concurring in part and dissenting in part:
I agree that the BOP regulation prohibiting electric and electronic instruments is a reasonable interpretation of the Zimmer Amendment and therefore does not violate the Administrative Procedure Act. I also have no doubt that the government can constitutionally restrict the use and possession of electric guitars in federal prisons. If the government had defended the electronic-music ban as reasonably necessary to promote internal order, for example, or even to promote a more efficient allocation of prison resources, then the Zimmer Amendment would easily qualify *236as a constitutional exercise of Congress’s authority to manage the federal prisons. Or if the government had denied funds only for the purchase of electric and electronic instruments, that too would easily pass constitutional muster, since the government has no obligation to subsidize the exercise of First Amendment rights. The government, however, has done neither of these things. This court nevertheless sustains the amendment’s constitutionality on the basis of a broad — indeed limitless— legal principle under which prison officials may ban not only the use and possession of electric guitars, but also the exercise of virtually any other constitutional right that requires electricity, guard supervision, or other prison resources — including the sending and receiving of inmate mail, and even the use and possession of books and magazines. Because the court’s holding conflicts with binding Supreme Court and circuit precedent, and because its alternative basis for affirmance was neither argued nor even mentioned by the BOP, I dissent.
I.
In Turner v. Safley, 482 U.S. 78, 107 S.Ct. 2254, 96 L.Ed.2d 64 (1987), which governs prisoners’ constitutional challenges, the Supreme Court struck a balance between two competing principles. Safley first recognizes that “[p]rison walls do not form a barrier separating prison inmates from the protections of the Constitution.” Id. at 84, 107 S.Ct. at 2258. In the case before us, the BOP concedes— and this court agrees — that the Zimmer Amendment’s electric- and electronic-music ban does restrict prisoners’ First Amendment rights. Appellee’s Br. at 8; cf. Ward v. Rock Against Racism, 491 U.S. 781, 109 S.Ct. 2746, 105 L.Ed.2d 661 (1989) (employing First Amendment scrutiny to analyze New York City’s sound-amplification controls for a music concert). At the same time, however, Safley recognizes the competing principle that “ ‘prison administrators ..., and not the courts, [must] make the difficult judgments concerning institutional operations,’ ” and that “[subjecting the day-to-day judgments of prison officials to an inflexible strict scrutiny analysis would seriously hamper their ability to anticipate security problems and to adopt innovative solutions to the intractable problems of prison administration.” Safley, 482 U.S. at 89, 107 S.Ct. at 2262 (quoting Jones v. North Carolina Prisoners’ Union, 433 U.S. 119, 128, 97 S.Ct. 2532, 2539, 53 L.Ed.2d 629 (1977)). Safley therefore establishes a “unitary, deferential standard for reviewing prisoners’ constitutional claims: ‘[W]hen a prison regulation impinges on inmates’ constitutional rights, the regulation is valid if it is reasonably related to legitimate penological interests.’ ” Shaw v. Murphy, 532 U.S. 223, 229, 121 S.Ct. 1475, 1479, 149 L.Ed.2d 420 (2001) (quoting Safley, 482 U.S. at 89, 107 S.Ct. at 2261-62). To evaluate a prison regulation’s reasonableness, Safley directs courts to consider four factors:
[F]irst, whether the restriction bears a “valid, rational connection” to the “legitimate governmental interest put forward to justify it,” such that the “asserted goal is [not] so remote as to render the policy arbitrary or irrational”; second, whether inmates retain alternative means of exercising the circumscribed right; third, the costs that accommodating the right imposes on other inmates, guards, and prison resources generally; and fourth, whether there are alternatives to the regulation that “fully accommodate[] the prisoners’ rights at de minimis cost to valid penological interests.”
Amatel v. Reno, 156 F.3d 192, 196 (D.C.Cir.1998) (quoting Safley, 482 U.S. at *23789-91, 107 S.Ct. at 2261-63) (citations omitted) (alterations in original).
In sustaining the Zimmer Amendment’s constitutionality, this court adopts a theory that effectively eviscerates Safley’s premise that prisoners generally retain their constitutional rights, as well as the decision’s carefully constructed four-part test for evaluating the constitutionality of prison regulations that restrict those rights. According to the court, because “the BOP has simply chosen not to subsidize inmates’ use or possession of a class of instruments requiring the expenditure of funds for electricity and care,” the BOP’s regulation “does not implicate the appellants’ First Amendment rights,” and the court “therefore need not invoke the four factor analysis the United States Supreme Court established in Turner v. Saf-ley.” Maj. Op. at 232-33. I would have no problem with this conclusion if the Zimmer Amendment were limited to denying funds for purchasing electric guitars. But the amendment also prohibits prisoners from using and possessing their own guitars, and there is an important difference between a regulation that denies funds to buy guitars and one that denies funds for electricity or other infrastructural resources necessary for the exercise of constitutional rights. All prison activities depend on an infrastructure of basic resources, such as electricity, guard supervision, and space, which the government both necessarily pays for and controls. If the government can “cho[ose] not to subsidize inmates’ use or possession of a class of instruments requiring the expenditure of funds for electricity and care,” Maj. Op. at 233, then the government could just as easily choose not to “subsidize” inmates’ use or possession of books, by withdrawing funds for electricity for light to read by, or it could decide not to “subsidize” inmate correspondence, by withdrawing funds for the distribution of prison mail.
To be sure, the costs associated with these various activities may differ, but as the court concedes, we have no evidentiary record to tell us how the costs of allowing prisoners to use and possess electric guitars compare with the costs of providing sufficient reading light or delivering prison mail. Maj. Op. at 234. In any event, the court’s rationale for sustaining the Zim-mer Amendment makes no distinction between activities that require large expenditures and those that do not. Indeed, by upholding the BOP’s ban on electric and electronic instruments as a reasonable interpretation of the Zimmer Amendment, which merely denies funding for the use or possession of such instruments, the court itself recognizes the proposition that an appropriations law denying funding for certain activities generally amounts to a substantive ban on those activities, regardless of the amount of funding involved. Maj. Op. at 232; see Robertson v. Seattle Audubon Soc’y, 503 U.S. 429, 440, 112 S.Ct. 1407, 1414, 118 L.Ed.2d 73 (1992) (“Congress ... may amend substantive law in an appropriations statute, as long as it does so clearly.”); Envtl. Def Ctr. v. Babbitt, 73 F.3d 867, 871-72 (9th Cir.1995) (noting that the “use of any government resources — -whether salaries, employees, paper, or buildings — ... entail[s] government expenditure,” and that “[t]he government cannot make expenditures, and therefore cannot act, other than by appropriation”) (emphasis added).
In the end, almost any restriction of prisoners’ constitutional rights can be recast as a ban on funding for those rights. But as this court observed in Amatel, in prisons, “[w]here the government absolutely monopolizes the means of speech or controls a bottleneck, ... a refusal to fund functions the same as an outright ban.” 156 F.3d at 194 n. 1. That the government “monopolizes” certain “means of *238speech” in prisons is what sets this case apart from Regan v. Taxation With Representation, 461 U.S. 540, 103 S.Ct. 1997, 76 L.Ed.2d 129 (1983). See Maj. Op. at 233-34. In Regan, the Supreme Court upheld a restriction on lobbying activities for tax-exempt charities on the ground that “a legislature’s decision not to subsidize the exercise of a fundamental right does not infringe the right, and thus is not subject to strict scrutiny.” Id. at 549, 103 S.Ct. at 2003. Describing the governing principle as “simple,” the Court explained that “ ‘although government may not place obstacles in the path of a [person’s] exercise of ... freedom of [speech], it need not remove those not of its own creation.’ ” Id. at 549-50, 103 S.Ct. at 2003 (quoting Harris v. McRae, 448 U.S. 297, 316, 100 S.Ct. 2671, 2687-88, 65 L.Ed.2d 784 (1980)). In other words, when the government denies tax-exempt status to organizations that engage in lobbying, those organizations remain free to continue their lobbying activities — the First Amendment expression in question. See Taxation With Representation, 461 U.S. at 545, 103 S.Ct. at 2000-01. Similarly, when the government denies tax deductions for lobbying activities, taxpayers remain free to engage in political advocacy, see Cammarano v. United States, 358 U.S. 498, 79 S.Ct. 524, 3 L.Ed.2d 462 (1959), and when the government denies public funding to candidates for public office, those candidates remain free to seek election, see Buckley v. Valeo, 424 U.S. 1, 90-107, 96 S.Ct. 612, 668-77, 46 L.Ed.2d 659 (1976) (per curiam).
Even in the prison context, if the government denies federal funds for purchasing magazines, books, stationery, or even electric guitars, it has “placefd] no governmental obstacle in the path” of prisoners seeking to read, write, or play, since they remain free to purchase those items on their own. Harris v. McRae, 448 U.S. 297, 315, 100 S.Ct. 2671, 2687, 65 L.Ed.2d 784 (1980). This is not so when the government denies funds for prisoners’ use of prison-controlled infrastructural resources to exercise constitutional rights. Because prisons absolutely control the electricity supply and storage space, as well as other infrastructural resources, such as mail distribution systems, denying federal funds for those resources is quite different from a refusal to “subsidize” First Amendment rights. Prisoners can neither generate their own electricity, nor construct their own storage space, nor deliver their own mail. Denying access to these resources therefore places an intractable “governmental obstacle in the path” of prisoners seeking to exercise their rights.
If this court’s subsidy rationale were correct, the reasoning of Amatel, Safley, and other like cases would have been quite different. Amatel dealt with an appropriations rider that barred the use of federal funds to distribute sexually explicit materials in prisons. Observing that the prison distribution system was a “bottleneck” under the government’s control, and that there was no suggestion that prisoners might be able to “obtain[ ] such material at their own expense” outside of that distribution system, we treated the government’s refusal to spend money on the distribution of such material as an “outright ban” on. sexually explicit materials in prison. Amatel, 156 F.3d at 194 n. 1. But if Congress’s decision to eliminate expenditures necessary for the exercise of First Amendment rights implicates no First Amendment concerns, as this court now holds, then we need not have spent pages running through the Safley factors to determine whether the ban on sexually explicit materials was reasonably related to the government’s interest in rehabilitating prisoners; we could have ended the opinion after the statement of facts. The same is true for Safley, part of which sustained *239restrictions on inmate-to-inmate correspondence, and Thornburgh v. Abbott, 490 U.S. 401, 109 S.Ct. 1874, 104 L.Ed.2d 459 (1989), which sustained restrictions on certain book and magazine subscriptions. Had the Government defended those restrictions on the theory that they merely eliminated “subsidies” for First Amendment rights, the Court’s careful application of the Safley factors would have been entirely unnecessary. Indeed, under this court’s holding, Missouri could now reenact the prohibition on inmate marriage that the Supreme Court struck down in the remainder of Safley simply by recasting it as a ban on funding for inmate marriage, since inmate weddings no doubt require expenditures for guard supervision and for electricity to light the wedding venue.
Of course, the government need not provide any and all resources necessary for prisoners to engage in constitutionally protected activities, but Safley teaches that if the government wants to cut back on infrastructural resources necessary for the enjoyment of constitutional rights, it must explain how the cutback is “reasonably related” to its “legitimate penological interest” in ensuring a more efficient allocation of BOP resources and show how the cutback satisfies the four-part Safley test. In other words, Congress may not bypass Safley through the simple expedient of using an appropriations law that functions, in effect, as an outright ban.
II.
Perhaps recognizing the weakness in its attempted end-run around Safley and justifiably dubious that this court might actually accept it, the BOP devotes most of its brief to defending the Zimmer Amendment on the ground that it does in fact satisfy Safley, arguing — apparently for the first time in litigation of this kind, see Kimber-lin v. Dep’t of Justice, 150 F.Supp.2d 36, 44 (D.D.C.2001)—that the government’s interest in enhancing the punitive and deterrent value of prison can justify a regulation that limits prisoners’ constitutional rights. Although both the district court and the concurring opinion embrace it, this argument is as flawed as it is novel.
Because punishment and deterrence are unquestionably among the fundamental “goals of the penal function in the criminal justice system,” Rhodes v. Chapman, 452 U.S. 337, 352, 101 S.Ct. 2392, 2402, 69 L.Ed.2d 59 (1981), I have no doubt that the government may enhance prison’s punitive and deterrent value by eliminating weightlifting equipment, in-cell coffee pots, and other “perks” not protected by the First Amendment. See Omnibus Consolidated Appropriations Act of 1997, Pub.L. No. 104-208, § 611, 110 Stat. 3009, 3009-66 (1996). The BOP’s defense of the Zim-mer Amendment, however, presents a very different question: Does the goal of enhancing the punitive and deterrent value of prison by making prison conditions more onerous justify limiting prisoners’ constitutional rights? As long as Safley is the law — that is, as long as prisoners generally retain their constitutional rights — the answer must be no, for there is no discerna-ble limit to the government’s ability to invoke punishment or deterrence as reasons for adopting regulations that restrict constitutional rights. The BOP’s rationale for banning electric guitars could also justify banning all musical instruments, or all music, or even all books, including the Bible and the Koran, on the ground that denying these “perks” will make prison more onerous and “more of a place of deterrence and punishment.” Concurring Op. at 236 (internal quotation marks and citation omitted). Under this theory, the government could, subject only to whatever limitations the Eighth Amendment im*240poses, reduce prisons to places of virtually silent, solitary confinement, where prisoners may not read, write, or engage in any other expressive activity.
The BOP’s theory conflicts with Safley in two additional respects. First, Safley does not contemplate punishment as one of the interests that justifies restrictions on prisoners’ constitutional rights. Its deferential standard of review is designed only to respond to the fact that “[rjunning a prison is an inordinately difficult undertaking that requires expertise, planning, and the commitment of resources, all of which are peculiarly within the province of the legislative and executive branches of government.” 482 U.S. at 84-85, 107 S.Ct. at 2259; see also Amatel, 156 F.3d at 196 (“[B]y their nature, [governmental institutions such as prisons] must allow regulation more intrusive than what may lawfully apply to the general public. In these environments, the government is permitted to balance constitutional rights against institutional efficiency in ways it may not ordinarily do.”) (citations omitted). The type of “legitimate penological interests” to which Safley refers are administrative matters such as “the preservation of internal order or discipline, the maintenance of institutional security against escape or unauthorized entry, and the rehabilitation of the prisoners,” all of which are particularly within prison officials’ expertise. Procunier v. Martinez, 416 U.S. 396, 412, 94 S.Ct. 1800, 1811, 40 L.Ed.2d 224 (1974). The BOP’s novel punishment rationale has nothing whatsoever to do with these operational, security, and management concerns. And while it is possible that a prison regulation might not be rationally related to such concerns, see, e.g., Safley, 482 U.S. at 97-98, 107 S.Ct. at 2265-66 (holding a ban on inmate marriage to be insufficiently related to the state’s asserted goals of promoting security and prisoner rehabilitation), regulations that deprive prisoners of their constitutional rights will always be rationally related to the goal of making prison more miserable. Banishing electric guitars from prison will certainly make prison less pleasant, but banishing all musical instruments — or even all music— would make it even less pleasant still, as would banning books, magazines, and letters from family and friends. While it is true that this case involves a relatively minor incursion on prisoners’ constitutional rights, the rationale the BOP offers, and that the district court accepted, would justify all of these far more significant constitutional deprivations.
Second, the remaining factors that Saf-ley instructs us to consider in evaluating the “reasonableness” of a prison regulation that impinges on constitutional rights make little sense when the government’s only reasons for the regulation are punishment and deterrence. All of those factors — “whether inmates retain alternative means of exercising the circumscribed right,” “the costs that accommodating the right imposes on other inmates, guards, and prison resources generally,” and “whether there are alternatives to the regulation that ‘fully accommodate[ ] the prisoners’ rights at de minimis cost to valid penological interests,’ ” Amatel, 156 F.3d at 196 (quoting Safley, 482 U.S. at 90-91, 107 S.Ct. at 2262-63) — are designed to ensure that legitimate management and security considerations do not unnecessarily limit prisoners’ constitutional rights. But if the government’s very purpose is to make prison more onerous, then why ask whether there are alternative means of expression available or less costly ways to accommodate the constitutional rights at issue? After all, the greater the deprivation, the stronger the connection between the restriction and the government’s goal of punishment and deterrence.
*241HI.
This brings me finally to the court’s alternative holding. Unable to accept the BOP’s assertion of punishment and deterrence as legitimate penological interests for purposes of the Safley test, and perhaps aware of the flaws in its primary holding that Safley does not apply at all, the court offers its own Safley-based reason for sustaining the Zimmer Amendment: that the regulation is “reasonably related” to the “legitimate penological interest” of conserving correctional funds. Maj. Op. at 233-34. Had the BOP made this argument, it might well have provided a valid basis for upholding the Zimmer Amendment. But the BOP identifies only deterrence and punishment as its “legitimate penological interests,” and explains only how the Zimmer Amendment relates to these two interests, never once even hinting — either in its brief or at oral argument — that the Zimmer Amendment might also reasonably relate to a governmental interest in conserving correctional funds. The only place the concept of money appears in the BOP’s discussion of the Zim-mer Amendment’s goals is in a quotation from the district court — “The District Court found from this legislative history that Congress enacted the Zimmer Amendment to ‘curb spending on prison amenities and to enhance the prison as a deterrent’ ” — to which the BOP attaches the following conclusion: “Therefore, Congress’ punitive intent is clear from the legislative history.” Appellee’s Br. at 12-13. The BOP’s focus on punishment and deterrence is hardly surprising, since, aside from a cursory reference to eliminating this “waste of taxpayer dollars,” the Zimmer Amendment’s legislative history makes quite clear that punishment and deterrence were in fact the amendment’s primary objectives. 147 Cong. Rec. H7751, H7768 (July 1995) (statement of Rep. Zimmer) (emphasizing that “inmate amenities are better than what lawabiding Americans have” and that such “prison perks undermine the concept of jails as deterrence”).
Although we are, of course, free to “consider any argument on appeal that supports the judgment of the District Court,” Dimond v. District of Columbia, 792 F.2d 179, 187 (D.C.Cir.1986), we may not decide a case on a claim that the parties have neither briefed nor argued nor even mentioned. Doing so is not only unfair to the parties, who have had no opportunity to respond to the court’s rationale or to object to the court’s willingness to decide the case without an appropriate evidentiary record, Corson & Gruman Co. v. NLRB, 899 F.2d 47, 50 n. 4 (D.C.Cir.1990) (“We require petitioners and appellants to raise all of their arguments in the opening brief to prevent ‘sandbagging’ of appellees and respondents and to provide opposing counsel the chance to respond.”), but it undermines “[t]he premise of our adversarial system[,] ... that appellate courts do not sit as self-directed boards of legal inquiry and research, but essentially as arbiters of legal questions presented and argued by the parties before them.” Carducci v. Regan, 714 F.2d 171, 177 (D.C.Cir.1983).
IV.
Whether Congress may ban electric guitars from federal prisons might seem an unimportant — even trivial — question, but courts are obligated to adjudicate even seemingly unimportant issues in accordance with the Constitution and prevailing case law. So long as Safley remains the law of the land, courts are obligated to ensure that any restriction on prisoners’ constitutional rights — whether in the form of an outright ban or a denial of funding that functions as an outright ban — is reasonably related to the government’s legiti*242mate penological interests. I would thus reject the BOP’s subsidy rationale and remand to the district court, where the BOP may, if it chooses, attempt to defend the Zimmer Amendment as a reasonable response to its financial concerns.