OPINION

California state prisoner Carl McQuillion appeals the district court’s denial of his two consolidated petitions for writs of ha-beas corpus, brought under 28 U.S.C. § 2254 . McQuillion alleges that his due process rights were violated when the California Board of Prison Terms, in 1994, rescinded as “improvidently granted” his parole date, which had been set in 1979. We hold that, under clearly established Supreme Court precedent, the parole scheme in California under which McQuil-lion was given his parole date in 1979 gave rise to a constitutionally protected liberty interest. The process that is due before a prisoner can be deprived of such an interest is a showing that there is “some evidence” in the record to support a later rescission of that date. Because the Board’s grounds for its later rescission reflect nothing more than a disagreement with the ultimate determination reached by the earlier granting panel, the “some evidence” standard has not been met. Accordingly, we reverse.

I

In 1973, McQuillion was convicted of two counts of murder and sentenced to seven years to life with the possibility of parole. In May 1979, a hearing panel of the Community Release Board, the California parole authority that has since been renamed the Board, of Prison Terms (“the Board”), found McQuillion suitable for parole. It calculated his total term of confinement to be 336 months (28 years). The 1979 panel reached its finding of suitability for parole after a three-hour hearing. At the hearing, the panel addressed the particulars of the commitment offenses, read from a probation officer’s report, heard additional details from the deputy district attorney, and questioned McQuillion about the specifics of his actions at the murder scene and in the aftermath of the crime. The panel discussed McQuillion’s criminal history, questioning McQuillion thoroughly regarding his actions and his motivations. McQuillion answered all ’of the panel’s questions, giving detailed accounts of all of his prior offenses, including crimes for which he could still have been charged. The panel discussed with McQuillion every psychiatric report'completed on him since his imprisonment, reading aloud both positive and negative comments by evaluators. McQuillion and others told the panel of his activities since incarceration, and McQuil-lion discussed at length his plans for employment if he were paroled. In both its oral discussion of its findings and its subsequent written report, the 1979 panel referred to the seriousness of McQuillion’s crimes, his criminal history, his psychiatric evaluations, and his post-conviction activities.

The panel granted McQuillion 32 months of post-conviction credit against his 28-year term, making his initial parole date October 7, 1998, and informed McQuillion that he could earn credit that would further advance his release date. At six progress hearings held after the grant of parole, McQuillion was granted this credit and his parole date was moved up. By his November 1991 progress hearing, McQuil-lion’s .release date had been advanced to March 7, 1995. McQuillion had consistently received the standard credit of four months for every year served. If he had received this same credit at his May 1994 progress hearing, he would have received 10 months credit for the two and a half years between November 1991 and May 1994, and his parole date would have been advanced to May 7, 1994, making the May 1994 progress hearing his last.

However, at the May 1994 progress hearing, the panel of the Board reviewing McQuillion’s case did not advance his parole date. Instead, it voted to refer the matter to the en banc Board for a vote as to whether a rescission hearing should be held. In a closed executive meeting on July 12, 1994, the Board voted to schedule a parole rescission hearing for McQuillion. The Board indicated that the hearing was to “determine whether an improvident grant of parole occurred as a result of the May 16, 1979 hearing panel’s failure to appropriately consider the following factors: (1) Gravity of the commitment offense; (2) The prisoner’s prior criminal history; (3) The prisoner’s ambiguous psychiatric reports; (4) No indications of remorse for the victims; (5) Lack of vocational programming.”

In September 1994, the rescission hearing was held before a panel of three Board members. After McQuillion’s counsel objected to issue four on the ground that remorse was not a factor for parole before 1977, the presiding commissioner eliminated the issue. The remaining four issues were addressed in questioning by the panel at the hearing. In his closing statement, the deputy district attorney, who argued in favor of rescission, acknowledged that each of the four factors addressed by the rescission panel had been considered by the granting panel.

California parole rescission proceedings take place in two phases. In the first phase, the panel determines only whether there is “good cause” to rescind the grant of parole. Cal.Code Regs. tit. 15 § 2450 (2000). If good cause is found, the panel moves to the second phase to determine whether, given that finding, the parole date should in fact be rescinded because an independent evaluation of the prisoner’s suitability for parole so dictates. See Cal Code Regs. tit. 15 §§ 2467, 2281.

After a 27-minute deliberation in the first phase, the panel announced that it had found good cause for rescission on all four grounds. After a 13-minute deliberation in the second phase, the panel announced that McQuillion’s parole date, granted 15 years earlier, was rescinded. MeQuillion’s challenge focuses on the panel’s action in the first phase.

McQuillion filed a petition for a writ of habeas corpus in a California state trial court, alleging that the rescission panel acted without good cause and thereby violated his due process rights. That petition was denied, and the California Court of Appeal and California Supreme Court also summarily denied McQuillion’s habeas petitions. After these denials, McQuillion filed two separate habeas petitions in federal district court that ultimately were consolidated. In his federal petitions, McQuillion again claimed, as he had claimed in state court, that the Board violated federal due process in rescinding his previously granted parole date. The district court denied habeas relief, and McQuillion timely appealed. We granted a Certificate of Appealability on the issue of whether McQuillion’s due process rights were violated when the Board, in 1994, rescinded as “improvidently granted” the parole date that had been established in 1979. We have jurisdiction under 28 U.S.C. §§ 1291 and 2253. We review de novo the district court’s decision to deny a 28 U.S.C. § 2254 habeas petition. See Alvarado v. Hill, 252 F.3d 1066, 1068 (9th Cir.2001).

II

As an initial matter, McQuillion argues that the State deprived him of due process by refusing to allow him to call the members of the granting panel as witnesses in his hearing before the rescission panel. McQuillion is correct that due process demands that he have a general right to call witnesses at his rescission proceeding. See Wolff v. McDonnell, 418 U.S. 539, 566 , 94 S.Ct. 2963 , 41 L.Ed.2d 935 (1974) (holding, in the context of prison disciplinary proceeding, that an inmate has a right to call witnesses unless doing so would be “unduly hazardous to institutional safety or correctional goals”); see also John v. United States Parole Comm’n, 122 F.3d 1278, 1284 (9th Cir.1997) (holding that blanket prohibition of right to call adverse witnesses at an institutional revocation hearing, without any specific determination of good cause, violated parolee’s right to due process). However, we hold that McQuillion was not deprived of his due process rights by the Board’s refusal to allow him to call these particular individuals as witnesses.

A right to call witnesses does not ordinarily encompass a right to call a fact-finder-decisionmaker to the stand. McQuillion’s request at the rescission hearing was the functional equivalent of seeking to call a judge from his earlier trial as a witness. Cf. Fed.R.Evid. 605 (stating that a judge is not competent to testify as a witness). Even in cases in which the question posed is whether a particular judge abused his or her discretion, we do not call the judge to the stand to analyze his or her deliberative process; rather, we examine the record. The rescission panel did not have a transcript of the actual deliberations of the granting panel. But, as the presiding commissioner on the rescission panel noted, there was sufficient evidence in the record (including both the full transcript of the grant hearing and the granting, panel’s official written decision), to allow the rescission panel to determine whether the parole date had been improvidently granted because the granting panel had failed to adequately consider particular factors. California law presumes that each piece of evidence presented to the granting panel was considered by it. See In re Caswell, 92 Cal.App.4th 1017, 1031 , 112 Cal.Rptr.2d 462 (2001). Moreover, at several points in the transcript, the granting panel explicitly stated what it intended to discuss, and what it had discussed, in its deliberations. The official written decision of the Board outlines in great detail the factors that went into the granting panel’s calculation of McQuillion’s term. McQuillion made no showing of any special need to call the members of the granting panel as witnesses, and his due process rights were not violated when the rescission panel rejected his request.

Ill

McQuillion also argues that his due process rights were violated because he was improperly deprived of his liberty interest in parole. “A procedural due process claim has two distinct elements: (1) a deprivation of a constitutionally protected liberty or property interest, and (2) a denial of adequate procedural protections.” Brewster v. Bd. of Educ. of Lynwood Unified Sch. Dist., 149 F.3d 971 , 982 (9th Cir.1998). In addressing McQuillion’s claim that his due process rights were violated when the Board rescinded his parole date as “improvidently granted,” we begin by examining whether he had a constitutionally protected liberty interest in a parole date.

Because McQuillion’s due process claim comes before us in a petition for habeas relief under 28 U.S.C. § 2254 , the State suggests that we are required to analyze his claim under the deferential standard of the Antiterrorism and Effective Death Penalty Act of 1996 (“AEDPA”). AEDPA provides that “[a]n application for a wilt of habeas corpus on behalf of a person in custody pursuant to the judgment of a State court shall not be granted with respect to any claim that was adjudicated on the merits in State court proceedings unless the adjudication of the claim resulted in a decision that was contrary to, or involved an unreasonable application of, clearly established Federal law, as determined by the Supreme Court of the United States.” 28 U.S.C. § 2254 (d)(1). McQuil-lion is a “person in custody pursuant to the judgment of a State court,” but he is not challenging anything that happened during the trial that led to his conviction and sentence. Nonetheless, it is possible that, because his claim of a due process violation by the state parole authority was heard by state courts on collateral review, that claim has been “adjudicated on the merits in [a] State court proceeding” within the meaning of AEDPA. Cf. White v. Indiana Parole Bd., 266 F.3d 759, 763-64 (7th Cir.2001). If the deferential AEDPA standard does apply in this case — which we assume without deciding — McQuillion qualifies for habeas relief on due process grounds only if the State has infringed a liberty interest that has been “clearly established” by the Supreme Court.

We have made clear that “the Supreme Court need not have addressed the identical factual circumstance at issue in a case in order for it to have created ‘clearly established’ law governing that case.” Van Tran v. Lindsey, 212 F.3d 1143, 1154 (9th Cir.2000). Rather, it is enough that the Supreme Court has prescribed a rule that plainly governs the petitioner’s claim. In this ease, the Court has done just that. We therefore hold that “clearly established Federal law, as determined by the Supreme Court of the United States” provides that California prisoners like McQuil-lion have a cognizable liberty interest in release on parole.

The governing rule in this area was articulated by the Supreme Court in Greenholtz v. Inmates of Nebraska Penal, 442 U.S. 1 , 99 S.Ct. 2100 , 60 L.Ed.2d 668 (1979), and Board of Pardons v. Allen, 482 U.S. 369 , 107 S.Ct. 2415 , 96 L.Ed.2d 303 (1987). Greenholtz and Allen established that, while “[t]here 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 , 99 S.Ct. 2100 , a state’s statutory scheme, if it uses mandatory language, “creates a presumption that parole release will be granted” when or unless certain designated findings are made, and thereby gives rise to a constitutional liberty interest. Id. at 12 , 99 S.Ct. 2100 ; Allen, 482 U.S. at 377-78 , 107 S.Ct. 2415 . The California parole scheme uses mandatory language and is largely parallel to the schemes found in Greenholtz and Allen to give rise to such an interest. California Penal Code § 3041(b) reads:

Id. (emphasis added). The Nebraska statute at issue in Greenholtz provided:

Neb.Rev.Stat. §§ 83-1,114(1) (1976) (emphasis added); see also Greenholtz, 442 U.S. at 11 , 99 S.Ct. 2100 . The Montana statute at issue in Allen provided:

Mont.Code Ann. § 46-23-201 (1985) (emphasis added); see also Allen, 482 U.S. at 376-77 , 107 S.Ct. 2415 .

Although we have never directly held that California’s parole scheme creates a protected liberty interest, we have repeatedly assumed that it does. See Powell v. Gomez, 33 F.3d 39, 40 (9th Cir.1994); Perveler v. Estelle, 974 F.2d 1132, 1134 (9th Cir.1992). Today we hold what we have previously assumed. Under the “clearly established” framework of Green-holtz and Allen , we hold that California’s parole scheme gives rise to a cognizable liberty interest in release on parole. The scheme “ ‘creates a presumption that parole release will be granted’ ” unless the statutorily defined determinations are made. Allen, 482 U.S. at 378 , 107 S.Ct. 2415 (quoting Greenholtz, 442 U.S. at 12 , 99 S.Ct. 2100 ).

In addition to the determinations set forth in § 3041(b), supra, the California parole scheme provides that release will not be granted if an already-granted parole date is rescinded for “good cause.” Cal.Code Regs. tit. 15 § 2450; see Caswell, 92 Cal.App.4th at 1026 , 112 Cal.Rptr.2d 462 (“Even after parole is granted, the Board is authorized to rescind the grant of parole, if unexecuted, for good cause after a rescission hearing.”). “Good cause” for rescission includes conduct enumerated in section 2451 of Title 15 of the California Code of Regulations, which at the time of McQuillion’s rescission hearing included: “(1) any disciplinary conduct subsequent to the grant of parole, (2) psychiatric deterioration of the prisoner, and (3) new information indicating parole should not occur, such as an inability to meet a special condition of parole, information significant to the original grant of parole being fraudulently withheld from the Board, or fundamental errors which resulted in the improvident grant of parole.” Caswell, 92 Cal.App.4th at 1026 , 112 Cal.Rptr.2d 462 (referencing former Cal Code Regs. tit. 15 § 2451). As in Greenholtz and Allen , the state parole scheme dictates that the Board “shall” release prisoners like McQuillion unless one or more of those specified criteria are satisfied. McQuillion therefore had an expectation of parole protected by the Due Process Clause.

The State argues that the case of Sandin v. Conner, 515 U.S. 472 , 115 S.Ct. 2293 , 132 L.Ed.2d 418 (1995), eliminated the “mandatory language” approach of Greenholtz and Allen . To a limited extent, the State is correct. The Court in Sandin addressed the separate but related question of when due process liberty interests are created by internal prison regulations. Noting that the Court in Hewitt v. Helms, 459 U.S. 460 , 103 S.Ct. 864 , 74 L.Ed.2d 675 (1983), had adopted the “mandatory language” approach in the context of administrative segregation, the Sandin Court indicated that after Hewitt , it had “wrestled with the language of intricate, often rather routine prison guidelines to determine whether mandatory language and substantive predicates created an enforceable expectation that the State would produce a particular outcome with respect to the prisoner’s conditions of confinement.” Sandin, 515 U.S. at 480-81 , 115 S.Ct. 2293 . Noting that this expansion had created disincentives for States to codify prison management procedures and had led to the involvement of federal courts in the day-to-day management of prisons, the Court announced that it would abandon the “mandatory language” framework and instead focus its liberty interest inquiry on ensuring “freedom from restraint which ... imposes atypical and significant hardship on the inmate in relation to the ordinary incidents of prison life.” Id. at 484 , 115 S.Ct. 2293 .

It is clear from the Court’s framing of the problem in Sandin , and from the fact that Sandin cited Allen with approval, see id., that Sandin’s holding was limited to internal prison disciplinary regulations. Courts and commentators that have considered the question in the wake of Sandin have reached this conclusion. See Ellis v. District of Columbia, 84 F.3d 1413, 1417-18 (D.C.Cir.1996) (Sandin does not deal with a prisoner’s liberty interest in parole and does not overrule Greenholtz and Allen ); Orellana v. Kyle, 65 F.3d 29, 32 (5th Cir.1995) (Sandin does not overrule previous law concerning the creation of liberty interests with respect to parole decisions); see also Susan N. Herman, “Slashing and Burning Prisoners’ Rights: Congress & the Supreme Court in Dialogue,” 77 Or. L.Rev. 1229, 1260 (1998) (citing eases for the proposition that “[cjhallenges to parole denial also, presumably, would not have to satisfy the new test” of Sandin). ' We agree.

Finally, we note that, while it suffices for purposes of this liberty interest analysis that the California parole scheme creates a protected interest in California prisoners in general, McQuillion’s position as a prisoner who had already been granted a parole date heightens that interest in this case. That is, not only did the mandatory language of the general parole scheme create a general expectation of parole; the decision of the Board in McQuillion’s case, containing the words “Parole Granted,” created a specific expectation. See Kelch v. Director, Nevada Dep’t of Prisons, 10 F.3d 684, 688 (9th Cir.1993) (order granting commutation of sentence created a legitimate liberty interest in commutation even before actual release). We therefore hold that McQuillion, who even without any further grants of credit was only six months shy of a release that he had been affirmatively promised a decade and a half earlier, had a constitutionally protected interest in freedom from confinement in accordance with the substantive criteria established by the State that would either require release or permit rescission.

IV

Having determined that McQuillion had a constitutionally protected liberty interest, we next examine whether the deprivation of that interest lacked adequate procedural protections and therefore violated due process. See Brewster, 149 F.3d at 982. We begin by examining “what process is due.” Morrissey v. Brewer, 408 U.S. 471, 481 , 92 S.Ct. 2593 , 33 L.Ed.2d 484 (1972). In Jancsek v. Oregon Bd. of Parole, 833 F.2d 1389 (9th Cir.1987), we held that the process that is due in the parole rescission setting is the same as the Supreme Court outlined in Superintendent v. Hill, 472 U.S. 445 , 105 S.Ct. 2768 , 86 L.Ed.2d 356 (1985), for prison disciplinary proceedings. The Court in Hill “concluded that the requirements of due process are satisfied if some evidence supports the decisión. It held, “We decline to adopt a more stringent evidentiary standard as a constitutional requirement.... The fundamental fairness guaranteed by the Due Process Clause does not require courts to set aside decisions of prison administrators that have some- basis in fact.’ ” Jancsek, 833 F.2d at 1390 (quoting Hill, 472 U.S. at 456 , 105 S.Ct. 2768 ) (alterations in original) (emphasis added); see also Perveler v. Estelle, 974 F.2d 1132, 1134 (9th Cir.1992). “Additionally, the evidence underlying the board’s decision must have some indicia of reliability.” Jancsek, 833 F.2d at 1390 .

Thus, in reviewing the parole rescission determination in this case, we ask only if the determination has been supported by “some evidence” having “some indicia of reliability.” Specifically, because McQuil-lion challenges the rescission panel’s initial determination that “good cause” for rescission existed, our particular inquiry is whether “some evidence” existed to support that “good cause” determination. As noted above, the panel’s “good cause” determination needed to be premised on a finding of “(1) any disciplinary conduct subsequent to the grant of parole, (2) psychiatric deterioration of the prisoner, [or] (3) new information indicating parole should not occur, such as an inability to meet a special condition of parole, information significant to the original grant of parole being fraudulently withheld from the Board, or fundamental errors which resulted in the improvident grant of parole.” Caswell, 92 Cal.App.4th at 1026 , 112 Cal.Rptr.2d 462 (referencing former Cal.Code Regs. tit. 15 § 2451).

The “improvidently granted” basis for good cause for rescission — the basis on which the rescission panel in this case purported to have acted — has been read somewhat broadly by California courts. These courts have held that good cause cán include not only a determination that parole was improvidently granted in light of circumstances appearing at a later time, but also a determination that it was improvidently granted under the circumstances appearing at the time of the grant. See Caswell, 92 Cal.App.4th at 1026 , 112 Cal.Rptr.2d 462 ; see also In re Powell, 45 Cal.3d 894, 902 , 248 Cal.Rptr. 431 , 755 P.2d 881 (1988); In re Johnson, 35 Cal.App.4th 160, 168 , 41 Cal.Rptr.2d 449 (1995); In re Fain, 139 Cal.App.3d 295, 304 , 188 Cal.Rptr. 653 (1983). Among other things, under California law a rescission panel may rescind parole based on a finding that the granting panel “failed to adequately consider” important factors already before , it when the grant was made. Caswell, 92 Cal.App.4th at 1027 , 112 Cal.Rptr.2d 462 ; Johnson, 35 Cal.App.4th at 168-69 , 41 Cal.Rptr.2d 449 .

However, the “improvidently granted” ground for rescission under California law is not a carte blanche for a later panel to re-decide parole eligibility. The rescission panel may base its “good cause” determination on a finding that the parole d.ate was improvidently granted because the granting panel gave inadequate consideration to important facts. But the rescission panel may not find “good cause” based on its own, differing assessment of facts that were thoroughly considered at the time of the grant. The most recent pronouncement of a California court on this issue has emphasized this distinction.

In Caswell , decided last year, the' California Court of Appeal stressed that “a rescission may not be upheld merely because the Board has mouthed words that have been held to constitute ‘cause’ for rescission. There must also be an adequate factual underpinning for the Board’s determination of cause.” 92 Cal.App.4th at 1027 , 112 Cal.Rptr.2d 462 (internal quotation marks and citation omitted) (emphasis in original). When the “basis for rescission [is] not new evidence, but a purportedly inadequate consideration of evidence by the granting Board,” id. at 1028 , 112 Cal.Rptr.2d 462 , the Caswell court held, the ground for rescission must “reflect[ ] more than a mere disagreement with the ultimate determination reached by the [granting] panel.” Id. at 1034 , 112 Cal.Rptr.2d 462 . Instead, the rescission panel must “target! ] a specific conclusion of the granting panel and establish[ ] the disparity between the conclusion and the evidence.” Id.

The earlier case of In re Johnson, 35 Cal.App.4th 160 , 41 Cal.Rptr.2d 449 (1995), had suggested that if “reasonable minds could differ” as to the conclusion to be reached on the suitability factors, there was good cause to rescind the parole date as improvidently granted. 35 Cal.App.4th at 169 , 41 Cal.Rptr.2d 449 . But the Cas-well court emphasized that this language “should not be misconstrued”:

Caswell, 92 Cal.App.4th at 1029 , 112 Cal.Rptr.2d 462 . The Caswell court held that, in a review of a rescission of parole based on the inadequate consideration of evidence, “the proper focus is on the findings and conclusions that were central to the original panel’s ultimate decision to grant parole.” Id. Only when “these findings or conclusions cannot be reconciled with the evidence before the granting panel, or when the granting panel misstated facts or explicitly declined to consider information germane to the [particular suitability factor]” will there exist “some evidence” of the panel’s failure to adequately consider the evidence, thereby justifying rescission of the parole release date on the grounds that “fundamental errors occurred, resulting in the improvident granting of a parole date.” Id.

We find the analysis of the Caswell court instructive in this case. In accordance with the United States Supreme Court’s decision in Hill and our- decision in Jancsek , the California Supreme Court has indicated that due process requires that there be “some evidence” supporting parole rescission. See In re Powell, 45 Cal.3d at 904 , 248 Cal.Rptr. 431 , 755 P.2d 881 (explicitly following Hill). Applying the “some evidence” standard and citing In re Powell, the court in Caswell examined each of the five “good cause” “improvidently granted” grounds cited by the rescission panel in that case, and found that four of the panel’s “good cause” determinations were not supported by “some evidence.” As to these determinations, the court noted that the rescission panel’s allegation of inadequate consideration “appeared] to be derived solely from the rescinding panel’s disagreement with the result of the [granting ] hearing.” Caswell, 92 Cal.App.4th at 1031 , 112 Cal.Rptr.2d 462 . The Caswell court highlighted the statements the prisoner and others had made at the granting hearing, stressed that the granting panel had been “provided a broad array” of material on the issues in question, and emphasized that the prisoner had answered questions presented to him by the granting panel on each of the issues. Id. at 1031-34 , 112 Cal.Rptr.2d 462 . Noting that, under California law, a court on review “may presume the panel considered the evidence before it,” id. at 1031 , 112 Cal.Rptr.2d 462 , the Caswell court found no support for the rescission panel’s findings of inadequate consideration on the issues that had been addressed at the time of the grant.

Only the fifth ground cited by the rescission panel was upheld in Caswell as an appropriate “good cause” determination. On that ground, the granting panel’s conclusion was “so at odds with the material before [it]” that it could be said there was “some evidence” supporting the conclusion that the panel failed to adequately consider the evidence before it. Id. at 1030 , 112 Cal.Rptr.2d 462 . The rescission panel had found that the granting’ panel misconstrued the criminal conduct relative to a particular victim and thereby inappropriately arrived at a mitigated term of imprisonment for its parole calculation. This finding, the court noted, was a “specific finding, which, in the context of the record before us, was central to [the granting panel’s] ultimate determination of the gravity of [the prisoner’s ] offenses and his suitability for parole. Furthermore, the [rescission panel] pointed to a factual basis in the record for concluding the granting panel’s finding was inconsistent with the facts before it.” Id. at 1034 , 112 Cal.Rptr.2d 462 (emphasis in original). This ground therefore justified the rescission of the prisoner’s unexecuted grant of parole. Id. at 1035 , 112 Cal.Rptr.2d 462 .

Like the Caswell court, we must apply the federal due process standard announced by the Supreme Court in Hill. We must determine whether the rescission panel had “some evidence” to support its finding of “good cause” to rescind McQuil-lion’s parole. We examine in turn each of the four “good cause” grounds set forth by the rescission panel to determine if there was “some evidence” supporting any of those grounds. We hold that none of the grounds was supported by “some evidence,” and that therefore McQuillion’s due process rights were violated when his parole was rescinded as improvidently granted.

1. Failure of the Granting Panel to Adequately Consider the Gravity of the Offense

The rescission panel first found that the granting panel did not adequately consider the gravity of McQuillion’s commitment offense. It- gave the following reasons in support of this finding:

In its statement after it had rescinded parole, the panel told McQuillion that it “did not feel that the Granting Panel really considered the fact that this was a double murder and granted you parole in a period of five years.”

As Caswell makes clear, it is not enough for a rescission panel simply to say it did not “feel” that a granting panel gave a fact enough consideration or that it did not think that the granting panel was thorough, without pointing to any “specific finding” in the record that was “central to the [granting panel’s] ultimate determination of the gravity of[the prisoner’s] offenses and his suitability for parole.” 92 Cal.App.4th at 1034 , 112 Cal.Rptr.2d 462 . Nor is it enough for the rescission panel to highlight that some of the material before the granting panel might have factored against a grant of parole, for, as noted in Caswell , that will almost always be the case. Id. at 1029 , 112 Cal.Rptr.2d 462 . The rescission panel’s statement gives no analysis as to why it might have thought that the granting panel’s conclusion was “so at odds with the material before [it]” that the inadequate consideration of that material constituted good cause for rescission. Id. at 1030 , 112 Cal.Rptr.2d 462 . Rather, its blanket statement that the granting panel’s “examination was not thorough” reflects nothing more than “mere disagreement with the ultimate determination reached” by that panel. Id. at 1034 , 112 Cal.Rptr.2d 462 .

The specifics that the rescission panel did list — the “execution” style of the crime, the presence of conflicting witness testimony as to the nature of the crimes, and McQuillion’s acceptance of responsibility— were all addressed in detail in the granting hearing. The granting panel questioned McQuillion on each of these issues, and, like the panel in Caswell , was “provided a broad array” of material regarding each of them. The granting panel read aloud the particulars of the commitment offense from a probation report and heard additional details from the deputy district attorney. McQuillion admitted his guilt, explaining how he and another man had killed the father-and-son proprietors of a sporting goods store in the course of what McQuillion said was planned to be only an armed robbery. McQuillion told the granting panel that he was in the process of tying up the son when he shot him. McQuillion did so, he said, because the son, startled at the sound of his father falling to the floor upstairs, turned for one of McQuillion’s guns. McQuillion admitted that the father’s shooting, which was carried out by his partner, was planned after McQuillion killed the son, but he indicated that it was not something the two had intended to do when they entered the store. He acknowledged that he was “responsible for the death of two people.”

The granting panel questioned McQuil-lion about statements made by a woman who claimed to have overheard McQuillion and his partner discussing the crime. The woman had said she heard McQuillion’s partner ask him, “Why did you put that over his head to shoot him?” and heard McQuillion reply, “Because I didn’t want to watch him die.” She also claimed to have heard McQuillion state that his partner had put a gun in the mouth of one of the victims before shooting. McQuillion disputed the woman’s statements, and told the panel that she, did not overhear what she claimed to have overheard. The panel pressed him on the inconsistencies between his account and the woman’s, and asked him for specific details regarding both the crime’s commission and its aftermath.

We presume that the panel considered all of this evidence that was squarely before it. But even in the absence of that presumption, there is sufficient evidence in the official report of decision by the granting panel to establish that it fully considered the gravity of McQuillion’s offense. In its report, the granting panel discussed the details of the killings, the witness’ report of overhearing McQuillion’s conversation, and the fact that McQuillion had initially denied guilt. In setting the parole date, the granting panel specifically found as aggravating factors the fact that “[d]ur-ing the commission of the crime, the prisoner had a clear opportunity to cease but instead continued,” and the fact that a crime “was committed to preclude testimony [of] an actual witness.” Further, it imposed discretionary enhancements for the use of the firearm and for the commission of a first-degree murder.

Given this depth of analysis by the granting panel, and the cursory disapproval statement by the rescission panel, we hold that the rescission panel’s “good cause” finding on this issue was not supported by “some evidence.” Accordingly, the rescission of the grant of parole as “improvidently granted” was not justified on grounds of the granting panel’s failure to adequately consider the evidence pertaining to the gravity of the offense. The rescission panel obviously disagreed with the substance of the granting panel’s decision. But this does not mean that the granting panel did not adequately consider the evidence before it.

2. Failure to Adequately Consider Pri- or Criminal History

The rescission panel also found that the granting panel’s failure to adequately consider MeQuillion’s prior criminal history constituted good cause for rescission. Its reasons for this finding were as follows:

If the granting panel is presumed to have considered all of the evidence before it— particularly the facts about which it extensively questioned the prisoner during the hearing—there is no disparity between its decision to grant parole and the evidence of McQuillion’s prior criminal history. All of McQuillion’s prior and subsequent offenses, including uncharged offenses and McQuillion’s prison escape, were discussed at the granting hearing. McQuillion gave detailed, first-person accounts of his crimes and answered panel questions as to what he did and why he did it. At one point, just before the granting panel began its deliberation, one panel member expressed concern about McQuillion’s criminal acts after the commission of the murders, and specifically indicated that this issue was “among the things I will be talking with my colleagues about when we are deliberating.”

The granting panel’s report of its decision confirms its consideration of McQuil-lion’s prior crimes. The panel indicated that it had “great concern about the severity and kinds of crimes” that McQuillion had committed and indicated that “[t]here was a lot of discussion as to whether [McQuillion] should be found suitable,” but noted that it was influenced by the fact that all of McQuillion’s crimes occurred within a three-year period and that he had no juvenile record. These factors were carefully articulated, and do not suggest that McQuillion’s history was inadequately considered by the granting panel.

The rescission panel did not point to a specific finding or to a factual basis in the record to justify a conclusion that the granting panel’s finding was inconsistent with the facts before it. Indeed, the rescission panel’s statement on this ground does not even allege that particular aspects of McQuillion’s criminal past were overlooked or inadequately considered by the granting panel; rather, it simply lists some aspects of the criminal history that it apparently considered important and thought pointed to a different end result. The granting panel’s conclusion, reached after weighing these very factors against other factors presented at the hearing, was not “so at odds with the material before [it]” that it can be said parole was improvidently granted. Caswell, 92 Cal.App.4th at 1030 , 112 Cal.Rptr.2d 462 . We therefore hold that the rescission panel’s “con-clusory disagreement with the granting panel’s ultimate decision” did not constitute “some evidence” of good cause for rescission on this ground. Id. at 1029 , 112 Cal.Rptr.2d 462 .

3. Failure to Adequately Consider Ambiguous Psychiatric Reports

The rescission panel next found that good cause existed because the granting panel had inadequately considered psychiatric evidence. The rescission panel wrote:

As the rescission panel’s statement suggests, McQuillion underwent a number of psychiatric evaluations in the time period between his imprisonment and the hearing at which parole was granted. Each of these was the subject of discussion and review by the granting panel. The panel read aloud from and questioned McQuillion regarding a series of psychiatric reports, noting specifically that a 1974 intake report had suggested that McQuillion struggled with a neurosis that prevented him from urinating in front of others, and that this report had suggested therapy for this problem. The panel also discussed a 1976 report from Dr. Gilbert, who had found that McQuillion had an “excellent” parole prognosis, that his offense was not the product of a diagnosable psychiatric disorder, and that he presented a “lower than normal expectation of violence potential.” The panel went on to discuss in some detail with McQuillion the fact that Dr. Dainard, the acting chief psychiatrist at the time, had reviewed -Dr.' Gilbert’s report and disagreed with his conclusion, writing by hand at the bottom of the report: “How do you rate this man’s prognosis as excellent? I would say this prognosis should be guarded at best,” McQuillion told the panel that he never met with Dr. Dainard, and said that he understood how someone whose analysis was based only on a paper record of McQuillion’s life and actions might reach the conclusion Dr. Dainard reached. The granting panel also read extensively from a then-very-recent report from Dr. Nuerenberger, who spoke positively of McQuillion’s progress while in prison and concluded that “his future use of psychiatric services can be left to discretion” and “[p]arole date and conditions can be determined on other than psychiatric considerations.”

The rescission panel’s “good cause” finding on this ground appears to be an attempt to point to a “specific finding, which, in the context of the record before us, was central to [the granting panel’s] ultimate determination of ... suitability for parole” and then to identify a “factual basis in the record for concluding the granting panel’s finding was inconsistent with the facts before it.” Caswell, 92 Cal.App.4th at 1034 , 112 Cal.Rptr.2d 462 . The granting panel told McQuillion that the absence of psychiatric contraindications to parole “weighed very heavily” in its decision to grant a parole date, and thus this issue was central to its ultimate determination of McQuil-lion’s parole suitability. It appears the rescission panel concluded that, to the extent the granting panel relied on the supportive report of Dr. Nuerenberger, the granting panel failed to address key psychiatric issues that were mentioned in the earlier reports and not addressed by Dr. Nuerenberger. If there were “some evidence” in support of this conclusion, it might well constitute good cause for rescission on the ground that inadequate consideration of the facts before the panel led to an improvident grant of parole.

But the rescission panel’s determination does not provide even “some, evidence” of good cause on this issue. Although the rescission panel hints at the nature of its objection to the granting panel’s analysis of the psychiatric evidence, the rescission panel did not point to any specific finding by the granting panel, such as, for example, a statement that the granting panel relied on Dr. Nuerenberger’s report and disbelieved the others. Indeed, the rescission panel could not have done so, because the granting panel’s official report does not elevate Dr. Nuerenberger’s report over the .others, but rather addresses each of the reports, including the 1974 finding of. “inadequate personality associated with some anxiety neurosis” and Dr. Dainard’s disagreement with Dr. Gilbert’s assessment. All of these reports were discussed and read from at length in the hearing, and we can assume that they-were given consideration.

Further, the rescission panel’s suggestion that greater weight should have been given to the earlier reports appears to be no more than a disagreement with the granting panel. The 1974 report was an intake report not designed to address parole suitability factors, and the neurosis that it diagnosed was not a disorder relevant to public safety concerns. The granting panel specifically discussed the relevance of the disorder to public safety. It also considered the comments of Dr. Dai-nard, and listened to — and apparently credited — McQuillion’s explanation for those comments (ie., that Dr. Dainard had never personally counseled him or even met him).

Parole suitability determinations are made based on the gravity of the underlying offense and past convicted offenseg and “consideration^] of the public safety.” Cal-.Penal Code § 3041(b). The granting panel may have concluded that neither the opinion of a psychiatrist who had never met the prisoner nor the particular neurosis with which McQuillion was diagnosed at intake related to McQuillion’s current dangerousness, and that Dr. Nuerenberger therefore had no reason to address them in his assessment of suitability for parole. The incongruity that the rescission panel suggested existed among the psychiatric reports may well have been more apparent than real; but even if it did exist, the record is clear that the granting panel gave it full consideration, and the rescis sion panel is not now free to reassess evidence that was adequately considered at the time of the grant.

4. Failure to Complete Vocational Training

The final “good cause” finding by the rescission panel dealt with McQuillion’s lack of vocational training. The panel’s statement of this ground for rescission was terse. It wrote, “The prisoner has not' completed vocational training and did not have a prior vocational skill.” It is not entirely clear whether the rescission panel concluded that McQuillion’s lack of vocational training at the time parole was granted was a factor that was inadequately considered by the granting panel, or whether it concluded that McQuillion’s failure to acquire vocational training in the years between his grant and the date of the rescission hearing constituted good cause for rescission. Neither conclusion is supported by “some evidence.”

First, a finding that McQuillion’s vocational training (or lack thereof) was inadequately considered by the granting panel is unsupported by the record. The granting panel engaged in a very detailed discussion with McQuillion regarding his prison adjustment and activities, his accomplishments while incarcerated, and the ways in which those accomplishments might translate into job opportunities upon McQuillion’s release. McQuillion and prison officials told the panel of his work as president of the Men’s Advisory Council, his involvement in a Crisis Intervention program, his position as coordinator of the Mensa high-IQ society, and his work as a motivational speaker and participant in life-success programs. The panel was told of a series of temporary releases granted to McQuillion through his participation in such programs, and it read aloud a letter from the Prison Warden, who said that he had “observed more growth and potential in Carl McQuillion in my tenure at San Quentin than any other inmate in my career at Corrections.”

A staff counselor discussed these accomplishments and indicated that McQuillion had “done an exceptional and above-average job in his programming” and that his “institutional conduct ... has been above average to exceptional.” The counselor called McQuillion’s parole plans, which involved the establishment of a community re-entry program for prison parolees, “realistic.” McQuillion addressed at length his plans for the program, outlining his funding prospects and project goals and answering specific questions from panel members. In granting parole, the panel commended McQuillion on his in-prison adjustment and stated that this adjustment greatly influenced its suitability determination.

Given this record of specific inquiry and deliberation by the granting panel on the issue of parole plans, any suggestion by the rescission panel that an inadequate consideration of vocational training constituted good cause for rescission of the parole as improvidently granted is not supported. To the extent that the rescission panel’s finding of good cause was based on McQuiUion’s failure, as of 1979, to complete a specific vocational training rather than the career preparations that the granting panel had acknowledged, such a finding constitutes the sort of “mere disagreement” with the conclusion reached by the granting panel that does not justify rescission. The rescission panel did not — • and could not — point to some inconsistency between the evidence before the granting panel on the issue of McQuillion’s post-release skills and the conclusions that the granting panel drew on that issue.

Second, if the rescission panel was instead holding that rescission was justified because McQuillion failed, after parole was granted, to obtain any further vocational training, this, too, is an insufficient ground for good cause. The rescission panel’s questioning of McQuillion on this issue centered on whether he had completed recognized Department of Corrections vocational programs in the years since the grant of parole in 1979. In its oral report at the conclusion of its deliberations, the rescission panel noted that “[t]he prisoner had not completed any vocational requirements ... up to the date of that continued grant.” This post-grant behavior is entirely irrelevant to the determination that the rescission panel was to make — ie., “whether an improvident grant of parole occurred as a result of the May 16, 1979 hearing panel’s failure to appropriately consider [certain relevant] factors.” A 1979 panel obviously could not be expected to consider behavior that had not yet occurred. Further, there is no basis in California law for the suggestion that the rescission panel’s own determination that McQuillion should have engaged in vocational training after parole was granted to him is ah adequate ground for rescinding that parole. The discretion granted to the Board in making rescission determinations is broad, but it is cabined by the criteria listed in CaLCode Regs. tit. 15 § 2451. None of these criteria even suggests that good cause for rescission would exist merely because a prisoner who had already been found worthy of a parole date did nothing more to add to his worthiness in any particular area.

Conclusion

None of the four “good cause” grounds for the rescission of McQuillion’s parole date was supported by “some evidence” of a failure by the granting panel to adequately consider the evidence before it. McQuillion is therefore entitled to habeas relief on the grounds that his parole rescission violated due process. We reverse the decision of the district court and remand with directions that it grant the writ.

REVERSED and REMANDED.

. In 2000, section 2451 was amended, making the occurrence of "[fjundamental errors which resulted in the improvident grant of parole” a separate good cause ground, rather than subsuming it as an example of "new information indicating parole should not occur.” See Cal.Code Regs. tit. 15 § 2451(a)-(d). A California court has held that the amendment made no substantive change, and therefore the alteration is of no consequence to our analysis. See Caswell, 92 Cal.App.4th at 1026, n. 2 , 112 Cal.Rptr.2d 462 .