dissenting:
From the beginning, the central issue presented by Steffan’s case has always been whether the military may constitutionally exclude from membership in the services individuals who admit to homosexual orientation, without any evidence of homosexual conduct or intent to engage in such conduct. Today’s majority reformulates Steffan’s appeal to avoid this critical issue. Through ingenious but totally unjustified uses of presumptions and inferences, the court seeks to transform Steffan’s case into one concerning homosexual conduct — when in fact the Navy has never even alleged that Steffan engaged or intended to engage in such conduct. See Majority opinion (“Maj. op.”) at 695.
The linchpin of the court’s transformation strategy is its assertion that “a statement that one is a homosexual” may be “used by the Navy as a proxy for homosexual conduct — past, present, or future.” Maj. op. at 688. We disagree in the most fundamental way with that claim, and believe that in the military context, where homosexual conduct results in automatic discharge or imposition of criminal sanctions, it is inherently unreasonable to equate an admission of homosexual identity with commission of or intent to engage in homosexual conduct. The in banc court’s attempt to recast the case to avoid the issue that the parties, the trial court, and the original panel opinion have identified as at the core of this litigation must ultimately fail. The critical issue posed in the starkest fashion by Steffan’s case is whether a member of the armed forces may be discharged on the sole basis of an admission of his homosexual orientation. We believe that he may not and we therefore dissent.
I. BACKGROUND
This case has a long and complicated history. The majority’s summary of that history is incomplete and in important respects misleading. A fair appraisal of the competing constitutional positions demands a more thorough exposition of the underlying events and earlier stages of this litigation.
In 1983, Joseph Steffan enrolled in the United States Naval Academy. During his four years there, his superiors variously praised him as “gifted,” “professional,” an “outstanding performer” who “exhibited excellent leadership,” and an “asset to the Academy.” None questioned that he would “undoubtedly make an outstanding naval officer.”
In February 1987, less than three months before Steffan was scheduled to graduate from the Academy, the Naval Intelligence Service (“NIS”) received a report alleging that Steffan had told two fellow students that he was homosexual. The NIS began an inquiry that continued until June 16, 1987. The final report from the NIS investigation found no evidence of homosexual conduct on the part of Steffan.1
In mid-March, Steffan learned of the NIS investigation from another midshipman. He *702asked the Chief of Chaplains at the Academy, Captain Byron Holderby, to intercede on his behalf with the Commandant of Midshipmen,2 Captain H.W. Habermeyer, in the hope that the Chaplain’s aid could “assure his graduation.” Captain Holderby’s efforts proved unavailing; accordingly, Captain Ha-bermeyer “advised [him] to advise Midshipman Steffan to seek legal counsel regarding this matter.”
Several days later, Steffan himself approached Captain Habermeyer with a special request to see the Superintendent. Steffan explained that he wished to tell the Superintendent personally of his desire to graduate with his class in June. Captain Habermeyer asked Steffan, “Are you willing to state at this time that you are a homosexual?” Stef-fan responded, “Yes, sir.” Captain Haber-meyer informed Steffan that he would not recommend approval of his request, and that he “seriously doubted whether [the Superintendent] would permit [Steffan’s] completion of his course of study to receive his diploma.”
The following day, March 24, Captain Ha-bermeyer convened a Brigade Military Performance Board to review Steffan’s situation.3 The hearing began with introductory remarks by Captain Konetzi, the presiding officer, including a description of Steffan’s performance at the Academy as “outstanding.” He then asked Steffan, “I’d like your word, are you a homosexual?” Steffan responded “Yes, sir.” Captain Konetzi continued, “Do you have anything else to add at this point?” Steffan replied, “No, sir.” Captain Konetzi thereupon concluded the hearing. The Board voted to forward Steffan’s ease to the Commandant with the recommendation that Steffan be separated from the Academy.
On March 26, the Commandant referred Steffan’s case to the Academic Board, "with the recommendation that “Midshipman Stef-fan be separated from the Naval Academy due to insufficient aptitude for commissioned service.” On April 1, the Board convened. It urged Steffan to “accept a qualified resignation in lieu of discharge.” Steffan gave a brief prepared statement in which he asked that he be allowed to graduate. The Board then voted unanimously to recommend discharge to the Superintendent.
That same day, the Superintendent advised Steffan that he intended to recommend Steffan’s discharge to the Secretary of the Navy. The Superintendent, however, exercised his authority to permit Steffan the option of submitting a qualified resignation; he stated that if Steffan did so, he would forego submitting a recommendation of discharge.
Also on April 1, Steffan met with the Academy Performance Officer, Major Funk, whose duty it was to counsel midshipmen faced with the choice between qualified resignation and discharge. Major Funk stated that discharge would require noting Steffan’s admission of homosexuality on his record. Major Funk expressed his opinion that Stef-fan’s job prospects would suffer from such a notation. He repeated the Academic Board’s urging that Steffan resign rather than wait to be discharged.
*703Still on April 1, approximately six weeks before graduation and after nearly four years of outstanding performance, Steffan agreed to submit his qualified resignation. He signed a statement of understanding reiterating that he would be discharged if he refused to resign, and warning that by submitting a qualified resignation, Steffan would “forfeit his right to show cause to higher authority why he should not be disenrolled from the Naval Academy.” On May 28, 1987, the Secretary of the Navy accepted Steffan’s resignation.
On December 9,1988, Steffan wrote to the Secretary of the Navy, requesting permission to withdraw his resignation and graduate from the Academy. The Secretary referred the matter to the Superintendent of the Academy, who “strongly” recommended denying Steffan’s request. He stated that because “Mr. Steffan admitted to being a homosexual” he “had insufficient aptitude to be a commissioned officer.” In accordance with this recommendation, the Secretary disapproved Steffan’s request.
Steffan filed this action on December 29, 1988. He requested a ruling that “the regulations pursuant to which the Naval Academy acted are unconstitutional ...” on the ground, inter alia, that the government had denied him equal protection under the Due Process Clause of the Fifth Amendment by forcing him to resign after a determination of “‘insufficient aptitude’ predicated solely on [his] sexual orientation.” The complaint did not specify which regulations were at issue. Indeed, several colloquies between this court and counsel for both Steffan and the Secretary during oral argument in banc demonstrate that neither Steffan nor the government were certain precisely which regulations authorized the Navy’s “processing” of Steffan. The majority correctly notes that “at various times different bodies within the military hierarchy relied on either [the Naval Academy’s own regulations] or [a Directive promulgated by the Department of Defense], or both.” Maj. op. at 684.
Steffan’s legal attack precipitated a flurry of opinions in the district court and in this court. In the first round, Steffan v. Cheney, 733 F.Supp. 115 (D.D.C.1989), the Secretary argued for dismissal on two threshold procedural grounds. First, the government claimed that Steffan lacked standing due to the “voluntary” nature of his resignation. The district court disagreed, holding that the purported “voluntariness” was no bar to jurisdiction. Steffan’s factual allegations established injury-in-fact (separation from the Academy) fairly traceable to defendants’ action (adoption and enforcement of regulations barring self-declared homosexuals from the Academy) and redressable by the court (through reinstatement and a declaratory judgment), thus clearly conferring Article III standing. Id. at 118-19.
The government also argued that Steffan had failed to exhaust his administrative remedies. It claimed that if Steffan’s resignation were truly involuntary, the Board for Correction of Naval Records (“BCNR”) could afford him “complete relief by reinstating him back into the Academy.”4 Id. at 119. The court rejected this argument on two grounds. First, the BCNR was not empowered to reinstate Steffan, but only to forward the matter to the Secretary of the Navy; yet the Secretary of the Navy had already once considered and rejected Steffan’s request to withdraw his resignation. Appeal to the BCNR thus *704could not have afforded Steffan “meaningful relief.” In addition, even if Steffan were permitted to withdraw his resignation, he would thereupon be “inevitably discharged” because he had “admitted] that [he] fit within the Navy’s definition of homosexuality.” Id. at 120.
A second district court Opinion, Steffan v. Cheney, 733 F.Supp. 121 (D.D.C.1989), addressed the government’s Motion for Rule 37 Sanctions, based upon Steffan’s refusal, on the grounds of irrelevance and Fifth Amendment privilege, to respond to deposition questions concerning whether he had engaged in homosexual acts. The district court acknowledged that Steffan was separated from the Academy based on his admissions rather than on any evidence of misconduct, but found inquiry into possible homosexual conduct “highly relevant” to Steffan’s qualification for reinstatement. Id. at 124-27.
In a per curiam opinion, this court (JJ. Wald, Ginsburg and Randolph) reversed and remanded. Steffan v. Cheney, 920 F.2d 74 (D.C.Cir.1990). We held that because “judicial review of an administrative record is confined to ‘[t]he grounds ... upon which the record discloses that [the] action was based,’” id. at 76, citing SEC v. Chenery Corp., 318 U.S. 80, 87, 63 S.Ct. 454, 459, 87 L.Ed. 626 (1943), and the record in Steffan’s case reflected an “administrative determination that he [was] unfit for continued service because he stated that he is a homosexual,” the district court had erred in requiring Stef-fan to answer conduct-related questions. We also rejected the government’s argument that homosexual conduct essentially had to have been involved in the administrative determination because “Steffan’s admission of homosexuality raised a ‘rebuttable regulatory presumption that he had a predilietion [sic] to commit, and had committed, homosexual acts.’ ” Id. at 76 n. *. This claim, the court wrote, was “not raised in the district court [and] finds no support in the record.” Id.
On remand to the district court, the government and Steffan cross-filed for summary judgment. The district court, in Steffan v. Cheney, 780 F.Supp. 1, 4 (D.D.C.1991), rev’d, 8 F.3d 57 (D.C.Cir.1993), vacated and rehearing in banc granted, (D.C.Cir. Jan. 7, 1994), granted the government’s motion and upheld “the regulations in question,” but did not specify which regulations were “in question.” Today’s majority correctly notes that “Midshipmen enrolled in the Naval Academy are subject to at least two sets of regulations relevant to homosexuality: the Naval Academy’s own regulations [“Navy regulations”] and the Directives of the Department of Defense [“DOD Directive”] applicable to the armed forces.” Maj. op. at 682. And the district court wrote that “[t]he plaintiff is suing for ... a declaration that the Department of Defense Directives 1332.14 and 1332.30, and all other regulations applied to the plaintiff prohibiting those with a homosexual orientation from serving in the Navy or attending the Naval Academy, are violations of the equal protection component of the fifth amendment to the Constitution.” Steffan, 780 F.Supp. at 2. Therefore, by granting the Secretary’s motion for summary judgment, the district court appears to have ruled that both sets of regulations were constitutionally valid.
The district court was more precise, however, concerning the focus of the case. Alluding to our earlier holding that Steffan’s conduct was irrelevant to this case, Steffan, 920 F.2d at 76 (D.C.Cir.1990), the court stated at the outset that “this is primarily a case about the plaintiffs status as a homosexual.” Steffan, 780 F.Supp. at 5. Nevertheless, it held, the “regulations in question are not violative of [ ] equal protection” because (1) homosexuals are not a suspect class, so regulations discriminating on the basis of sexual orientation are subject only to rational basis review, id. at 5-10, and (2) the regulations bear a rational relation to the legitimate government goals of “maintenance of discipline, morale, good order, a respected system of rank and command, ... morality and respect for [ ] privacy interests,” id. at 16, and to the goal (not raised by the government) of preventing the spread of AIDS in the armed forces. Id. at 13-16. Essentially, the court reasoned that because Steffan’s statement of orientation indicated he might “one day” engage in misconduct, all of the justifications that the Navy could offer in a case involving *705conduct “have equal application to this non-conduct case.” Id. at 12.
Steffan appealed. His brief to the panel stated the principal issue to be “[w]hether the military regulations that exclude persons from the armed services solely on the basis of their status as homosexuals ... violate the constitutional guarantee of equal protection of the laws.” Appellant’s Brief at 1, Steffan v. Aspin, 8 F.3d 57 (D.C.Cir.1993). He included both the Navy and DOD regulations in his addendum of challenged regulations and discussed both in his brief.5 The government’s brief presented the question as “[wjhether the military’s ‘old policy’ on service by homosexuals ... violates equal protection.” Appellees’ Brief at 1. The government’s brief expressly discusses only the DOD Directive; it nowhere mentions the Navy regulations.
The panel unanimously reversed the district court’s judgment, and ordered Steffan reinstated in military service, graduated from the Academy, and commissioned in the Navy. Steffan v. Aspin, 8 F.3d 57, 70 (D.C.Cir.1993), vacated and rehearing in banc granted (D.C.Cir. Jan. 7, 1994). Following the lead of the district court and the parties, the panel did not consider the Navy regulations and DOD directive to pose separate constitutional questions, and specifically addressed itself only to the DOD Directive.
The panel emphasized that because the “district court [had] acknowledged that Stef-fan was discharged solely because of his status as a homosexual,” the critical issue in the case concerned the constitutionality of excluding from the services a “class [of individuals] ... defined by homosexual orientation, not conduct.” Id. at 63. The panel declined to consider whether homosexual orientation constitutes a suspect classification for purposes of equal protection analysis; instead it rejected the district court’s reasoning that the government may rationally infer from a serviceman’s admission of “homosexuality”— defined by the Directives to include mere “desires” — a “propensity” to engage in repeated homosexual conduct so as to justify separation.
The government determined to forego appeal of the panel’s decision on the merits. Instead, the government asked this Court to consider in banc whether that portion of the panel’s remedial order that required Steffan to be commissioned as an officer, see Steffan, 8 F.3d at 70, violated separation of powers principles. This court voted sua sponte to rehear Steffan’s entire case, including the merits, in banc. This extraordinary measure brought this case back before us for a fourth time.6
II. Analysis
A. The Importance of Steffan’s Concession
The majority declares that Steffan’s concession that “the military may discharge those who engage in homosexual conduct whether on or off duty”7 “frames the dispute.” Maj. op. at 684 (emphasis added). We agree, but Steffan’s concessions should be understood for what they are — concessions on issues not properly presented by this case, and that certainly do not warrant the majority’s intimation that he was obliged to make them.8
*706Like the district court and earlier panels of this court, Steffan believed that the question whether homosexual conduct may be proscribed in the armed services had no bearing on his appeal. Indeed, from the time of this court’s decision in Steffan v. Cheney, 920 F.2d 74 (D.C.Cir.1990), until today, everyone in this litigation reasonably believed that it concerned only the constitutionality of excluding from the services a “class [of individuals] ... defined by homosexual orientation, not conduct.” Steffan, 8 F.3d at 10. In that context, Steffan’s concession at argument that “homosexual conduct” (of all varieties and in all circumstances) may be proscribed in the services must be viewed simply as an attempt to avoid argument about issues irrelevant to the appeal, as he and all other participants had pursued it up to that point, and not an authoritative statement of constitutional law. It is only in light of the majority’s transformation of Steffan’s case to one involving homosexual conduct that Steffan’s concession assumes the critical importance that it has today.
B. The Issue Presented by this Case: Discharge for Orientation
The majority disposes of Steffan’s appeal by transforming his unadorned admission of “homosexuality” from a statement of homosexual orientation into a declaration of past or intended homosexual conduct, thus avoiding the difficult question this case actually presents: whether an individual may be constitutionally discharged from the military on the sole basis of an admission of homosexual orientation. This alchemy is not, of course, an original idea — the government has argued from the outset that an admission of homosexuality both raises a “presumption” of past homosexual conduct and indicates a “propensity” to engage in future homosexual conduct. The policy of excluding from the ser-viees all who profess homosexual orientation is thus — according to the government — just a way of regulating conduct.
Today’s majority, in contrast to the Ninth Circuit’s recent decision in Meinhold v. U.S. Dep’t of Defense, 34 F.3d 1469 (9th Cir. 1994),9 embraces the government’s position. Its only change is a linguistic one; the government’s “presumption” of past conduct now becomes a “rational inference” of past conduct. But whether the Navy predicts future homosexual conduct from an admission of “homosexuality,” or “infers” past homosexual conduct from the same statement, its action is still impermissible absent evidence corroborating such conduct or intent. Because the Navy may not rationally draw either inference, we conclude that Steffan was unconstitutionally discharged solely for his homosexual orientation.
C. Both Sets of Regulations Present the Same Issue
We begin our analysis by emphasizing that although Steffan was “processed” under two sets of regulations — the Department of Defense Directives and the Naval Academy’s own regulations — both present the same constitutional issue. Indeed, although the court’s new emphasis on the Navy regulations and the bifurcated structure of its opinion might suggest otherwise, the crux of its argument is the same as to both sets of regulations. It is a “Heads, we win, Tails, you lose” proposition for Steffan, but ultimately a frolic and detour in terms of constitutional analysis.
The relevant portions of the Department of Defense Directive in effect at the time of Steffan’s discharge read:
H. Homosexuality
1. Basis
a. Homosexuality is incompatible with military service. The presence in the mili*707tary environment of persons who engage in homosexual conduct or who, by their statements, demonstrate a propensity to engage in homosexual conduct, seriously impairs the accomplishment of the military mission. The presence of such members adversely affects the ability of the Military Services to maintain discipline, good order, and morale; to foster mutual trust among servicemembers; to ensure the integrity of the system of rank and command; to facilitate assignment and worldwide deployment of servicemembers who frequently must live and work under close conditions affording minimal privacy; to recruit and maintain members of the Military Services; to maintain the public acceptability of military service; and to prevent breaches of security.
b. As used in this section:
(1) Homosexual means a person, regardless of sex, who engages in, desires to engage in, or intends to engage in homosexual acts;
(3) A homosexual act means bodily contact, actively undertaken or passively permitted, between members of the same sex for the purpose of satisfying sexual desires.
c. The basis for separation may include preservice, prior service, or current service conduct or statements. A member shall be separated under this section if one or more of the following approved findings is made:
(2) The member has stated that he or she is a homosexual or bisexual unless there is a further finding that the member is not a homosexual or bisexual.
DOD Directives 1332.14 and 1332.30, 32 C.F.R. Part 41, App. A (1991).
The Navy regulations provide that midshipmen who “possess certain traits which are undesirable in commissioned officers” “may be processed for separation in accordance with this instruction.” United States Naval Academy Regulation, COMDTMIDN Instruction 1610.6Í Ch-2.15.1 (July 16, 1987). They go on to list a number of “problems ... sufficient in and of themselves to warrant separation from the Naval Academy.” Id. at Ch-2.15.3. Among these “problems” is:
e. Homosexuality. The basis for separation may include previous, prior service or current service conduct or statements. Homosexuality includes the member engaging in, attempting to engage in or soliciting another to engage in a homosexual act or acts. It also includes statements by the member that he or she is homosexual or bisexual, or the member marrying or attempting to marry a person known to be of the same biological sex.
Id. at Ch-2.15.3.c. The Navy regulations contain no further definition of “homosexual.”
The court devotes the bulk of its opinion to the Navy regulations. The majority offers two different theories by which to equate Steffan’s bare-boned statement of “homosexuality” with past or intended homosexual conduct. First, it says the Navy could properly have concluded from Steffan’s statement that he was “likely” to engage in homosexual conduct in the future. Maj. op. at 685-86.10 Second, the Navy could have taken his statement as an admission that Steffan had already engaged in past homosexual conduct. The nub of the majority’s argument appears to be that when Steffan admitted his “homosexuality,” he was actually “saying” any one of three things: (1) “I have already engaged in homosexual conduct”; (2) “I intend to engage in homosexual conduct”; or (3) “I desire to engage in homosexual conduct, but I do not engage or intend to engage in such conduct — I am simply homosexual by orientation.” 11 Given this trinity of possible *708meanings to Steffan’s declaration of homosexuality, the argument continues, the military was entitled to infer that Steffan meant the first or second, but not the third. See id.
The court’s disposition of Steffan’s challenge to the DOD Directive is basically the same. First, the court contends that Stef-fan’s “as-applied” challenge to the regulations fails because the Navy was entitled to infer “conduct” or “intent” from his admission of “homosexuality”; he therefore has not shown affirmatively that the “desires” prong of the regulation — the prong Steffan claims is unconstitutional “as applied” — was ever actually “applied” to him. Maj. op. at 693-96. Second, Steffan lacks “prudential standing” because if the Navy discharged him on the basis of “conduct” or “intent,” inferred from his statement, then he was not harmed by the “desires” (or “orientation”) portion of the regulation.12 See Maj. op. at 698-99. Therefore, although the court addresses the two sets of regulations separately, both analyses present the same issue: whether the Navy could constitutionally discharge Steffan based solely on an inference of homosexual conduct — past or future — from his admission of homosexual orientation, without corroborating evidence of conduct or intent.
D. Rationality Review
A government action that burdens individuals unequally but does not implicate a fundamental right or burden a “suspect” or “quasi-suspect” class is subject to “rational-basis” or “rationality” review. Rationality review requires, “at the minimum,” that legislative classifications must be “rationally related to legitimate governmental objectives.” See, e.g., Schweitzer v. Wilson, 450 U.S. 221, 230, 101 S.Ct. 1074, 1081, 67 L.Ed.2d 186 (1981); Plyler v. Doe, 457 U.S. 202, 216, 102 S.Ct. 2382, 2394, 72 L.Ed.2d 786 (1982) (Under rationality review, “[t]he Equal Protection Clause directs that ‘all persons similarly circumstanced [must] be treated alike.’ ”) (citation omitted). Rationality review also compels that the challenged legislation “find some footing in the realities of the subject addressed by the legislation.” Heller v. Doe, -U.S.-,-, 113 S.Ct. 2637, 2643, 125 L.Ed.2d 257 (1993).
Thus certain kinds of motivations or justifications for discriminatory government behavior have been found categorically “irrational.” For example, government actions depriving individuals of the equal protection of the laws on the sole basis of invidious prejudice or unreasoned antipathy can never be deemed “rational.” See City of Cleburne v. Cleburne Living Center, Inc., 473 U.S. 432, 446-48,105 S.Ct. 3249, 3257-59, 87 L.Ed.2d 313 (1985), citing United States Dep’t of Agriculture v. Moreno, 413 U.S. 528, 533-35, 93 S.Ct. 2821, 2826, 37 L.Ed.2d 782 (1973) (holding a “bare ... desire to harm a politically unpopular group” insufficient to justify a statutory scheme). Similarly, government actions reflecting adherence to naked stereotypes cannot be “rational.” See Stanton v. Stanton, 421 U.S. 7, 13-15, 95 S.Ct. 1373, 1377-78, 43 L.Ed.2d 688 (1975) (“role-typing” and “old notions” could not provide rational basis for a state statute specifying for males a greater age of majority than for females); Reed v. Reed, 404 U.S. 71, 75-76, 92 S.Ct. 251, 253-254, 30 L.Ed.2d 225 (1971) (mandatory statutory preference for men over women in the appointment of estate administrators failed rational-basis test).
The limited role that the courts continue to exercise under rationality review thus remains a significant bulwark against unreasonable and illegitimate classifications. When government action deprives individuals of the equal protection of the laws for arbitrary reasons, or for reasons founded solely upon irrational and invidious prejudices, a court must declare the action unconstitutional under rational-basis review. Cleburne, 473 U.S. at 450, 105 S.Ct. at 3259-60. Otherwise, *709rationality review would be tantamount to no review at all — a result manifestly inconsistent with Supreme Court precedent. See, e.g., Zobel v. Williams, 457 U.S. 55, 61-64, 102 S.Ct. 2309, 2313-15, 72 L.Ed.2d 672 (1982); Moreno, 413 U.S. at 533-38, 93 S.Ct. at 2825-27 (1973).
At the same time, however, rationality review imposes no burden on the government to make any particularized justification of its behavior. So long as the action taken is in fact reasonable, the government need not make findings or produce evidence to support its decision. Heller v. Doe, — U.S. at —, 113 S.Ct. at 2643. Before striking down a statute or regulation on rationality review, the court must find wanting any justification actually proffered by the government, as well as other potentially legitimate grounds for the action. Id. at —, 113 S.Ct. at 2642-43.
The presence of a military branch as defendant in equal protection litigation does not eviscerate the courts’ role. It is certainly true that we accord great deference to the judgment of military decisionmakers within their areas of expertise. For example, in Goldman v. Weinberger, 475 U.S. 503, 508, 106 S.Ct. 1310, 1313-14, 89 L.Ed.2d 478 (1986), the Supreme Court upheld this court’s ruling that a military prohibition on servieemembers wearing yarmulkes survived strict scrutiny, although the regulation would likely have been struck down as violative of the First Amendment in other contexts. In effect, the Court deferred to the military’s judgment that its interest in uniformity of dress — which may not have been even an “important” governmental interest in another context — rose to a “compelling” level in the military. Surely the military itself is most competent to determine whether uniformity of dress is merely “important” or positively “compelling” in the military context.
Steffan’s case presents quite a different picture, however. Steffan has conceded, “for today,” that keeping individuals who engage or intend to engage in homosexual conduct out of. the military is a “legitimate” interest. The issue here is thus limited to whether the government may “rationally” infer past or future conduct from Steffan’s admission of “homosexuality.” The military has no special competence to decide this question. To the contrary, reviewing the “rationality” of such inferences — with attendant legal consequences — drawn from status to conduct falls more properly within the expertise of the courts. Such analyses are an important part of our work. See, e.g., Aptheker v. Secretary of State, 378 U.S. 500, 509-11, 84 S.Ct. 1659, 1665-66, 12 L.Ed.2d 992 (1964) (government cannot punish status of Communist on the theory that subversive conduct would follow); Robinson v. California, 370 U.S. 660, 667, 82 S.Ct. 1417, 1420-21, 8 L.Ed.2d 758 (1962) (disallowing punishment of drug addicts, absent evidence of drug use). We therefore decline the government’s invitation to establish the military branches as ultimate arbiters of the “rationality” of their inferences under the Equal Protection Clause. With due respect for military expertise in appropriate areas, we employ the traditional rational basis tools in this situation.
We must, however, take issue with the majority’s assertion that we hold the mistaken view that “the government’s position is weakened if it does not produce evidence to support (‘demonstrate’) its regulatory proposition.” See Maj. op. at 685, 689-90. No such thing. Our position is that the inferences drawn in the regulations are not in fact rational ones — we do not rely on any argument that the government has failed to support them with evidence. In particular, as we explain in greater detail later, there are fundamental impediments deeply rooted in our constitutional jurisprudence that will not permit the government to treat a service-member’s statement of homosexual identity as a proxy for proscribed homosexual conduct. Governmental actions predicated on such a presumption must therefore fail rational-basis review, quite apart from eviden-tiary disputes. That the government has never tried to demonstrate — either in this case or in any other ease upon which the majority relies — that its inference of homosexual conduct following from an admission of orientation is rooted in “realit[y],” see Heller v. Doe, — U.S. at —, 113 S.Ct. at 2643, serves only to reinforce our view of its basic irrationality.
*710E. Inferring Future Homosexual Conduct From “Homosexuality"
Although the questions of whether the Navy may infer past or future homosexual conduct from an admission of homosexuality are closely related, the history of this case requires that we address them separately. Inferring future homosexual conduct from an admission of homosexuality presents the now-familiar “propensity” issue — whether the admission itself indicates that an individual will “one day” actually engage in proscribed homosexual conduct.
We address this question in three parts. Initially, because the majority’s analysis rests almost entirely on its untenable conflation of homosexual status and homosexual conduct, see Maj. op. at 689-90, we emphasize that this view has been rejected both by the military itself and by expert authority. We then demonstrate that, particularly in the military, there is no “rational connection” between orientation/status and conduct as a factual or experiential matter. Finally, we point out that even if a rational connection between these concepts could be shown, the Constitution prohibits presuming, on the sole basis of an admission of homosexual status, that a servieemember will “one day” violate military regulations governing sexual conduct.
1. Military and Expert Recognition of the Orientation/Conduct Distinction
The military itself recognizes a fundamental distinction between homosexual orientation and homosexual conduct. The DOD Directives under which Steffan was separated expressly distinguish between them:
Homosexual means a person, regardless of sex, who engages in, desires to engage in, or intends to engage in homosexual acts....
DOD Directive 1332.14(H)(l)(b)(l). Because the Directives’ inclusion of individuals with homosexual “desires” — as distinct from those who engage or intend to engage in homosexual conduct — would otherwise be wholly redundant, the Directives clearly embrace an orientation/conduct distinction.13 Indeed, the military has taken action against service-members on the basis of orientation alone. See, e.g., Selland v. Aspin, 832 F.Supp. 12,13 (D.D.C.1993) (quoting the General Counsel of the Defense Department as advising the Attorney General of two separations “based only on acknowledged homosexual ‘status,’ as opposed to homosexual conduct”).
Second, an orientation/conduct distinction is found in the Secretary’s interpretation of the very DOD Directives under which Stef-fan was separated. In his brief to the in banc court, the Secretary acknowledges that under these regulations status ought not be conflated with conduct. He counsels the court not to read the regulations to require discharge in every case when a servicemem-ber states that he is a homosexual. Instead, the Secretary asks us to read into the regulations a “rebuttable presumption” for celibate homosexuals.14 The Secretary obviously appreciates the irrationality- — even under the “old” regulations at issue in this case — of a rigid rule equating homosexual orientation and conduct.
Third, the military’s recognition of the distinction between homosexual orientation and conduct rises to a full-blown status in the Secretary’s newest policy on homosexuals in the military. On July 19, 1993, after Stef-fan’s separation, the Secretary issued a memorandum to the Defense Department command structure, stating:
[I]t is the policy of the Department of Defense to judge the suitability of persons to serve in the armed forces on the basis of their conduct. Homosexual conduct will be grounds for separation from the military *711services. Sexual orientation is considered a personal and private matter, and homosexual orientation is not a bar to service entry or continued service unless manifested by conduct.
Department of Defense, “Policy on Homosexual Conduct in the Armed Forces,” cited in Brief for Appellee, Addendum at 1, Stejfan v. Aspin, 8 F.3d 57 (D.C.Cir.1993). The most recent policy not only explicitly acknowledges the distinction between homosexual status and homosexual conduct, but, even more significantly, admits that homosexual orientation by itself is not incompatible with military service. See DOD 1332.14(H)(1)(a) (Dec. 21, 1993) (“[S]exual orientation is considered a personal and private matter, and homosexual orientation is not a bar to continued service ... unless manifested by homosexual con-duct_”). The new DOD Directives no longer include any reference to “desires”; they focus instead exclusively on homosexual conduct and the intent to engage in such conduct. Id.
Particularly in light of this change in the military’s stance toward homosexuality per se — which we must assume represents the result of the military’s best reasoning and experience15 — we can only be stunned by the court’s attempt to justify as rational a conflation of conduct and orientation that it attributes to the military, but that the military may never have endorsed and has now explicitly forsworn.16 This flaw infects all the stages of the majority’s “rationality” analysis.
Indeed, several recent court decisions have criticized this position of the majority and the cases on which it relies, see, e.g., Ben-Sha-lom v. Marsh, 881 F.2d 454 (7th Cir.1989), on the ground that they offer no convincing rationale. In the words of one jurist, Ben-Shalom “abandoned explicitly the distinction between orientation and conduct which has been stressed in other eases.... without any evidence in the ease before it and without citation to authority of any kind.” Jantz v. Muci, 759 F.Supp. 1543, 1547 n. 2 (D.Kan. 1991) (emphasis added), rev’d on other grounds, 976 F.2d 623 (10th Cir.1992), cert. denied, — U.S. —, 113 S.Ct. 2445, 124 L.Ed.2d 662 (1993).17 The contrary view, in contrast, has documented support. In Cammermeyer v. Aspin, 850 F.Supp. 910, 919 (W.D.Wash.1994), the district judge pointed to “substantial uncontroverted evidence” in support of his conclusion that “a distinction between homosexual orientation and homosexual conduct is well grounded in fact.” The judge reproduced much of this testimony in his opinion, including statements by clinical and research psychologists indicating that “a person’s public identification of his or her sexual orientation does not necessarily imply sexual conduct, past or present, or a future desire for sexual behavior.” Id.
Similarly, the court in Equality Foundation of Greater Cincinnati v. Cincinnati, 860 F.Supp. 417, 437 (S.D. Ohio 1994), concluded that “evidence amply established ... that there is a broad distinction between sexual orientation, and sexual conduct,” citing expert witness testimony that “[sjexual orientation ... ‘is a predisposition towards one’s own and/or the other gender’ and is not [] defined by any conduct.” Id. Indeed, evidence in Equality Foundation demonstrated that “sexual activity is not even necessarily a good predictor of one’s sexual orientation.” Id. Based on this testimony in the record, *712the court explicitly rejected the “fundamental underpinning of [Ben-Shalom, Woodward, and High Tech Gays ] — that homosexuality is a status defined by conduct.” Id.
2. The Lack of Any “Rational” Factual Connection Between Orientation and Conduct
Given, then, that homosexual orientation and conduct are analytically distinct concepts, the “propensity” question reduces to whether an admission of homosexuality alone, without elaboration of any kind, may rationally give rise to an inference that a particular individual will “one day” engage in homosexual conduct, regardless of the inhibitions of his or her environment. Neither the majority nor the government offers any indication that such a presumption is rooted in reality. The government’s brief baldly claims that there is a “sound factual connection between the proved and inferred facts,” Appellees’ Brief at 15, but offers no further support for this proposition. At oral argument, the government repeatedly relied on the incantation that the “nature of human sexuality” supported the inference. In Banc Transcript at 49, 55, 62, 68. The majority, for its part, simply declares: “[W]e are persuaded that in this case the correlation is more than sufficient to justify the government’s policy.” Maj. op. at 688.
We are not “persuaded.” The government’s contention in this case smacks of precisely the sort of stereotypical assessment forbidden by Stanton and Reed, see swpra at 688; at bottom, the government and the majority seem to be saying that gay service-members — unlike heterosexuals — must be presumed incapable of controlling their sexual “desires” in conformity with the law.18 While the government is not obliged to offer evidence to support the rationality of an inference, neither are courts obliged to accept the naked assertion of an untenable position.
The irrationality of the government’s inference is particularly patent in the military, where homosexual conduct is grounds for automatic discharge and, in the case of homosexual sodomy, punishable by incarceration. Indeed, it is much more reasonable to infer that a servicemember who admits to “homosexuality” will thereafter assiduously forego homosexual conduct. After all, servicemem-bers are surely aware that statements of homosexual orientation or desire will trigger close scrutiny of their subsequent behavior for evidence of homosexual “conduct” or “intent,” as indeed occurred in Steffan’s case.19 It would be foolhardy for servicemembers to freely admit “homosexuality,” unless they were quite confident that no additional evidence of conduct or intent existed. ’
The Ninth Circuit is in agreement with this view, and has interpreted the regulation at issue here as not reaching simple, unadorned admissions of homosexuality. In so doing, it cited numerous inconsistencies in the military’s position that raised, in the court’s view there, serious doubts about the “rationality” of the same inference urged upon us here. In Meinhold, 34 F.3d at 1478 n. 11, Judge Rymer noted that when an individual admits or has been found to engage in past homosexual acts, the military does not necessarily infer future homosexual conduct. Rather, so long as certain “approved findings” are made — including that the individual no longer “desires” to engage *713in homosexual acts — the DOD Directive permits such servicemembers to remain in the military. DOD Directives 1332.14 and 1332.30. Judge Rymer correctly concludes that it is not “wholly rational” to infer future homosexual conduct from a mere statement of “homosexuality” at the same time a similar inference is not necessarily made from past homosexual conduct by professed heterosexuals.
The Meinhold court also observed that the military’s “propensity” inference treated homosexuals and heterosexuals differently for no reason. Judge Rymer wrote:
Although courts defer to the military’s judgment about homosexual conduct, and classifications having to do with homosexuality may survive challenge if there is any rational basis for them [citations omitted], at least a serious question is raised whether it can ever be rational to presume that one class of persons (identified by their sexual preference alone) will violate regulations whereas another class (identified by their preference) will not.
Id. at 1478. It is telling that the majority does not seriously attempt to distinguish Meinhold, but rather limits its criticism of the holding to two points of secondary importance. See Maj. op. at 687 n. 7 (definition of the class of persons at issue) & 694 (exhaustion of remedies).
3. The Constitution Prohibits Inferring Proscribed Conduct
Finally, and most fundamentally, presuming that servicemembers who admit to homosexual orientation will inevitably violate military regulations conflicts with bedrock principles of our legal and constitutional order. For that reason, it is inherently irrational.
The Supreme Court has repeatedly emphasized that even prior conduct does not demonstrate a “propensity” to engage in the same actions after they later become illegal. See, e.g., Jacobson v. United States, — U.S. —,—, 112 S.Ct. 1535, 1542, 118 L.Ed.2d 174 (1992) (“Evidence of predisposition to do what once was lawful is not, by itself, sufficient to show predisposition to do what is now illegal, for there is a common understanding that most people obey the law even when they disapprove of it.”). And if prior conduct does not permit an inference regarding future conduct, such a conclusion is still less justifiable when based on mere orientation or “desire.” “[A] person’s inclinations and ‘fantasies ... are his own and beyond the reach of government....” Id. (quoting Paris Adult Theatre I v. Slaton, 413 U.S. 49, 67, 93 S.Ct. 2628, 2641, 37 L.Ed.2d 446 (1973)). Indeed, as Justice Marshall wrote in Stanley v. Georgia, 394 U.S. 557, 565, 89 S.Ct. 1243, 1248, 22 L.Ed.2d 542 (1969), “Our whole constitutional heritage rebels at the thought of giving government the power to control men’s minds.” Clearly the Navy may not exercise this power over the minds of servicemembers by discharging them on the basis of inquiries about “homosexuality.”
The “constitutional heritage” to which Justice Marshall referred in Stanley is evident in the evolution of the law of treason. Under a statute of Edward III, it was a crime to “compass or imagine the Death of ... the King.” Statute of Treasons 25 Edw. III. This became the crime of “constructive treason,” which was enforced against supposed “compassers” and “imaginers” even when no overt act (other than mere words) or agreement corroborated an intent to carry out the regicide. See, e.g. Case of Thomas Burdet, 79 Eng.Rep. 706 (1477); Trial of Sir John Perrot, 1 How.St.Tr. 1315, 1318 (1592); Trial of Thomas Hardy, 24 How.St.Tr. 199, 894 (1794) (all cited in Watts v. United States, 394 U.S. 705, 709-10 & n. 1 (1969) (Douglas, J. concurring)).
Our Constitution expressly repudiates constructive treason. Article III, section 3 declares: “treason against the United States Shall consist only in levying War against them, or in adhering to their Enemies, giving them Aid and Comfort. No person shall be convicted of Treason unless on the Testimony of two Witnesses to the same overt Act, or on Confession in open Court.” U.S. Const, art. Ill, § 3 (emphasis added). The restrictions imposed by our Constitution, limiting the definition of treason to particular conduct and requiring an “overt act” for conviction, express the fundamental constitutional principle that a person’s thoughts are his own— *714however distasteful they may be to the state or to the populace.
This same principle was accepted even in the cases that upheld the Smith Act’s broad proscriptions of subversive activities. Despite the nation’s widespread fears of Communist threats to overthrow the state, the Supreme Court never allowed prosecutions merely for private Communist sympathies. For example, in Scales v. United States, 367 U.S. 203, 221-24, 81 S.Ct. 1469, 1482-84, 6 L.Ed.2d 782 (1961), the Supreme Court construed the Smith Act’s “membership clause” to allow conviction only upon proof of both “active membership” in a communist-affiliated organization and a “specific intent” to overthrow the Government of the United States. Only if intent accompanied membership, the Court held, would the statute be brought “within established, and therefore presumably constitutional standards of criminal imputability.” Scales, 367 U.S. at 228, 81 S.Ct. at 1485. Any broader construction would render the statute constitutionally doubtful. Id. at 222, 224, 81 S.Ct. at 1482, 1483-84.
Similarly, in Dennis v. United States, 341 U.S. 494, 499-502, 71 S.Ct. 857, 862-63, 95 L.Ed. 1137 (1951), the Supreme Court construed the subversive advocacy provisions of the Act to require both specific conspiratorial intent to overthrow the government at some future date, and active advocacy of (as opposed to a mere statement of support for) that position. Later, as the nation’s fear of domestic communism waned, the Supreme Court in Aptheker v. Secretary of State, 378 U.S. 500, 84 S.Ct. 1659, 12 L.Ed.2d 992 (1964), reasserted its view that mere “thoughts and desires” — as evidenced by membership in a “subversive organization”— could never be sufficient grounds for deprivation of civil rights. In Aptheker, the Court held unconstitutional a law that indiscriminately deprived Communist Party members of their passports. Id. at 509-11, 84 S.Ct. at 1665-67. Cf. Robinson v. California, 370 U.S. 660, 665-67, 82 S.Ct. 1417, 1419-21, 8 L.Ed.2d 758 (1962) (unconstitutional to criminalize narcotics addiction in absence of proof of use); Powell v. Texas, 392 U.S. 514, 532-36, 88 S.Ct. 2145, 2154-56, 20 L.Ed.2d 1254 (1968) (upholding conviction for public intoxication because it was based on conduct, not status as chronic alcoholic).
Thus, even Cold War fears of internal subversion could not induce the Supreme Court to countenance the kind of presumption that the government argues and the majority adopts here — an inference of future misconduct on the basis of an admission of inchoate “desire,” unaccompanied by any specific intent to engage in misconduct. Such an inference is repugnant to time-honored legal principles that guard the sanctity of a person’s “thoughts and desires” against governmental control.
Indeed, numerous circuits have already applied this axiom to homosexual status in the context of gay and lesbian student groups denied recognition by state universities. The universities argued that because such organizations would encourage homosexuals to congregate and fraternize, they would facilitate the commission of criminal acts — homosexual sodomy or “deviate” sex acts. See Gay Students Org. of Univ. of New Hampshire v. Bonner, 509 F.2d 652, 662 (1st Cir. 1974); Gay Alliance of Students v. Matthews, 544 F.2d 162, 166 (4th Cir.1976); Gay Lib v. University of Missouri, 558 F.2d 848, 853 (8th Cir.1977). Therefore, the universities contended — much like the military here — they could deny university recognition to inhibit anticipated offensive behavior. The courts roundly rejected this line of argument. The Fourth Circuit, for example, wrote that while “the University could constitutionally regulate [ ] conduct,” its argument for denying official recognition to homosexual groups on campus could not be squared with the Supreme Court’s holding in Robinson prohibiting punishment for status. See Gay Alliance, 544 F.2d at 166 (citing Robinson v. California, 370 U.S. 660, 82 S.Ct. 1417, 8 L.Ed.2d 758 (1962)); see also Gay Lib, 558 F.2d at 856 (university cannot “ascribe singularly evil connotations to the group simply because they are homosexuals”); Gay Students, 509 F.2d at 662 (“speculation that individuals might at some time engage in *715illegal activity is insufficient to justify regulation by the state”).20
The majority’s attempt to analogize this ease to Massachusetts Bd. of Retirement v. Murgia, 427 U.S. 307, 96 S.Ct. 2562, 49 L.Ed.2d 520 (1976), upholding a mandatory retirement age of 50 for police officers on the ground that it rationally furthered a legitimate purpose of excluding officers lacking necessary physical fitness, misses the mark, as do its other examples of height and blindness disqualifications from government service. Of course the Navy can exclude individuals based on disqualifying physical or mental characteristics beyond their control. It is, however, an altogether different proposition to predicate exclusion on the assumption that certain individuals will not exert their will to prevent mere “desires” from translating into illegal actions.
To put the argument plainly, there is nothing a blind or tall person can do to negate those characteristics that disqualify him from military service. There is no will power that will spare an aging person the eventual loss of physical ability. And certainly there is no way that these people merely by conforming to the law can qualify for military service. But a servicemember who has homosexual desires can; he need only refrain from engaging in prohibited homosexual conduct, and by the Navy’s own admission he will be as “fit” as the next person.21 Since a decision not to act is within the control of the individual servicemember — unlike the “decision” whether to age or be blind — it is not rational to assume that he will choose to engage in conduct that would subject him to discharge or even incarceration.22
The majority’s citation to New York Transit Authority v. Beazer, 440 U.S. 568, 99 S.Ct. 1355, 59 L.Ed.2d 587 (1979), is no more convincing. In Beazer, the Supreme Court upheld a regulation disqualifying methadone users who had formerly been heroin addicts from employment with the NYTA. But two crucial elements of Beazer are absent from Steffan’s case. First, the district court there had found that methadone use itself can cause “drowsiness, insomnia, excess sweating, constipation, and perhaps some other symptoms.” 440 U.S. at 588 n. 32, 99 S.Ct. at 1367 n. 32. The exclusion policy for methadone was thus based, at least in part, on a lack of current capacity to perform at an appropriate level, not solely on a prediction of future misconduct. In addition, even had Beazer involved such a prediction, evidence of past misconduct (heroin use) might have lent some support to the inference. A ser-vieemember’s mere admission of “homosexuality,” on the other hand, carries with it no inference of past conduct; indeed, such statements will likely be made only when no evidence of past or intended misconduct exists.
*716Finally, the majority suggests that the constitutional prohibition against inferring illegal conduct from mere thoughts or desires does not apply in a case not involving “criminal punishment,” but only an “employment deeision[ ].” Maj. op. at 682. The majority is mistaken. It is just as irrational and fundamentally unfair for the government to draw such an inference in the “employment” context as in any other. Indeed, the Supreme Court’s decisions indicate that the government may neither punish nor make “employment decisions” solely on the basis of citizens’ political affiliations or membership in “subversive organizations.” See Elrod v. Burns, 427 U.S. 347, 372-73, 96 S.Ct. 2673, 2689, 49 L.Ed.2d 547 (1976) (prohibiting dismissal of non-policymaking county employees based on political affiliation); Elfbrandt v. Russel, 384 U.S. 11, 17-19, 86 S.Ct. 1238, 1241-12, 16 L.Ed.2d 321 (1966) (invalidating law denying employment on grounds of membership in subversive organization). Cf. Meinhold, 34 F.3d at 1478 (pointing out “the constitutionally significant danger of making status a surrogate for prohibited conduct”).
Stripped down, the majority’s distinction is constitutionally unpalatable. Any government action with serious civil or economic consequences for an individual admitting homosexual (as opposed to heterosexual) “desire,” predicated on the assumption that only the homosexual individual will violate government regulations prohibiting certain behavior, is equally defective.
F. Inferring Past Conduct From an Admission of “Homosexuality”
The majority advances the alternative position that the Navy could — on the basis of Steffan’s admission of “homosexuality” — discharge him for past conduct, quite apart from any “propensity” to engage in future conduct. In our view, any assumption of past conduct from such an admission is subject to all the same objections made against inferring “propensity” to future conduct. And more.
A rational connection between an admission of “homosexuality” and homosexual conduct is equally lacking whether the conduct inferred is past or future; and our Constitution prohibits the government from drawing that inference in either case. The Supreme Court’s ruling in Robinson v. California— holding unconstitutional a state’s attempt to criminalize narcotics addiction in absence of proof of use — did not hinge on whether California anticipated future use or suspected past use.
The majority’s assertion that the Navy may infer that Steffan “admitted” past conduct when he acknowledged his “homosexuality” is just another variation on the same old theme — twice rejected by this court at earlier stages of this case — that the Navy can “presume” past conduct from his statement, which he is then required to explain away. The majority would place the “burden” on Steffan to have explained what he meant by “homosexual” when he first answered “yes” to the government’s inquiry as to whether he was one, and conclude that because he failed to carry that burden, he “really” meant that he had engaged in homosexual conduct. This theory is unacceptable constitutionally, rationally, and indeed borders on sophistry.
It is, moreover, inconsistent with the regulations under which Steffan was “processed.” The court seems unable to grasp that there was, under any reasonable reading of the regulations, no “burden” that Steffan could possibly have carried to forestall discharge— and therefore no “presumption” or “inference” of past conduct that could have arisen from his silence. Indeed, nothing Steffan could have said before the Performance Board would have been remotely relevant to his predicament.23 The DOD directive explains, in the plainest prose, that “[a] member shall be separated ... if [a] finding is made [that] [t]he member has stated that he or she is a homosexual ... unless there is a further finding that the member is not a homosexual_” DOD Directive 1332.14.-H.l.e.2. The term “homosexual” is defined *717as “a person, regardless of sex, who engages in, desires to engage in, or intends to engage in homosexual acts.” Id. at 1332.14.H.l.b.l. Steffan does not now nor has he ever denied that he “stated that he ... is a homosexual,” under 1332.14.H.1.C.2. And he has from the outset straightforwardly admitted that he is a person aptly described by the definition of “homosexual” contained in 1332.14.H.l.b.l. Therefore, even had Steffan given all of the explanations and earned all of the “burdens” that the majority can conjure up, he would have gained no advantage whatever. He would still have been a “homosexual” under the regulations and he would still have been turned out of the Navy. Steffan did not “join[ ] the issue,” Maj. op. at 695, simply because there was no “issue” to join. As the Navy itself argued to the district court, “Officers who admit that they fit within the Navy’s regulatory definition of homosexual are inevitably discharged.” Steffan, 733 F.Supp. at 120.
The majority does not offer even a color-able explanation why this catch-22 predicament was not controlling in Steffan’s case. It argues only that if an individual who admits to being a “homosexual” goes on to explain in detail what he means by “homosexual” in his own case then the military might find that he is not a “homosexual” after all. See Maj. op. at 695. It never points to any authority under the old regulation that might have given rise to such a “hope.” Far from being “rational,” the argument is decidedly circular.
The majority’s error is thus one of 20/20 hindsight. As a court deciding a thorny appeal, we might well wish that Steffan had himself articulated the constitutional issue at his original hearing. But Steffan was not a court, nor even an attorney planning a constitutional appeal. He was a midshipman, a man who had dedicated four years to a career in the Navy, and he wished to remain in the Navy. Yet he knew that having once admitted his homosexuality, nothing he could say would keep him in service, unless it led ultimately to the determination that he was not a homosexual. By the plain language of the DOD Directive, no amount of emphasis on the word “desire” would do that. He therefore, quite understandably, said nothing further. The government cannot later— whether two years later, in district court, or seven years later in today’s majority opinion — inform Steffan that because he remained silent when further explanations would have been of no avail, his admission of homosexuality became, presumptively, an admission of conduct. It would be the worst kind of due process deprivation to retroactively apply a “presumption” or “inference” not even plausibly suggested by the regulations at the time.24
This court has already recognized this fact by rejecting the government’s earlier arguments that Steffan’s discharge involved homosexual conduct. In Steffan v. Cheney, 920 F.2d 74 (D.C.Cir.1990), we reversed the district court’s dismissal of the case as a sanction for Steffan’s refusal to answer questions about whether he had engaged in homosexual conduct. The district court had found that “[t]he record is clear that [Steffan] was separated from the Naval Academy based on his admissions that he is a homosexual rather than on any evidence of misconduct,” but nonetheless held that Steffan’s conduct was relevant to his eligibility for reinstatement. Id. at 76. We disagreed, stating that because the courts may review administrative action only on the bases contained in the administrative record, Steffan could not be held to answer conduct-related questions “unless [ ] conduct was a basis for his separation.” Id. at 76. With both the administrative record and the district court’s finding before us, we determined that conduct had not been the reason for Steffan’s discharge and therefore reversed the district court. Id. We also specifically rejected the government’s argument that because of a “rebutta-ble presumption” or “celibate homosexual ex*718ception”25 in the regulations, Steffan was presumptively caught by the conduct or intent prongs of the regulation until he demonstrated otherwise. The majority’s ruling today, that Steffan was discharged on the basis of homosexual conduct, thus effectively reverses the panel decision of four years ago— in the crucial part of the litigation that framed the issue now before us on appeal— that Steffan was not discharged for homosexual conduct, but for status.
Law of the case principles normally require a court to avoid thus disavowing an earlier “legal decision ... unchallenged in a subsequent appeal when the opportunity to do so existed.” Williamsburg Wax Museum, Inc. v. Historic Figures, Inc., 810 F.2d 243, 250 (D.C.Cir.1987). We recognize, of course, that an in banc court is not always “bound by the law of the case established by a panel on an earlier appeal, when that ruling was not reviewed in banc.” See, e.g., Van Ge-mert v. Boeing Co., 590 F.2d 433, 437-38 n. 9 (2d Cir.1978). Nevertheless, law of the case doctrine binds this court even in in banc review when “a party would [otherwise] be seriously prejudiced.... ” Id.; see also First Nat. Bank of Hollywood v. American Foam Rubber Corp., 530 F.2d 450, 453 n. 3 (1976) (law of the case doctrine need not apply if “no prejudice results from its omission”). We find it difficult to conceive how prejudice could be greater than that which Steffan suffers from today’s reversal of this court’s first Steffan ruling. It was, after all, as a direct result of that decision that the question whether Steffan had actually engaged or intended to engage in conduct was not joined in the district court. The combination of the first panel opinion and today’s majority opinion forecloses any opportunity for Steffan to challenge his discharge on any ground.
G. Effect on Morale, Discipline, and Recruitment
Because the majority concludes that Stef-fan was properly discharged based on an inference of past or future homosexual conduct drawn entirely from his admission of homosexuality, it does not address the legitimacy of other justifications — unrelated to conduct — offered by the government. Under rationality review the regulations pass constitutional muster if any rational basis can justify his discharge. See Heller v. Doe, — U.S. at —, 113 S.Ct. at 2642-43. Thus we consider the nonconduct rationales proffered.
The first justification offered by the DOD Directive itself is that the mere presence of homosexuals, even those who admit only homosexual orientation or “desire,” will have a negative impact on morale, discipline and the recruitment of new servieemembers. The Directives state, in part:
Homosexuality is incompatible with military service. The presence in the military environment of [homosexuals, later defined to include mere “desirers”] ... adversely affects the ability of the Military Services to maintain discipline, good order, and morale; ... [and] to recruit and retain members of the Military Services....
DOD Directive § H.l.a.
The argument appears to be that even if military policymakers did not make the irrational inference that servieemembers who admit homosexual orientation engage in homosexual conduct, the military could nonetheless exclude those servieemembers on the ground that heterosexual servieemembers would be appalled at the requirement that they serve alongside openly gay soldiers— even those who neither engage in homosexual conduct nor intend to do so. We are not told the source of the heterosexual service-members’ presumed distaste. Perhaps the assumption is that servieemembers would make the same irrational inference of conduct as military policymakers; perhaps it is that members dislike anyone with a homosexual orientation, no matter how he be*719haves. In either ease, the justification is without rational basis.
Under rationality review, we must assume the premise that heterosexuals would not wish to serve with individuals of homosexual orientation to be well-founded, even though we may personally hold a higher opinion of our servicemembers’ maturity than do their policymakers. We must accept too the unlikely assumption that forcing heterosexuals to serve with homosexuals will, in fact, lower morale, impair discipline, and discourage enlistment — despite our experience with the utter falsity of similar predictions voiced by opponents of President Truman’s 1948 executive order requiring racial integration of the armed forces. See Dahl v. Secretary of United States Navy, 830 F.Supp. 1319, 1330 (E.D.Cal.1993) (quoting Defense Personnel Security Research and Education Center, Nonconforming Sexual Orientations and Military Suitability at 8-10 (1988)). Still, given all that, it is a cardinal principle of equal protection law that the government cannot discriminate against any class of its citizens solely to give effect to the likes or dislikes of others. Such discrimination plays directly into the hands of bigots; it ratifies and encourages their prejudice.
In Palmore v. Sidoti, 466 U.S. 429, 104 S.Ct. 1879, 80 L.Ed.2d 421 (1984), the Supreme Court overturned a state court’s order that took a child away from her mother and delivered custody to her father solely because the mother had married an African-American. The state court reasoned that the child would be subject to discrimination in her community if she were to live with parents of mixed races. In a unanimous opinion, the Supreme Court held that even if the child’s welfare would be impaired, the court could not determine custody on that basis because the decision would give effect to the prejudice of others. This the Constitution would not allow:
Private biases may be outside the reach of the law, but the law cannot, directly or indirectly, give them effect. ‘Public officials sworn to uphold the Constitution may not avoid a constitutional duty by bowing to the hypothetical effects of private racial prejudice that they assume to be both widespread and deeply held.’
Id. at 433, 104 S.Ct. at 1882.
The Court extended this reasoning to rational-basis review in City of Cleburne v. Cleburne Living Center, Inc., 473 U.S. 432, 105 S.Ct. 3249, 87 L.Ed.2d 313 (1985). In Cleburne, the city had denied a zoning permit to construct a home for the mentally disabled in a residential neighborhood, arguing that neighborhood residents, angered by the presence of the home, would pose a danger to the mentally disabled residents. The Court (citing Palmare) rejected this argument as giving effect to irrational private biases, and held that it could not provide a rational basis for denying the permit. Cleburne, 473 U.S. at 448, 105 S.Ct. at 3258.
Palmore, Cleburne and the fundamental constitutional principle that they embody compel us to reject the government’s argument that individuals of homosexual orientation may be excluded from the military because others may be offended or angered by their mere presence. The Constitution does not allow government to subordinate a class of persons simply because others may not like them.
H. Privacy Rationale
The Directives also offer the justification that the mere presence of individuals of homosexual orientation in the military will invade the privacy of heterosexual servicemem-bers. DOD Directive § H.l.a. This argument could mean one of two things. The concern is either that homosexual service-members will ogle their heterosexual colleagues in close quarters, or that — regardless of whether such ogling takes place — heterosexuals will experience an invasion of personal privacy merely from their own fears.
Neither argument detains us long. The argument that homosexual “desirers” will stare reproduces the flawed presumption that mere thoughts and desires will translate into offensive conduct. The contention that heterosexuals must be protected from their fears of ogling, in turn, replicates the argument that governmental action should be controlled by the prejudices and stereotypes *720of third parties. Both purported justifications fail rational-basis review.
I. Rationales Not Offered by the Secretary
The Supreme Court’s recent decision in Heller v. Doe suggests rationality review must consider whether arguments not proffered by the government might provide a rational basis for the regulations. — U.S. at —, 113 S.Ct. at 2642-43. The district court relied on one such justification — preventing the spread of AIDS in the armed forces. The government on appeal has not endorsed this rationale for the military’s policy, and neither do we. Homosexual orientation cannot spread the AIDS virus. Homosexual, or heterosexual, conduct can.— if one participant carries the virus. This justification relies on the illegitimate assumption that homosexual servieemembers will break the rules and engage in prohibited homosexual conduct that may spread the disease, but heterosexuals — to whom sexual conduct is not forbidden — will not pose any such danger.
A final rationale on which the Secretary does not rely, but to which the Directives allude, is the “security” threat posed by homosexuals’ alleged susceptibility to blackmail. DOD Directive § H.l.a. As Judge Norris in the Ninth Circuit has aptly pointed out, however, the military’s policy increases the risk of blackmail by making gays and lesbians remain in the closet for fear of forfeiting their careers.26 Watkins v. United States Army, 875 F.2d 699, 730-31 (9th Cir. 1989) (Norris, J., concurring), cert. denied, 498 U.S. 957, 111 S.Ct. 384, 112 L.Ed.2d 395 (1990).
We conclude that the military’s discharge of Joseph Steffan on the sole basis of his admission of homosexuality was not rationally related to a legitimate government goal. We disagree with the majority that this case concerns only conduct, whose past or future existence may be inferred solely from the mere acknowledgement of “homosexuality.” To permit discharge on the basis of such an inference would contravene both principles of rationality review and constitutional guarantees. Additionally, we find no rational non-conduct justification for the regulations, insofar as they penalize a simple admission of homosexuality. We therefore believe that Steffan should be reinstated into military service, permitted to graduate from the Academy, and commissioned as a military officer.27
*721ConClusion
For the government to penalize a person for acknowledging his sexual orientation runs deeply against our constitutional grain. It has, we believe, no precedent or place in our national traditions, which spring from a profound respect for the freedom to think and to be what one chooses and to announce it to the world. The majority’s ingenious plays on presumptions and inferences cannot disguise the injustice that lies at the heart of this case. In years to come, we will look back with dismay at these unconstitutional attempts to enforce silence upon individuals of homosexual orientation, in the military and out. Pragmatism should not be allowed to trump principle or the soul of a nation will wither. We respectfully dissent.
. In light of this investigation, we find it curious that the majority would define the class of persons at issue in this case not as those "who say they are gay but have not acted in accordance with their propensity in the past,” as the Ninth Circuit recently did in Meinhold v. United States Dep't of Defense, 34 F.3d 1469 (9th Cir.1994), but rather as “persons who say they are gay but as to whom the military has no additional evidence as to their conduct.” Maj. op. at 687 n. 7. To suggest that the military had "no [] evidence” regarding Mr. Steffan’s conduct following a four-month investigation is surely incorrect — the plain truth is that the evidence they did find showed no homosexual conduct.
. The Commandant of Midshipmen is the officer second-in-command at the Academy, responsible only to the Superintendent. The Superintendent, in turn, answers to the Chief of Naval Operations, and is subject to the policies of the Secretary of the Navy.
. The Naval Academy’s Military Performance System provides for several levels of review before a midshipman may be separated from the Academy. Steffan’s case began at the Brigade Performance Board level. When, as in Steffan’s case, the Brigade Performance Board finds in favor of separation, the Board’s authority is limited to recommending the midshipman's separation to the Commandant of Midshipmen.
If the Commandant finds in favor of separation, he makes that recommendation to the Naval Academy Academic Review Board. That Board, if it also believes separation necessary, refers the case to the Superintendent of the Academy.
Upon receiving a recommendation of separation, the Superintendent must forward it to the Secretary of the Navy. The Superintendent may also, however, elect to give the midshipman the option of qualified resignation rather than outright discharge. If the midshipman receives and elects to exercise this option, by so doing he forfeits the right to show cause to the Secretary why the Superintendent's recommendation for separation should not be approved. The Secretary of the Navy makes the final decision whether to approve the separation — whether discharge or resignation. Once the Secretary has reached a decision, the Naval Academy regulations do not provide for further review.
. For example, Steffan's Statement of the Case specifically mentioned only the “Navy’s regulations,” while his Argument more frequently referred to "the servicewide regulations promulgated in 1981 as part of Department of Defense Directives 1332.14 and 1332.30_"
. We have discussed only three; on the remaining occasion, not relevant to this appeal, we upheld in an unpublished opinion the district judge’s refusal to recuse himself.
. Counsel for Mr. Steffan also conceded "for today" that the military could constitutionally discharge individuals who “haven't yet engaged in homosexual conduct but [] intend to.”
.The majority writes:
Steffan concedes that the military may constitutionally terminate service of all those who engage in homosexual conduct — wherever it occurs and at whatever time the conduct takes place. See Dronenburg v. Zech, 741 F.2d 1388, 1398 (D.C.Cir.1984) ...; Beller v. Middendorf, 632 F.2d 788, 812 (9th Cir.1980)....
Maj. op. at 685 & n. 4. Neither case, however, suggests that the military may constitutionally discharge members who engage in any "homosexual conduct” at any time, as the regulations under review in this case purport to do. See infra at 685. The issue in Dronenburg was whether the Navy was constitutionally entitled to discharge a “27-year-old petty officer [who] had *706[had] repeated sexual relations with a 19-year-old seaman recruit” “in a barracks on the Navy base.” 741 F.2d at 1389, 1398. Belter involved the discharges of two servicemembers who admitted repeatedly engaging in homosexual conduct while in service, and one who refused to terminate an active homosexual relationship with a fellow servicemember. 632 F.2d at 792-95.
. Meinhold, like this case, presented the question whether the services could constitutionally discharge a member for acknowledging his homosexuality. In an opinion by Judge Rymer, the court construed the DOD Directive so as not to require discharge for a mere admission of homosexuality, thus avoiding the "serious constitutional problem! ]” that would otherwise arise. Id. at 1476. The government has not sought in banc review in the Meinhold case.
. This, of course, is the Navy’s "propensity" argument, familiar from the earlier stages of this litigation. The court, however, now goes a step further, and suggests that the connection between homosexual orientation and conduct is far closer than mere “propensity”; there is actually no distinction between the two. See Maj. op. at 689-90.
. These three meanings of "homosexual” appear explicitly in the DOD Directive, which states in part: "Homosexual means a person, regardless of sex, who engages in, desires to engage in, *708or intends to engage in homosexual acts.” DOD Directives 1332.14 and 1332.30, 32 C.F,R. Part 41, Appendix A.
. This curious "prudential standing” argument does not even convince a majority of the in banc court. As the concurrences of Judges Buckley and Randolph suggest, the court’s purported "standing" argument is simply a restatement of the substantive argument that the Navy was entitled to discharge Mr. Steffan on the basis of "conduct” or "intent” inferred from his bare admission of homosexuality.
. It is a settled rule of statutory and regulatory construction that, where possible, each term should be given independent meaning. See Weinberger v. Hynson, Westcott and Dunning, Inc., 412 U.S. 609, 633, 93 S.Ct. 2469, 2485, 37 L.Ed.2d 207 (1973).
. Essentially, the Secretary claims that the DOD regulations give rise to a "rebuttable presumption” because even an individual who claims that he or she is a "homosexual” may remain in the service if there is a "further finding” that he or she is not a "homosexual.” See Appellees’ Brief at 688-90. We later explain our view that the language of the regulations as applied to Steffan simply will not bear the Secretary's construction. See infra at 696-98.
. The majority’s position is in considerable tension with its own assertion of "the special deference we owe the military’s judgment....” Maj. op. at 686. Here, the military has determined that conduct and orientation are distinct concepts; for purposes of its own analysis, the court collapses them nonetheless.
. The majority's lengthy citation from Judge Reinhardt’s dissenting opinion in Watkins v. Army advances its cause hardly at all. 847 F.2d 1329, 1361 n. 19 (9th Cir.1988) (suggesting that the class of individuals who have a "homosexual orientation” is somehow "define[dj” by "homosexual conduct”); accord Ben-Shalom v. Marsh, 881 F.2d 454, 464 (7th Cir.1989). Judge Reinhardt’s position was roundly rejected by the Watkins majority itself, see 847 F.2d at 1346-47, and is unsupported.
.Judge Kelly also noted that the Ninth and Federal Circuits have made similar statements. Both, "without citation to any evidence in the record or to a single medical authority” have announced that “[h]omosexuality ... is behavioral.” Jantz, 759 F.Supp. at 1547 n. 3 (1990), quoting High Tech Gays v. Defense Indus. Sec. Clearance Office, 895 F.2d 563, 573 (9th Cir. 1990); Woodward v. United States, 871 F.2d 1068, 1076 (Fed.Cir.1989).
. The majority's attempt to explain this irrational difference ■ in treatment on the ground that "heterosexuals have a permissible outlet for their particular sexual desires whereas homosexuals in the military do not," Maj. op. at 692, is undercut by the military's own definition of "homosexual," which includes individuals who admit to being “bisexual.” DOD Directive 1332.14(H)(l)(b)(2). Bisexuals obviously have a “permissible outlet for their sexual desires.”
Moreover, recent events belie the notion that because heterosexuals have a “permissible outlet for their particular sexual desires,” their "desires” are less likely than those of homosexuals to translate into forbidden conduct. Five West Point football players are currently facing charges of "groping” female cadets during a pep rally; seventy-six percent of the 1993 class of women cadets there report experiencing some form of harassment. Eleanor Randolph, Army Players are Accused of 'Groping,' Washington Post, Nov. 2, 1994, at Al & A16. Under Academy disciplinary rules, the most severe punishment the accused athletes face for their alleged misconduct is suspension for ninety days. Id.
. For three months after the initial admission, the NIS unsuccessfully sought evidence of homosexual conduct on the part of Mr. Steffan.
.The majority’s protest that the cited cases involve the First Amendment, Maj. op. at 692, misses our point. The fundamental due process notion that the Constitution prevents the government from inferring that individuals will one day engage in proscribed conduct on the sole basis of their homosexual orientation is unaffected by the level of scrutiny applied in a given case.
Moreover, as the Gay Alliance court's reliance on both First Amendment and equal protection rationales suggests, see 544 F.2d at 167, a line between the First Amendment and equal protection harms wreaked by the government’s discriminatory treatment of individuals admitting homosexual orientation may not so facilely be drawn. Although Mr. Steffan has chosen to frame his case in equal protection terms, he might well have challenged the government's attempt to control his speech on sexual identity on First Amendment grounds. See David Cole & William Eskridge, Jr., From Hand-Holding to Sodomy: First Amendment Protection of Homosexual (Expressive) Conduct, 29 Harv.C.R.-C.L.L.Rev. 319 (1994) (detailing the First Amendment problems presented by the military's homosexual ban).
. Throughout these proceedings, the Navy has strenuously urged that “celibate homosexuals” are not subject to discharge. This supposed exception alone casts doubt on the rationality of the inference that those who are homosexual by orientation will fail to exercise their self-control.
. The single piece of evidence in this record regarding what homosexual servicemembers actually do when faced with the prospect of harsh penalties for homosexual conduct can be found in Dr. Richard Green’s Declaration to the District Court. Dr. Green, a Professor of Psychiatry at UCLA, asserted that "substantial anecdotal evidence [suggests] that during World War II, gays in the militaiy simply refrained from engaging in homosexual conduct in order to avoid the harsh penalties that could be imposed” (J.A. 1137, citing Coming Out Under Fire: The History of Gay Men and Women in World War Two (Free Press, 1990)).
. At the Brigade Military Performance Board hearing convened to review Mr. Steffan’s situation, Captain Konetzi — the presiding officer— asked Mr. Steffan, "I’d like your word, are you a homosexual?” Mr. Steffan responded "Yes,- sir." Captain Konetzi then asked, "Do you have anything else to add at this point?,” and Mr. Steffan replied, “No, sir.”
. Cf. Satellite Broadcasting Co., Inc. v. FCC, 824 F.2d 1, 3 (D.C.Cir.1987) (Silberman, J.) (“Traditional concepts of due process incorporated into administrative law preclude an agency from penalizing a private party for violating a rule without first providing adequate notice of the substance of the rule.”). There is no indication in the record that Steffan ever received notice of any "rebuttable presumption”; hence, the district court's grant of summary judgment was inappropriate under the circumstances.
. These are two names for the same idea. As the government's attorney explained at oral argument in banc: "[W]hat was loosely and perhaps inartfully called the celibate homosexual exception is [] exactly the same point I am making today about rebutting the presumption."
In Steffan v. Cheney, the government used the label "rebuttable presumption.” In Steffan v. Aspin, 8 F.3d 57, 64 (D.C.Cir.1993), vacated and rehearing in banc granted (D.C.Cir. Jan. 7, 1994), they gave the argument another try as a "celibate homosexual exception," and we again rejected it.
. Indeed, although the government offers no evidence that the military’s policy decreases the risk of blackmail from external sources, amici cite research indicating that the policy increases the opportunities for sexual harassment within the services. A particularly notorious example involved a group of servicemen stationed at Kais-erslautern, Germany, who called themselves "dykebusters.” These men reputedly made systematic sexual advances to military women, and reported those who refused as lesbians. See Randy Shilts, Conduct Unbecomtng: Gays and Lesbians in the U.S. Military 496-97 (1993). See also, Michell M. Benecke & Kirstin S. Dodge, Military 'Women in Nontraditional Job Fields: Casualties of the Armed Forces' War on Homosexuals, 13 Harv. Women's LJ. 215 (1990).
. The Secretary’s request for rehearing in banc raised the single issue whether this court has the authority to order Steffan commissioned. Specifically, while the Secretary did not question our authority to order Steffan's reinstatement into the Navy and his graduation from the Academy, he claimed that requiring the Navy to commission Steffan somehow intrudes upon the authority granted to the President and the Senate under the appointments clause of the Constitution. U.S. Const, art. II, § 2 (“The President ... shall nominate, and by and with the Advice and Consent of the Senate, shall appoint ... Officers of the United States....’’).
We are told that ”[e]very physically qualified graduate [of the Naval Academy] is commissioned in the unrestricted line of the Navy or the Marine Corps.’’ Naval Academy Catalog for Prospective Midshipmen 51 (1982-83). Indeed, even before graduation, Mr. Steffan himself was as a matter of course nominated for a commission by President Reagan on May 11, 1987. The nomination was then referred to the Senate Armed Services Committee, which reported favorably on May 14, 1987. The Senate thereafter confirmed Steffan's nomination by unanimous consent. 133 Cong.Rec. 11,997, 12,033, 12,046, 12,-486, 12,636 (1987).
So far as the record reflects, then, the only impediment to Steffan's receiving his commission was his constructive discharge from the Naval Academy. See Appellees’ Letter of May 20, 1994 at 1 (any commission prepared for Steffan "would have been ineffective because [he] ... left the Naval Academy prior to completing the academic and military requirements necessary for commissioning”). The Secretary concedes that this court has the authority to remove that impediment by ordering Steffan reinstated *721and graduated. The specter raised by the Secretary, that the President might then refuse to perform the essentially ministerial duty of handing over Steffan’s commission, appears to be at this juncture a hypothetical case that is unnecessary to resolve.