OPINION
FISHER, Circuit Judge.
This appeal concerns the restriction on viewing lethal injection executions imposed on the public and the press by San Quentin Institutional Procedure 770. We hold that Procedure 770 is an exaggerated, unreasonable response to prison officials’ legitimate concerns about the safety of prison staff and thereby unconstitutionally restricts the public’s First Amendment right to view executions from the moment the [871]*871condemned is escorted into the execution chamber.
FACTUAL AND PROCEDURAL BACKGROUND
In assessing the reasonableness of Procedure 770’s viewing restriction, it is helpful to understand the execution process. California executions take place in San Quentin State Prison, in a sealed octagonal room that has four large windows facing an adjoining witness area. The witness area accommodates as many as 50 people to watch the execution, including four prison guards, 12 official witnesses, 17 news media witnesses and up to five individuals requested by the prisoner. A curtain may be drawn over the windows to obscure the witnesses’ view of the execution chamber. Approximately 25 minutes before the execution is scheduled to take place, four guards escort the condemned inmate from a special overnight holding cell to the execution chamber. Though his legs are free, the condemned is handcuffed and his wrists are shackled to his waist. Upon entering the execution chamber, the condemned is laid on a gurney, to which he is secured with six straps. Next, two of the four guards leave and two medical technicians enter to insert two intravenous lines (one is redundant, in the event the main line fails). Once the intravenous lines are inserted, a saline solution begins to flow into the inmate’s veins and all staff exit the chamber. The warden then gives the order to dispense a progression of chemicals — sodium pentothal, to render the inmate unconscious, followed.by pancuroni-um bromide, to stop his lungs, and finally potassium chloride, to stop his heart.
Historically, representatives of the public and the press have been allowed to witness California’s entire execution process from start to finish. During the era of the gas chamber (beginning in 1987), that meant watching the prison staff escort the prisoner into the execution chamber (the same chamber where lethal injection executions now take place), strap him into the chair and administer the lethal gas until he was declared dead. However, for the execution of William Bonin — the first lethal injection execution in California— San Quentin officials implemented Procedure 770, which prohibits witnesses from observing the execution until after the execution team members exit the chamber. Thus, witnesses were not permitted to watch Bonin as the guards brought him into the chamber, tied him down to the gurney, inserted the intravenous lines and left him alone to await the warden’s order to dispense the chemicals. Rather, by the time prison officials opened the chamber curtains, permitting the witnesses to see inside the chamber, Bonin lay motionless on the gurney, appearing to be asleep or sedated. (Bonin had not, in fact, been sedated.) The lethal chemicals were then administered — without any announcement to the witnesses — and after several minutes, Bonin was declared dead. The witnesses, therefore, observed Bonin as he died, but were unable to see the processes leading to that point.
Following Bonin’s execution, the California First Amendment Coalition and the Society of Professional Journalists, Northern California Chapteb (“plaintiffs”), whose members attend and report on executions in California, sued in federal court and obtained a preliminary injunction prohibiting the named prison officials (“defendants”) “from preventing witness observations of executions from at least just before the time intravenous tubes are inserted to at least just after death.” Two days later, Keith Daniel Williams was executed, and witnesses were permitted to observe the insertion of the intravenous lines. We affirmed the preliminary in[872]*872junction in an unpublished memorandum disposition. Cal. First Amend. Coalition v. Calderon, 92 F.3d 1191, 1996 WL 442471 (9th Cir. Aug.5, 1996) (“Cal. First Amend. I ”).
The district court subsequently granted summary judgment to the plaintiffs and entered a slightly broader permanent injunction, directing defendants to “allow the witnesses to executions by lethal injection to view the procedure at least from the point in time just prior to the condemned being immobilized, that is[,] strapped to the gurney or other apparatus of death, until the point in time just after the prisoner dies.” Cal. First Amend. Coalition v. Calderon, 956 F.Supp. 883, 890 (N.D.Cal.1997). On appeal to this Court, we reversed the district court’s summary judgment for plaintiffs because we concluded that Procedure 770 did not — on the record then before the Court — violate whatever First Amendment right plaintiffs might have to view executions. Rather, absent substantial evidence that Procedure 770 was an exaggerated response by prison officials to asserted concerns “directly related to prison security, staff safety, and the orderly operation of the institutional procedure,” we were prepared to defer to those officials. Cal. First Amend. Coalition v. Calderon, 150 F.3d 976, 982-83 (9th Cir.1998) (“Cal. First Amend. Ill ”) (internal quotation marks omitted). We remanded to the district court “to determine whether [plaintiffs have] presented ‘substantial evidence’ that Procedure 770 represents an exaggerated response to [defendants’] security and safety concerns.” Id. at 983.1
In accordance with our remand instructions, the district court heard evidence during two days of trial regarding the prison officials’ security and safety concerns and the reasonableness of Procedure 770 in addressing them. The district court agreed that ensuring prison staff safety is a legitimate safety concern, but found on the evidence presented that “restricting public access to view lethal injection executions to a degree greater than that afforded to view lethal gas executions is an exaggerated response to defendants’ safety concerns.” Cal. First Amend. Coalition v. Woodford, No. C-96-1291-VRW, 2000 WL 33173913 at *6 (N.D.Cal. July 26, 2000); see also id. at *4-6 (discussing specific reasons for so finding). It therefore permanently enjoined the defendants “from preventing uninterrupted viewing of executions from the moment the condemned enters the execution chamber through to, and including, the time the condemned is declared dead.” Id. at *11.
As a result of the appellate proceedings, which reversed the district court’s injunction, Procedure 770 was in effect at the executions of Thomas M. Thompson on July 14, 1998; Jaturan Siripongs on February 9, 1999; Manuel Babbitt on May 4, 1999 and Darrell Keith Richard on March 15, 2000. Following the district court’s entry of its post-trial injunction, California executed Robert Lee Massie on March 27, 2001 and Stephen Wayne Anderson on January 29, 2002.
STANDARD OF REVIEW
We review de novo the constitutionality of Procedure 770, the district court’s conclusions of law and its determinations on [873]*873mixed questions of law and fact. See Neal v. Shimoda, 131 F.3d 818, 823 (9th Cir.1997). We review the district court’s findings of fact for clear error, and we will not disturb those findings without a “definite and firm conviction that a mistake has been committed.” Jones v. United States, 127 F.3d 1154, 1156 (9th Cir.1997). “If the district court’s’ account of the evidence is plausible in light of the record viewed in its entirety,’ we may not reverse even if we are convinced that had we been sitting as the trier of fact, we would have weighed the evidence differently.” Id.
ANALYSIS
The issues presented involve the balance between the State’s ability to carry out executions in a safe and orderly manner and the public’s right to be informed about how the State and its justice system implement the most serious punishment a state can exact from a criminal defendant — the penalty of death. Given our previous opinion in California First Amendment III, we do not write on a blank slate. Nonetheless, because the issues are important and Procedure 770 will govern future executions — indeed, there have been seven during the pendency of this litigation — we review the historical and constitutional context informing our analysis.
We first consider whether the public has a First Amendment right to view executions. In California First Amendment III, we assumed without deciding that the public had a “severely limited” First Amendment right to view executions. Cal. First Amend. III, 150 F.3d at 982. In Section I, we reach the question and conclude that the public does indeed enjoy a First Amendment right of access to view executions from the moment the condemned is escorted into the execution chamber. However, as we discuss in Section II — where we articulate the level of scrutiny that applies to Procedure 770 — • the public’s right of access may be reasonably limited. Finally, in Section III, we reach the merits of this case and determine that Procedure 770 is an exaggerated response to defendants’ legitimate peno-logical concerns.
/.
The First Amendment Right to View Executions
To challenge the constitutionality of Procedure 770 successfully, plaintiffs must demonstrate that they have a First Amendment interest in viewing the portion of the execution that Procedure 770 seeks to conceal. It is well-settled that the First Amendment guarantees the public — and the press — a qualified right of access to governmental proceedings.2 See Press-Enter. Co. v. Superior Court, 478 U.S. 1, 8-14, 106 S.Ct. 2735, 92 L.Ed.2d 1 (1986) (preliminary hearings) (“Press-Enter.II ”); Press-Enter. Co. v. Superior Court, 464 U.S. 501, 510-11, 104 S.Ct. 819, 78 L.Ed.2d 629 (1984) (voir dire) (“Press-Enter.I”); Globe Newspaper Co. v. Superior Court, 457 U.S. 596, 603-11, 102 S.Ct. 2613, 73 L.Ed.2d 248 (1982) (testimony of child vic[874]*874tim of sex offense); Richmond Newspapers, Inc. v. Virginia, 448 U.S. 555, 579, 100 S.Ct. 2814, 65 L.Ed.2d 973 (1980) (criminal trials). This right of access is premised on “the common understanding that ‘a major purpose of [the First] Amendment was to protect the free discussion of governmental affairs.’ ” Globe Newspaper, 457 U.S. at 604, 102 S.Ct. 2613 (quoting Mills v. Alabama, 384 U.S. 214, 218, 86 S.Ct. 1434, 16 L.Ed.2d 484 (1966)). By guaranteeing that “the individual citizen can effectively participate in and contribute to our republican system of self-government,” the First Amendment right of access ensures that “this constitutionally protected ‘discussion of governmental affairs’ is an informed one.” Id. at 604-05, 102 S.Ct. 2613 (citations omitted). Therefore, although the right of access is not enumerated in the First Amendment, it is encompassed within the Amendment as a right that is “nonetheless necessary to the enjoyment of other First Amendment rights.” Id. at 604, 102 S.Ct. 2613.
Our precedent extends the right of access to a broad range of criminal proceedings. For example, we have recognized that “the public and the press have a right of access to criminal proceedings and documents filed therein,” and we have found “no principled basis for affording greater confidentiality to post-trial documents and proceedings than is given to pretrial matters.” CBS, Inc. v. United States Dist. Court, 765 F.2d 823, 825 (9th Cir.1985). We have also found a public right of access to pretrial suppression hearings, United States v. Brooklier, 685 F.2d 1162, 1170-71 (9th Cir.1982), pretrial release proceedings and documents, Seattle Times Co. v. United States Dist. Court, 845 F.2d 1513, 1517 (9th Cir.1988), transcripts of closed hearings that occurred during the course of jury deliberations, Phoenix Newspapers, Inc. v. United States Dist. Court, 156 F.3d 940, 949 (9th Cir.1998), and plea agreements and related documents, Oregonian Publ’g Co. v. United States Dist. Court, 920 F.2d 1462, 1465-66 (9th Cir.1990).
The question before us now is how the constitutional principles that animate granting public access to governmental proceedings — particularly those comprising the process of trying, convicting and sentencing criminal defendants — carry over to the process of executing a condemned inmate within the confines of the prison. Both the Supreme Court and this circuit have held that, under the First Amendment, there are at least qualified rights of access to gather information from prison inmates and to observe some prison operations. In Pell v. Procunier, 417 U.S. 817 (1974), the Court recognized that “the conditions in this Nation’s prisons are a matter that is both newsworthy and of great public importance.” Id. at 831 n. 7, 94 S.Ct. 2800. In affirming the California prison regulation prohibiting face-to-face interviews between the press and individual inmates, the Court noted that “this regulation is not part of an attempt by the State to conceal the conditions in its prisons or to frustrate the press’ investigation and reporting of those conditions. Indeed, the record demonstrates that ... both the press and general public are accorded full opportunities to observe prison conditions.” Id. at 830, 94 S.Ct. 2800. The Court similarly stressed in Saxbe v. Washington Post Co., 417 U.S. 843, 94 S.Ct. 2811, 41 L.Ed.2d 514 (1974), that, except for the nearly identical federal regulation forbidding face-to-face interviews between the press and individual inmates, “members of the press are accorded substantial access to the federal prisons in order to observe and report the conditions they find there.” Id. at 847, 94 S.Ct. 2811. As emphasized in California First Amendment III, “the Supreme Court has never flatly held that the press has no First [875]*875Amendment right to view events inside prison walls; only that such a right is coextensive with the public’s right to the same information.” 150 F.3d at 982.3
Finally, two “complementary considerations” inform our determination that the public has a First Amendment right of access to governmental proceedings in general and executions in California in particular: (1) “whether the place and process have historically been open to the press and general publicf ]” and (2) “whether public access plays a significant positive role in the functioning of the particular process in question.” Press-Enter. II, 478 U.S. at 8-9, 106 S.Ct. 2735. The Supreme Court’s two-pronged test leads us to conclude that the public has a First Amendment right to view executions.
A. The historical tradition of public access to executions
Historically, executions were open to all comers. In England, from 1196 to 1783, the city of Tyburn hosted up to 50,000 public executions. See John Laurence, A History of Capital Punishment 177-178, 179 (1960). The Old Bailey, opposite New-gate, was the site of public executions from 1783 to 1868. See id. at 179-180. Tyburn and Newgate both drew “large and disorderly” crowds; in 1807 a crowd as large as 40,000 congregated at Newgate. See id. at 180; see also David D. Cooper, The Lesson of the Scaffold 1-26 (1974).
Executions were fully open events in the United States as well. See John D. Bes-sler, Televised Executions and the Constitution: Recognizing a First Amendment Right of Access to State Executions, 45 Fed. Comm. L.J. 355, 359-64 (1993); Neil E. Nussbaum, “Film at Eleven ... ”— Does the Press Have a Right to Attend and Videotape Executions ?, 20 N.C. Cent. L.J. 121, 122-23 (1992); Roderick C. Patrick, Hiding Death, 18 New Eng. J. on Crim. & Civ. Confinement 117, 118 (1992). California abolished public executions in 1858, moving them within prison walls, and the last “town square” execution in the United States took place in 1937. 'See Bessler, 45 Fed. Comm. L.J. at 365; Nuss-baum, 20 N.C. Cent. L.J. at 124. Approximately 500 people watched a hanging in Galena, Missouri. Nussbaum, 20 N.C. Cent. L.J. at 124. The year before, thousands observed a public hanging in Owens-boro, Kentucky. Id.
When executions were moved out of public fora and into prisons, the states implemented procedures that ensured executions would remain open to some public scrutiny. In abolishing public executions in 1858, for example, California provided that a minimum of “twelve respectable citizens” should be present at the private execution. Cal. First Amend. III, 150 F.3d at 978 (noting also that the current version of the statute, Cal.Penal Code § 3605, is virtually identical to the 1858 statute). Every state authorizing the death penalty currently requires that official witnesses be present at each execution. See Bessler, 45 Comm. L.J. at 368-[876]*87672. Further, when public executions were first abolished in America, the press was still allowed to attend. See Louis P. Massur, Rites of Execution 114-16 (Oxford University Press 1989). But see Bessler, 45 Fed. Comm. L.J. at 364 (noting that Minnesota did not permit press access to executions); Holden, 137 U.S. at 486, 11 S.Ct. 143. In California, “the press has been a constant presence” since executions were moved into prisons. Cal. First Amend. III, 150 F.3d at 978. Currently, in addition to the 12 official witnesses who attend California executions, 17 news media witnesses are also invited. Thus, there is a tradition of at least limited public access to executions. That only select members of the public attend does not erode the public nature of executions— these official witnesses act as representatives for the public at large. Cf. Richmond Newspapers, 448 U.S. at 573, 100 S.Ct. 2814 (noting that people now acquire information about trials chiefly through the media rather than first hand, and validating the media’s claim that it functions as a “surrogate! ] for the public”).
Defendants argue that the public does not have a right to view the “initial execution procedures,” but rather only the execution itself, which defendants define as beginning when the lethal chemicals start to flow. This definition, however, is simply of defendants’ own making. The public and press historically have been allowed to watch the condemned inmate enter the execution place, be attached to the execution device and then die. As we noted in California First Amendment III, before California adopted the lethal gas method of execution, witnesses were permitted to view hangings “in their entirety, from the condemned’s ascent up the gallows to the fall of the trap door.” 150 F.3d at 978. Thereafter, witnesses were also permitted to observe lethal gas executions “from the time the condemned was escorted into the gas chamber until pronouncement of death.” Id. Accordingly, historical tradition strongly supports the public’s First Amendment right to view the condemned as the guards escort him into the chamber, strap him to the gurney and insert "the intravenous lines.
B. The functional importance of public access to executions
Independent public scrutiny — made possible by the public and media witnesses to an execution — plays a significant role in the proper functioning of capital punishment. An informed public debate is critical in determining whether execution by lethal injection comports with “the evolving standards of decency which mark the progress of a maturing society.” Trop v. Dulles, 356 U.S. 86, 101, 78 S.Ct. 590, 2 L.Ed.2d 630 (1958). To determine whether lethal injection executions are fairly and humanely administered, or whether they ever can be, citizens must have reliable information about the “initial procedures,” which are invasive, possibly painful and may give rise to serious complications. Cf. Globe Newspaper, 457 U.S. at 606, 102 S.Ct. 2613 (“Public scrutiny of a criminal trial enhances the quality and safeguards the integrity of the factfinding process, with benefits to both the defendant and to society as a whole.”). This information is best gathered first-hand or from the media, which serves as the public’s surrogate. See Richmond Newspapers, 448 U.S. at 572, 100 S.Ct. 2814 (“People in an open society do not demand infallibility from their institutions, but it is difficult for them to accept what they are prohibited from observing.”). Further, “public access ... fosters an appearance of fairness, thereby heightening public respect for the judicial process.” Globe Newspaper, 457 U.S. at 606, 102 S.Ct. 2613; accord Richmond Newspapers, 448 U.S. at 572, 100 S.Ct. [877]*8772814. Finally, public observation of executions fosters the same sense of catharsis that public observation of criminal trials fosters. Although this may reflect the dark side of human nature, the Supreme Court has recognized that the public must be permitted to see justice done, lest it vent its frustration in extralegal ways. See Richmond Newspapers, 448 U.S. at 571, 100 S.Ct. 2814 (“The crucial prophylactic aspects of the administration of justice cannot function in the dark; no community catharsis can occur if justice is done in a corner or in any covert manner.” (internal quotation marks and brackets omitted)). Accordingly, the same functional concerns that drove the Court to recognize the public’s right of access to criminal trial proceedings compel us to hold that the public has a First Amendment right to view the condemned as he enters the execution chamber, is forcibly restrained and fitted with the apparatus of death.
Because there-is both an historical tradition — beginning with entirely public executions and continuing with the practice of inviting official witnesses — and a functional importance of public access to executions, both prongs of the test articulated in the Richmond Newspapers line of cases have been satisfied. We therefore hold that the public enjoys a First Amendment right to view executions from the moment the condemned is escorted into the execution chamber, including those “initial procedures” that are inextricably intertwined with the process of putting the condemned inmate to death.
II.
The Proper Level of Scrutiny
Given the public’s First Amendment right to view executions, the question then is whether Procedure 770 permissibly burdens that right by prohibiting observers from seeing all of the proceedings in the execution chamber. In evaluating defendants’ justification for their restrictive policy, we must first address the proper level of scrutiny that governs our review.
Under the public right of access line of cases, once the right of access attaches to a governmental proceeding, that right “may be overcome only by an overriding interest based on findings that closure is essential to preserve higher values and is narrowly tailored to serve that interest.” Press-Enter. II, 478 U.S. at 9, 106 S.Ct. 2735 (internal quotation marks omitted); accord Press Enter. I, 464 U.S. at 510, 104 S.Ct. 819; see also Globe Newspaper, 457 U.S. at 607, 102 S.Ct. 2613; Richmond Newspapers, 448 U.S. at 581, 100 S.Ct. 2814. As Richmond Newspapers itself recognized, however, penal institutions “by definition, are not ‘open’ or public places.” 448 U.S. at 577 n. 11, 100 S.Ct. 2814. It is a “truism that prisons are institutions where public access is generally limited.” Saxbe, 417 U.S. at 849, 94 S.Ct. 2811 (internal quotation marks omitted). Because the executions at issue here take place within prison walls, are administered by the same individuals who run San Quentin and are staffed by the same personnel who participate in the daily operations of the prison, our level of scrutiny must be guided by the line of cases addressing constitutional challenges to prison regulations, rather than by those governing access to. governmental proceedings.
The Supreme Court has formulated a “unitary, deferential standard for reviewing prisoners’ constitutional claims.” Shaw v. Murphy, 532 U.S. 223, 229, 121 S.Ct. 1475, 149 L.Ed.2d 420 (2001) (emphasis added). In general, federal courts take a “hands-off attitude towards problems of prison administration” because they “are complex and intractable and, more to the point, they are not readily [878]*878susceptible of resolution by decree. Most require expertise, comprehensive planning, and the commitment of resources, all of which are peculiarly within the province of the legislative and executive branches of government.” Procunier v. Martinez, 416 U.S. 396, 404-05, 94 S.Ct. 1800, 40 L.Ed.2d 224 (1974), limited on other grounds by Thornburgh v. Abbott, 490 U.S. 401, 109 S.Ct. 1874, 104 L.Ed.2d 459 (1989); accord Lewis v. Casey, 518 U.S. 343, 386-87, 116 S.Ct. 2174, 135 L.Ed.2d 606 (1996) (Thomas, J., concurring). Thus, in reviewing a challenge to a prison regulation that “burdens fundamental rights,” we are directed to ask whether the regulation “is ‘reasonably related’ to legitimate penological objectives, or whether it represents an ‘exaggerated response’ to those concerns.” Turner v. Safley, 482 U.S. 78, 87, 107 S.Ct. 2254, 96 L.Ed.2d 64 (1987) (quoting Pell, 417 U.S. at 827, 94 S.Ct. 2800). The “legitimate policies and goals of the corrections system” are deterrence of future crime, protection of society by quarantining criminal offenders, rehabilitation of those offenders and preservation of internal security. Pell, 417 U.S. at 822-23, 94 S.Ct. 2800. In determining whether a restriction on the exercise of rights is reasonable or exaggerated in light of those penological interests, four factors are relevant: (1) whether there is “a ‘valid, rational connection’ between the prison regulation and the legitimate governmental interest put forward to justify it”; (2) “whether there are alternative means of exercising the right that remain open to prison inmates”; (3) what “impact accommodation of the asserted constitutional right will have on guards and other inmates, and on the allocation of prison resources generally” and (4) whether there exist “ready alternatives ... that fully accommodate[ ] the prisoner’s rights at de minimis cost to valid penological interests.” Turner, 482 U.S. at 89-91, 107 S.Ct. 2254.
The Supreme Court has never applied Turner in a case such as this one, where the regulation promulgated- by prison officials is centrally concerned with restricting the rights of outsiders rather than prisoners.4 In Procunier v. Martinez, the Supreme Court held that regulations censoring outgoing prisoner mail must be justified by a showing (1) that the regulation furthers “an important or substantial governmental interest unrelated to the suppression of expression” and (2) that “the limitation of First Amendment freedoms must be no greater than is necessary or essential to the protection of the particular governmental interest involved.” 416 U.S. at 413, 94 S.Ct. 1800. The Court noted the argument that prisoners enjoy limited free speech rights, but it declined to adopt a more deferential approach on that basis, reasoning that “[wjhatever the status of a prisoner’s claim to uncensored correspondence with an outsider, it is plain that the latter’s interest is grounded in the First Amendment’s guarantee of freedom of speech.” Id. at 408, 94 S.Ct. 1800. The Court subsequently determined that, in light of Turner, the standard of review advocated in Martinez would be restricted to regulations concerning outgoing prisoner correspondence. Abbott, 490 U.S. at 413, 109 S.Ct. 1874.5 The Court also rec[879]*879ognized, however, that “[w]here, as in Martinez, the nature of the asserted governmental interest is such as to require a lesser degree of case-by-case discretion, a closer fit between the regulation and the purpose it selves may safely be required.” Id. at 412, 94 S.Ct. 1800 (emphasis added).
In California First Amendment III, we decided that the district court should apply Pell’s deferential “exaggerated response” test — the test the Supreme Court originally formulated to determine whether prisoners’ rights had been unconstitutionally restricted and that is incorporated in Turner’s reasonableness standard. See Cal. First Amend. III, 150 F.3d at 983 (remanding case and quoting language of Pell, 417 U.S. at 827, 94 S.Ct. 2800); see also Tuner, 482 U.S. at 87, 107 S.Ct. 2254 (stating inquiry as “whether a prison regulation that burdens fundamental rights is ‘reasonably related’ to legitimate penological objectives, or whether it represents an ‘exaggerated response’ to those concerns”). We are bound by our previous panel decision. See United States v. Scrivner, 189 F.3d 825, 827 (9th Cir.1999) (“The ‘law of the case’ doctrine provides that one panel of an appellate court will not as a general rule reconsider questions which another panel has decided on a prior appeal in the same case.” (citations and internal quotation marks omitted)). Therefore, we will apply the “exaggerated response” test here. However, our prior panel did not consider the relevance of the Court’s statement in Abbott that regulations that are broad in nature and do not require substantial case-by-case discretion must exhibit a closer fit to their purported purposes.
As we shall see, Procedure 770 is by its nature similar in material respects to the outgoing correspondence regulation in Martinez — -in every case it automatically censors a communication to outsiders that poses no serious threat to security inside the prison. Cf. Pell, 417 U.S. at 830, 94 S.Ct. 2800 (noting that the challenged regulation was “not part of an attempt by the state to conceal the conditions in its prisons or to frustrate the press’ investigation and reporting of those conditions” and that the press and public were “accorded full opportunities to observe prison conditions”). Accordingly, as we review the district court’s findings, we shall do so with the Supreme Court’s own qualifier in mind- — requiring a “closer fit” between Procedure 770 and defendants’ legitimate penological interest in the security of the execution team.
III.
Procedure 770 is Not Reasonably Related to a Legitimate Penological Interest
Defendants try to short-circuit our review, arguing that we are bound by the law of the case to reverse the district court’s conclusion that Procedure 770 is unconstitutional. Isolating our statement that “we hold that Procedure 770 does not violate the First Amendment rights of the either the press or the public[,]” 150 F.3d at 982, defendants insist that California First Amendment III predetermined that Procedure 770 does not violate the First Amendment. Not so. California First Amendment III made clear that it was ruling on the limited record before it, a record it expressly recognized was subject to further development. Id. at 983 (“We do not have substantial evidence indicating an exaggerated response here ....’’ (emphasis added)). Indeed, we could not have been conclusively deciding the constitutional question, because we delegated that task in the first instance to the district [880]*880court, remanding with the instruction to determine as a matter of fact whether Procedure 770 represented an exaggerated response to the defendants’ security concerns. Id. The record we now review is different from the summary judgment record that was before us in California First Amendment III. Whereas the California First Amendment III panel had before it declarations, submitted in support of the parties’ summary judgment papers, we now have before us the record of the bench trial, where the declarants testified live and were subjected to extensive cross-examination. It is the superseding record upon which we determine Procedure 770’s constitutionality.
We turn now to the central issue of whether Procedure 770 is a reasonable or an exaggerated response to legitimate penological interests- — defendants’ concern for staff safety and institutional security. Cal. First Amend. III, 150 F.3d at 979. Specifically, defendants have attempted to justify Procedure 770 as necessary to protect the anonymity of the execution staff, who, according to defendants, face possible retaliation from prisoners in the general population or — as defendants assert for the first time in their reply brief on appeal — from death penalty opponents. The district court agreed that “ensuring staff safety is a legitimate safety concern,” but found that defendants had presented no evidence that execution staff had ever been, or were ever likely to be, publicly identified or attacked. Moreover, the district court found — consistent with Turner — that there was a ready alternative to Procedure 770 that would fully accommodate defendants’ concerns and plaintiffs’ First Amendment interests at a de minim-is cost. Defendants assert that the suggested alternative — -that execution team members wear surgical garb to conceal their identities from the witnesses — would be ineffective and unsafe, but the district court found otherwise. Finally, the district court found that “Procedure 770 was motivated, at least in part, by a concern that the strapping of a condemned inmate, the injection of intravenous lines or other aspects of a lethal injection execution would be perceived as brutal by the public and thus was, to that extent, prompted by considerations other than legitimate concerns for prison personnel safety.” Cal. First Amend. Coalition v. Woodford, 2000 WL 33173913 at *6. As we explain, the district court’s factual findings are fully supported by the evidence and are not clearly erroneous. In light of those findings, we agree that Procedure 770 is an exaggerated response to defendants’ concerns about the safety of prison personnel.
A. Valid, rational connection between the prison regulation and the legitimate governmental interest put forward to justify it
As its third reason for striking down Procedure 770, the district court found that the procedure was motivated at least in part by a desire to conceal the harsh reality of executions from the public. We find no clear error in the district court’s factual finding. However, even if we were to give defendants the benefit of the doubt by assuming that Procedure 770 is “neutral” and reflects their legitimate concern for the security of prison personnel, Abbott, 490 U.S. at 415, 109 S.Ct. 1874, we must still conclude that it fails constitutional scrutiny. Based on the evidence in the record, defendants’ fear that execution team members will be publicly identified and retaliated against is an overreaction, supported only by questionable speculation. Moreover, Procedure 770 does not rationally address the purported problem. See Morrison v. Hall, 261 F.3d 896, 902 (9th Cir.2001) (explaining that the first Turner prong requires the reviewing court [881]*881to “(1) determine whether the defendant’s regulation is legitimate and neutral; and (2) assess whether there is a rational relationship between the governmental objective and the regulation” (internal quotation marks omitted)).
Ordinarily, “even in the absence of institution-specific or general social science evidence, as long as it is plausible that prison officials believed the policy would further a legitimate objective, the governmental defendant should prevail on Turner’s first prong.” Frost v. Symington, 197 F.3d 348, 355 (9th Cir.1999). However, if the regulation’s challengers “present[ ] sufficient (pre or post) trial evidence that refutes the common-sense connection between a legitimate objective and a prison regulation, ... the state must present enough counter-evidence to show that the connection is not so ‘remote as to render the policy arbitrary or irrational.’ ” Id. (quoting Mauro v. Arpaio, 188 F.3d 1054, 1060 (9th Cir.1999)).
For several reasons, plaintiffs have refuted the “common-sense” connection between Procedure 770 and the safety of execution staff and demonstrated that the connection is too “remote” to satisfy Turner’s first prong. First, witnesses were as likely to identify execution team members under the lethal gas regime (and presumably during hangings, earlier) as under the lethal injection regime. During lethal gas executions, execution team members were fully visible for at least one minute. Yet Warden Vasquez testified at trial that it takes only a few seconds to catalog an individual’s defining physical features. Indeed, during Robert Harris’ execution, witnesses saw the execution team escort Harris into the chamber more than once. Although lethal injection executions require 'the team to be present in the chamber for well more than one minute, the district court made three relevant findings that diminish the importance of this fact. First, the time for preparing inmates for execution by lethal injection has shortened with each execution. During the Bonin execution, 17 minutes passed between the time Bonin entered the execution chamber and the time the saline solution was set and running, which is the point at which execution staff leave the chamber. Similarly, during the Williams execution, it took 17 minutes to escort Williams into the chamber, secure him to the gurney and insert the intravenous lines. For the Siripongs execution, the same process took only 10 minutes, and it took about six minutes for the Babbitt execution. Second, it takes only about one minute for ordinary medical personnel to insert an intravenous line. Thus, the individuals who insert the intravenous lines into the inmates should not have to be present in the room for much longer than that amount of time. Third, execution team members have their backs turned to the witnesses for most of their time in the chamber. Thus, although the duration of exposure is greater, the evidence supports the district court’s finding that the difference between the potential for identification during lethal gas and lethal injection executions is not material. Cf. Morrison, 261 F.3d at 902 (rejecting prison’s argument that mail restriction was necessary to prevent introduction of contraband because “defendants have failed to present any evidence that the risk of contraband in first or second class mail is any lower than the risk of contraband in mail that is sent bulk .rate, third, or fourth class”).
Second, as the district court found, Procedure 770 contains loopholes that undermine its rationality and the credibility of defendants’ concerns for anonymity. Most significantly, even with Procedure 770 in place, the condemned inmate — who arguably has the strongest motive to seek retaliation — has the opportunity to disclose [882]*882the execution team members’ identities to the outside world. Warden Woodford testified that “[mjembers of the execution team are with the condemned inmate continuously from 6:00 p.m. the day before the execution until the execution is completed after midnight.” During this same period of time, the condemned inmate may freely call his attorney, friends and family. Even though the inmate knows he is interacting with the same individuals who will ultimately put him to death, there is nothing to prevent him from revealing the identities — or at least detailed physical descriptions- — of these execution team members. A similar loophole exists with respect to the five to ten prison guards who stand in the witness area during the entire execution. These guards have always been present and their physical features have never been concealed, even under Procedure 770.
Third, even assuming an execution team member were identified by a witness, the notion of retaliation is pure speculation. No execution team member has ever been threatened or harmed by an inmate or by anyone outside the prison because of his participation in an execution. The guards standing in the witness area — who have always been visible to the witnesses for a prolonged period of time — have never been threatened or harmed in any way. Additionally, as the district court pointed out, there are also “many high-profile individuals whose participation in the implementation of executions is essential, including the warden, the governor and judges of the courts who reject the condemned’s appeals.” If defendants’ professed concern about politically motivated retaliation by death penalty opponents is to be credited, retaliation is at least as likely to be directed against these decision makers as against low level execution staff. Yet defendants presented no evidence of such retaliation. Cf. Pell, 417 U.S. at 831, 94 S.Ct. 2800 (noting that restriction on media contact with prisoners was imposed in response to a violent episode and that prisoners who became virtual public figures often became the source of severe disciplinary problems); Turner, 482 U.S. at 91, 107 S.Ct. 2254 (identifying specific security problems involving correspondence between inmate gang members). Although prison officials may pass regulations in anticipation of security problems, Casey v. Lewis, 4 F.3d 1516, 1521 (9th Cir.1993); Friedman v. Arizona, 912 F.2d S28, 332-33 (9th Cir.1990), they must at a minimum supply some evidence that such potential problems are real, not imagined. Here, defendants fail to explain why we should disregard the history of safety surrounding those officials whose identities have always been publicly known.
In sum, defendants’ fear that execution team members will be identified and retaliated against is speculative and contradicted by history. This is relevant not only to our discussion of the first Tuner factor, but also to our conclusion in Part II, that Procedure 770 is analogous to the outgoing correspondence restriction in Martinez— neither restriction reasonably implicates security inside the prison. See Abbott, 490 U.S. at 411-12, 109 S.Ct. 1874 (noting that outgoing correspondence by its very nature does not “pose a serious threat to prison order and security”; even correspondence that “magnifies grievances or contains inflammatory racial views cannot reasonably be expected to present a danger to the community inside the prison” (emphasis in original)). Moreover, the connection between execution team safety and Procedure 770 is questionable in light of Procedure 770’s loopholes. The Turner “reasonableness standard is not toothless,” Abbott, 490 U.S. at 414, 109 S.Ct. 1874, particularly as applied with Martinez in mind. Thus, although we credit defen[883]*883dants’ safety concerns to a point, we conclude that defendants failed to establish the requisite rational link between execution team safety and Procedure 770.
We shall consider the remaining Turner factors, although the first factor is arguably dispositive. Compare Morrison, 261 F.3d at 904 (holding that once the first Turner factor is resolved in either the plaintiffs or defendant’s favor, other factors need not be considered), and Prison Legal News v. Cook, 238 F.3d 1145, 1151 (9th Cir.2001) (“The rational relationship of the Turner standard is a sine qua non.”), with Mayweathers v. Newland, 258 F.3d 930, 938 (9th Cir.2001) (holding that plaintiffs had a strong likelihood of success on the merits of a prison regulation challenge even though the first Turner factor favored governmental defendants).
B. Alternative means of exercising the right
Of great significance, Procedure 770 prevents the public from having any firsthand knowledge of the events that take place behind the curtain. The witnesses may not observe the condemned inmate’s demeanor as he enters the execution chamber, has intravenous lines inserted into his body and realizes that the saline solution has begun to flow. The witnesses may not observe the manner of the guards as they restrain the prisoner. Indeed, a 1996 memorandum from Warden Calderon, articulating his concerns about the present lawsuit, suggests — as the district court found — that Procedure 770 reflects defendants’ interest in avoiding any public perception that 'executions involve the use of excessive force:
In the event of a hostile and combative inmate, it -will be necessary to use additional force and staff to subdue, escort and secure the inmate to the gurney. It is important that we are perceived as using only the minimal amount of force necessary to accomplish the task. In reality, it may take a great deal of force. This would most certainly be misinterpreted by the media and inmate invited witnesses who don’t appreciate the situation we are faced with.
Because witnesses cannot see first-hand the manner in which the intravenous lines are injected, they will not be privy to any complications that may arise during this initial, invasive procedure. Consequently, the public will be forced to rely on the same prison officials who are responsible for administering the execution to disclose and provide information about any difficulties with the procedure. For example, during the Bonin execution, the execution staff encountered difficulty in inserting the intravenous lines. The witnesses were unable to see for themselves whether Bonin experienced pain as a result of the complications, or how severe the problem was, and instead had to rely on the prison officials’ version of events. Procedure 770 thus entirely eliminates independent, public eyewitness observation of several crucial steps of the execution process.
By way of contrast, in Pell the Court stressed that the inmates had at least two alternative means of communication with the media. First, “the medium of written correspondence affords inmates an open and substantially unimpeded channel for communication with persons outside the prison, including representatives of the news media.” 417 U.S. at 824, 94 S.Ct. 2800. Second, inmates enjoyed “an unrestricted opportunity to communicate with the press” through their permitted visitors (families, friends, clergy or attorneys). Id. at 825, 94 S.Ct. 2800. Here, once the curtain is drawn, there is no alternative to accommodate the public’s First Amendment right to view the execution.
[884]*884We consider this factor to be particularly relevant because of our conclusion that it is critical for the public to be reliably informed about the lethal injection method of execution. In California First Amendment III, we noted that “[ejyewitness media reports of the first lethal gas executions sparked public debate over this form of execution and the death penalty itself.” 150 F.3d at 978-79. An informed public debate is the main purpose for granting a right of access to governmental proceedings. Prison officials simply do not have the same incentives to describe fully the potential shortcomings of lethal injection executions. As Warden Calderon’s memo demonstrates, a prison official’s perception of the execution process may be vastly different — and markedly less critical— than that of the public.
C. The impact that accommodation would have on guards, other inmates and the allocation of prison resources
In Turner, the Court explained that “[w]hen accommodation of an asserted right will have a significant’ ripple effect’ on fellow inmates or on prison staff, courts should be particularly deferential to the informed discretion of corrections officials.” 482 U.S. at 90, 107 S.Ct. 2254. This factor does not weigh in favor of deference here. Abolition of Procedure 770’s viewing restriction is straightforward and will not unduly strain prison resources. As discussed above, there is no evidence that execution staff will be retaliated against with threats or violence; and even with Procedure 770 in place, the public may have the opportunity to learn the identities of the execution staff. Defendants suggest that qualified prison personnel might be unwilling to volunteer for execution duty because of a perceived risk that they will be publicly identified. Again, that speculation is not borne out by actual events. None of the team members involved in the Williams execution withdrew from duty upon being informed that Procedure 770 would not be fully implemented. Although Warden Calderon testified that he had to convince these individuals to remain on the execution team, nothing in the record suggests that other qualified individuals were not available to serve. Defendants have not suggested— much less proffered evidence — that they were unable to locate qualified individuals to staff the Massie and Anderson executions. Again, we will not accord defendants deference on the basis of mere speculation. The third Turner factor does not weigh in favor of enforcing Procedure 770.
D. The presence or absence of ready, low-cost alternatives
The fourth Turner factor also weighs strongly against the reasonableness of Procedure 770: there exists a ready, low-cost alternative that would fully accommodate the public’s First Amendment right of access and adequately address defendants’ security concerns as well. Based on the evidence presented, the district, court found that “[t]he use of surgical garb is a practical alternative to restricting access to witness lethal injection executions in order to conceal the identity of such execution staff should security concerns warrant such concealment.” Cal. First Amend. Coalition v. Woodford, 2000 WL 33173913 at *6. This finding is not clearly erroneous.
Trial testimony supported the district court’s finding that “masks are an effective means of concealing the identity of the wearer.” Id. at *5. The testimony of a newspaper reporter confirmed the commonsense understanding that an individual wearing a surgical cap, mask and gloves— which, when worn together, cover the fore[885]*885head, nose, cheeks, mouth and hands— cannot be identified with any meaningful degree of specificity.
Contrary to defendants’ assertions, trial testimony supported the district court’s finding that the “[u]se of surgical garb would ... not impede execution staff in performing executions.” Id. Plaintiffs’ expert witness, Dr. Shavelson, an emergency room physician, testified that surgical garb did not hamper his abilities to communicate with others or to carry out medical procedures, including insertion of an intravenous line, and that, in his extensive experience dealing with struggling patients who sometimes have to be restrained, he had never seen a mask dislodged. The district court also enjoyed the benefit of in-court demonstrations of how surgical garb might be used during an execution. The court was entitled to credit Dr. Shav-elson’s testimony over that of Wardens Calderon and Vasquez, both of whom believe that a surgical mask will be dislodged during a struggle with an inmate. Unlike Dr. Shavelson, neither warden has any personal experience with such masks. Defendants’ position that surgical garb will compromise the security of the execution team is further weakened by the fact that there are up to 11 guards readily available to deal with a struggling inmate. Finally, in the six executions that Warden Calderon had witnessed, none of the condemned inmates struggled, and defendants did not introduce any evidence that condemned inmates have ever attempted to fight off execution staff.
We are not persuaded by Warden Wood-ford’s contention that the use of masks might interfere with the bond that develops between the execution staff and the condemned inmate. Warden Woodford failed to explain how bonding would be disrupted if prison personnel refrained from wearing any surgical attire until just before they escorted the condemned into the execution chambér. Nor is there any evidence that explaining the procedure to the prisoner would be inadequate for maintaining any bond.
For these reasons, we find no error in the district court’s finding that surgical garb can be worn without compromising the guards’ security and efficacy and is thus an available, effective alternative to Procedure 770.
CONCLUSION
In sum, each Turner factor weighs against our concluding that Procedure 770 is reasonably related to the prison officials’ legitimate interest in the safety of prison staff and instead demonstrates that the viewing restrictions are an exaggerated response, as the district court found. We agree with the district court and hold that Procedure 770 is unconstitutional.6 Ac[886]*886cordingly, we affirm the district court’s permanent injunction, which prohibits defendants “from preventing uninterrupted viewing of executions from the moment the condemned enters the execution chamber through, to and including, the time the condemned is declared dead.”
AFFIRMED.