (1990). Since both our own and the D.C. Court of Appeals' reading of the District's regulations indicate that parole is never "required after the Board determines that the necessary prerequisites exist" (Allen, 482 U.S. at 376) and that a prisoner's low total point score does not "compel the Board to grant a prisoner release" (Price, 53 F.3d at 371), we hold that the regulations do not give any prisoners a liberty interest in parole. The procedures the district court imposed in the name of due process were therefore unwarranted.

II.

This brings us to the matter of parole revocation. Morrissey v. Brewer, 408 U.S. 471 (1972), established that a state may not terminate a parolee's conditional liberty consistent with the Due Process Clause unless it follows "some orderly process, however informal." Id. at 482. Our concern therefore is with the constitutional adequacy of the procedures the District of Columbia already has in place.

A.

Under the Board's regulations, whenever there is probable cause to believe that a parolee has committed a crime or otherwise violated the conditions of parole, the Board or a member of the Board may issue a warrant for the parolee's arrest. D.C. Mun. Regs. tit. 28, Section(s) 217.1-.7. After the warrant is executed, the parolee "has the right to have a preliminary interview . . . at or reasonably near the place of the alleged parole violation or arrest, without unnecessary delay." Section(s) 219.1. In the preliminary interview, the parolee is informed of the parole conditions allegedly violated and is informed of his right "to written notice of the claimed violations . . .; disclosure . . . of evidence against him . . .; and opportunity to be heard in person, to present witnesses and documentary evidence, and to confront and cross-examine adverse witnesses (unless the hearing officer specifically finds good cause for not allowing confrontation) at a hearing before the Board or a member of the Board; and a written statement of the Board's final determination." Section(s) 219.1(b). The parolee is also told of the "approximate time, place, and purpose(s) of the revocation hearing." Section(s) 219.1(c). Revocation hearings must be held "at or reasonably near the place of the alleged parole violation or arrest, within sixty (60) days of the preliminary interview." Section(s) 219.3. Board policy requires that revocation hearings be held within 30 days after the Board is notified of the execution of a warrant.

In its memorandum opinion, the district court concluded that "[s]ince the defendants do not offer preliminary hearings to any alleged parole violators, they are in clear violation of the requirements of Morrissey." Furthermore, because the evidence showed that "Defendants also deny revocation hearings within 60 and even 90 days to a small but significant number of alleged parole violators," they were in "clear violation of Morrissey."

B.

1. Preliminary "hearings." The first problem — one unfortunately not mentioned by the parties — is the incongruity between the district court's opinion and its order. While the opinion speaks of the need for "preliminary hearings," the order requires something different: the Board must "provide a prompt preliminary interview when a warrant issued by the Board has been executed against a parolee for a parole violation . . . ." (emphasis added). As a matter of constitutional law, we cannot make sense of this portion of the injunction. The Board's rules already grant each parolee the right to a preliminary interview, "without unnecessary delay." Section(s) 219.1. There is no indication that the Board is ignoring the rule. One might suppose there is difference between a "prompt" preliminary interview, as the court ordered, and one held "without unnecessary delay," as the Board's rule provides. But we cannot imagine how that would amount to a difference of constitutional magnitude and, in any event, the district court never hinted that it did. This would be cause enough to reverse the portion of the order regarding preliminary interviews. It is, after all, the order which is appealable, not the court's opinion. 28 U.S.C. Section(s) 1291. We suppose it is possible that the district court considered an interview and a hearing as synonymous, although the terms have always carried quite different legal connotations. In its opinion, the court spoke about allowing the parolee to "speak, present information, and question adverse witnesses." This seems to assume that the Board must call witnesses, which is scarcely what one would expect to occur in an interview. As we have said, the parties make nothing of the terms of the court's order. They pay attention only to the court's opinion. Their arguments proceed on the assumption that the court enjoined the Board to hold an evidentiary hearing for the purpose of determining whether probable cause supported the warrant. Even if that assumption is correct, we would still reverse this portion of the order.

In discussing what sort of informal process is required before parole can be revoked, the Morrissey Court began with the proposition that "due process is flexible and calls for such procedural protections as the particular situation demands." 408 U.S. at 481. In many cases, the Court said, there is "a substantial time lag between the arrest and the eventual determination" and parolees are often held at places far from where they were arrested. Id. at 485. The two parolees in Morrissey were held 100 and 250 miles away from where they were arrested, and the parties disputed whether either parolee ever received a revocation hearing of any kind. Id. at 473. The Court held that "[g]iven these factors, due process would seem to require that some minimal inquiry be conducted at or reasonably near the place of the alleged parole violation or arrest as promptly as convenient after arrest." Id. at 485 (emphasis added). The purpose of this minimal inquiry was to determine if probable cause supported the arrest. Id. at 487. The decisionmaker could be an administrative officer, but not the person — such as the parole officer — "initially dealing with the case." Id. at 486. Before the final revocation determination, the parolee also had a right to a hearing within a reasonable time, at which he could present evidence and examine witnesses. Id. at 488.

The Morrissey Court disclaimed any intent of imposing "an inflexible structure for parole revocation procedures." Id. at 490. The task of writing such "a code of procedure" was the responsibility of each state. Id. at 488. The Court specifically rejected the argument that persons facing parole revocation must receive the same level of procedural protection provided to criminal defendants. "No interest would be served by formalism in this process; informality will not lessen the utility of this inquiry in reducing the risk of error." Id. at 487. While stating broadly that parolees have to be given the right to confront and cross-examine witnesses in both preliminary and final revocation hearings, the Court recognized that parolees need not receive these rights when "an informant would be subjected to risk of harm if his identity were disclosed" or when a hearing officer found "good cause" to deny such process. Id. at 487, 489.

The following year, in Gagnon v. Scarpelli, 411 U.S. 778 (1973), the Court held that a probationer arrested in Cook County, Illinois, and later incarcerated in Green Bay, Wisconsin, without a hearing was entitled to a preliminary and final revocation hearing under the conditions specified in Morrissey. Id. at 782. The Court cautioned, however, that Morrissey did not require live testimony in all cases; "affidavits, depositions, and documentary evidence" might suffice. Id. at 783 n. 5. Neither did Morrissey "foreclose the states from holding both the preliminary and the final hearings at the place of violation or from developing other creative solutions to the practical difficulties of the Morrissey requirements." Id.

A few years later, the Court said that no preliminary hearing is required if the parolee has already been convicted of a subsequent offense. Moody v. Daggett, 429 U.S. 78, 86-89 n. 7 (1976). One state supreme court has held that preliminary hearings are unnecessary if final revocation hearings are held within 30 days of arrest. State v. Myers, 545 P.2d 538, 544 (Wash. 1976) (in banc), cited with approval in Pierre v. Washington State Bd. of Prison Terms, 699 F.2d 471, 473 (9th Cir. 1983); see also State v. DeLomba, 370 A.2d 1273, 1275 (R.I. 1977) (probation).

To this line of authority must be added Gerstein v. Pugh, 420 U.S. 103 (1975), decided under the Fourth Amendment. The defendant had been arrested on the authority of a prosecutor's information. The Court held that a judicial probable cause determination was necessary but that the state need not provide an adversary proceeding with the attendant rights to counsel, confrontation, cross-examination, and compulsory process. Id. at 119-22, 123. Probable cause, the Court held, "traditionally has been decided by the magistrate in a nonadversary proceeding on hearsay and written testimony, and the Court has approved these informal modes of proof." Id. at 120.

There is a close analogy between the procedures of the Board of Parole and the issuance of arrest warrants in criminal cases and there is a large difference between the situation here and the one presented in Morrissey. Parole violators in the District of Columbia are taken into custody for parole violations only after the Board of Parole or a member thereof issues a parole violator warrant based on a determination of probable cause to believe the parolee committed a new offense or violated a condition of release. D.C. Mun. Regs. tit. 28, Section(s) 217.3, 217.5, 217.7. In this respect, the Board functions as does a magistrate in deciding whether to issue an arrest warrant. The standard in both instances is the same-probable cause; both proceedings are nonadversary; the evidence considered may be hearsay and in writing; live witnesses need not be called; and the target of the arrest is, for obvious reasons, not present. If that is constitutional in the criminal context — and it is — one would suppose it to be constitutional in the context of parole revocation. On the other hand, the parole system in Morrissey authorized the parole officer supervising the parolee's release to direct his arrest. For that reason (and because the parolee might be arrested far from the place of his later incarceration) the Morrissey Court held that the parolee was entitled to an informal proceeding before an independent decisionmaker — such as member of the Board of Parole — in order to determine whether probable cause existed for the arrest. The difference here is significant — the Board or its member makes the probable cause determination before the arrest. A neutral decisionmaker is in all cases interposed between the parole officer and the parolee before the parolee is taken into custody. This, in the words of Gagnon, 411 U.S. at 783 n. 5, is the District's "creative solution" to the problem Morrissey addressed.

Our only hesitation in pronouncing the District's pre-revocation hearing procedures constitutionally sufficient stems from the second of Gerstein's two grounds for distinguishing Morrissey. A preliminary parole revocation hearing, the Gerstein Court wrote, "more than the probable cause determination required by the Fourth Amendment, serves the purpose of gathering and preserving live testimony, since the final revocation hearing frequently is held some at some distance from the place where the violation occurred." 420 U.S. at 121 n. 22. The Court added that "revocation proceedings may offer less protection from initial error than the more formal criminal process. . . ." Id. at 121-22 n. 22. The Court's first ground for differentiating the process of issuing arrest warrants — in which live testimony is not required — cannot support a parolee's constitutional right to live testimony before his revocation hearing. There is no reason to believe parole violators in the District of Columbia are frequently detained far from the place of their violations. Given the small size of the District, there is every reason to believe the opposite. As to the Court's second ground for distinguishing Morrissey, it is doubtless true that revocation hearings in the District are less formal than criminal trials. But we are not sure where this consideration should lead. The Gerstein Court could hardly have meant that because a revocation hearing is less formal and less elaborate than a criminal trial, pre-revocation procedures must be more formal and more elaborate than pre-trial criminal procedures.

Morrissey stressed again and again that informal procedures would suffice and that the "full panoply of rights due a defendant in [a criminal prosecution] does not apply to parole revocations." 408 U.S. at 480, 482, 487. Furthermore, it is far from clear what the Gerstein Court sought to convey by comparing criminal trials with revocation hearings in terms of "protection against initial error" — if, by "initial error," the Court meant an improper arrest. A criminal trial, we would have thought, serves a function other than giving "protection" against arrests made on less than probable cause. A defendant's acquittal does not show that he was wrongly arrested; it shows only that the government did not prove his guilt beyond a reasonable doubt. And a defendant may be convicted through overwhelming proof even if, at the time of his arrest, the police did not have enough evidence to support the warrant. Much the same reasoning applies to revocation hearings, although the standard of proof is preponderance of the evidence.

At any rate, the Gerstein Court's lone statement about the need for protecting against initial errors strikes us as an extremely thin reed on which to rest a constitutional mandate that the Board must hold evidentiary hearings to determine probable cause despite the Board's already having made the necessary probable cause determinations in the same manner as magistrates issuing arrest warrants in criminal prosecutions. Morrissey, 408 U.S. 481, 487, 490, and later Gagnon, 411 U.S. at 783 n. 5, 788, 790, said that due process requirements are flexible and variable, that the procedures may be informal; that there should be "an uninvolved person to make [the] preliminary evaluation of the basis for believing the conditions of parole have been violated." Morrissey, 408 U.S. at 486. In an analogous circumstance, Gerstein also emphasized the need to view the State's pretrial procedure "as a whole." 420 U.S. at 124. Here, the Board's determination of probable cause is not the last step before the revocation hearing. Each parolee taken into custody has the right to a prompt preliminary interview. D.C. Mun. Regs. tit. 28, Section(s) 219.1. Although there is no requirement under District law for live witnesses at the interview, the parolee is notified of the substance of his alleged violations and of the rights that will be afforded him in the revocation hearing, which fulfill at least part of the function of a preliminary hearing. Furthermore, all parties agree that under the Board's policy — of which more in a moment — final revocation hearings must be held within 30 days of execution of parole violator warrants. Consolidating the preliminary and final revocation hearings into a single proceeding is constitutionally permissible. E.g., Gagnon, 411 U.S. at 783 n. 5; Pierre, 699 F.2d at 473. Without the pre-detention determination of probable cause and without the preliminary interview, the Board's holding of final revocation hearings within the short space of 30 days may thus itself fulfill the demands of due process. With those other procedures in place, the Board's system is constitutionally sufficient.

2. Revocation Hearings. The final issue is presented by the district court's order that the Board conduct all final parole revocation hearings within 90 days of the warrant's execution, despite the fact that the Board has neither a policy nor a practice of refusing to provide revocation hearings within that time. The district court was impressed by evidence that the Board "den[ies] revocation hearings within 60 days and even within 90 days to a small but significant number of alleged parole violators." This evidence cannot support the court's order. See Rizzo v. Goode, 423 U.S. 362, 379-80 (1976). Injunctive relief is warranted in this type of Section(s) 1983 action only if there is "a pervasive pattern . . . flowing from a deliberate plan by the named defendants." Id. at 375 (internal quotations omitted). To establish the existence of such a pattern, plaintiffs must show either that the local officials had direct responsibility for allegedly unconstitutional behavior or that the incidence of such misconduct was more severe than elsewhere; an "unadorned finding of a statistical pattern" is not sufficient. Id. at 375-76. Applying Rizzo in Washington Mobilization Committee v. Cullinane, 566 F.2d 107 (D.C. Cir. 1977), this court declined to sustain an injunction absent any showing that the defendants "directed, authorized or approved" the allegedly unconstitutional conduct. Id. at 122 (citing also Washington Free Community, Inc. v. Wilson, 484 F.2d 1078, 1081-82 (D.C. Cir. 1973)). Even if mistakes were made in some instances, "it is not reasonable to extrapolate a general policy of lawlessness from such mistakes." 566 F.2d at 123.

The District's regulations require revocation hearings to be held within 60 days of the preliminary interview (Section(s) 219.3), and the parties agree that Board policy requires revocation hearings to be held within 30 days after notice of a warrant's execution. If the Board's policy is followed, the District's system clearly comports with due process. Morrissey, 408 U.S. at 488. There is no indication that any of the individual defendants ordered, authorized, or approved deviations from the norm. Washington Mobilization Committee, 566 F.2d at 122. In the end, the district court simply relied on the kind of "unadorned statistical evidence" deemed insufficient in Rizzo, 423 U.S. at 375. This was especially unwarranted because extended delays may well be reasonable in individual cases; everything depends on the reason for the delay. Moody, 429 U.S. at 86-87; cf. Barker v. Wingo, 407 U.S. 514, 530 (1972). We therefore conclude that the evidence is insufficient as a matter of law to support the injunction ordering the Board to conduct revocation hearings within 90 days.

III.

We reverse the district court's order requiring the District of Columbia Board of Parole to hold parole hearings far enough in advance of prisoners' parole eligibility dates so that prisoners may be released on that date if the Board's decisions are in favor of parole; requiring the Board to provide certain material to prisoners to assist them in understanding the Board's parole decisions in their cases; and, with respect to parole revocations, requiring the Board to offer prompt "preliminary interviews" after parole warrants are executed and to provide parole revocation hearings no more than 90 days thereafter. We affirm the district court's order refusing to enjoin the defendants to comply with the Board's procedural regulations.

Affirmed in part, reversed in part.