Opinion for the Court filed by Circuit Judge SILBERMAN.
Concurring statement filed by Circuit Judge BUCKLEY, with whom RANDOLPH, Circuit Judge, joins.
Concurring statement filed by Circuit Judge RANDOLPH.
Statement concurring in part, dissenting in part, and concurring in the judgment filed by Circuit Judge GINSBURG.
Dissenting opinion filed by Circuit Judge WALD, with whom EDWARDS, Chief Judge, and ROGERS, Circuit Judge, join.
SILBERMAN, Circuit Judge:Joseph Steffan, a former Navy midshipman who admitted to being a homosexual, appeals from the judgment of the district court sustaining the constitutionality of the regulations pursuant to which he was discharged from the Naval Academy. We affirm.
I.
Midshipmen enrolled in the Naval Academy are subject to at least two sets of regulations relevant to homosexuality: the Naval Academy’s own regulations and the Directives of the Department of Defense applicable to the 'armed forces generally.
Academy regulations provide a number of “separation criteria” applicable to the “small minority of midshipmen” who “either [do] not perform to standards” or who “possess certain traits which are undesirable in commissioned officers.” United States Naval Academy Regulation, COMDTMIDN Instruction 1610.6F Ch-2.15.1 (July 16, 1987).1 A number of such deficiencies are considered to be “sufficient in and of themselves to warrant separation from the Naval Academy.” Id. at Ch-2.15.3. The Academy regulations provide a “listing” of those shortcomings, explaining that the “listing is not all-inclusive, but rather serves as examples which severely limit a midshipman’s aptitude and potential for commissioned service.” Id. With regard to “homosexuality,” one such concern, the regulations state:
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 (emphasis added). The Academy regulations do not further define the term “homosexual.”
The Department of Defense Directives applicable to homosexuality are more detailed. They begin with a statement describing their “basis” which provides:
Homosexuality is incompatible with military service. The presence in the military 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 and confidence among service members; to ensure the integrity of the system of rank and command; to facilitate assignment and worldwide deployment of service members who frequently must live and work under close conditions affording minimal privacy; *683to recruit and retain members of the Military Services; to maintain the public acceptability of military service; and to prevent breaches of security.
DOD Directive 13S2.14.H.l.a, 32 C.F.R. Pt. 41, App. A (1991) (superseded) (emphases added). The Directives mandate that a “member shall be separated ... if one or more of the following approved findings is made.” Id. at 1332.14.H.1.C. One such finding is 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.” Id. at 1332.14.H.l.c.(2). And the term “homosexual” is defined as “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).
Joseph Steffan enrolled in the Naval Academy in 1983 and successfully completed three of his four years of training, consistently being ranked near the top of his class. During the fall of his senior year, Steffan confided in two fellow midshipmen that he was a homosexual. One of the two reported Stef-fan’s conversation to Academy officials and on the basis of this report the Naval Investigative Service began an investigation of Stef-fan’s homosexuality. Steffan was informed of that inquiry by a fellow midshipman in March 1987. When questioned by Naval investigators, Steffan “invoked his right to remain silent,” but did confide his homosexuality to a chaplain in the Academy. Subsequently, in a meeting with the Commandant of the Academy, Steffan stated that he was a homosexual.
On March 24,1987, the Academy convened a meeting of its Performance Board. At that hearing, Steffan was asked, “I’d like your word, are you a homosexual?” He replied, “Yes, sir.” Steffan was then asked whether he had “anything else to add at this point,” and he answered “no.” Based on this hearing the Performance Board recommended to the Commandant of the Academy that “Stef-fan be separated from the Naval Academy due to insufficient aptitude for commissioned service.” The Board did not state explicitly whether it was relying on the Academy’s regulations or the Directives, although its conclusion appears to paraphrase the Academy regulation’s wording. The Commandant accepted this recommendation and forwarded it to the Academic Board, chaired by the Superintendent of the Academy. That Board met on April 1 and voted to recommend Steffan’s discharge from the Academy to the Secretary of the Navy, again based on “insufficient aptitude for commissioned service.”
Following that meeting, Steffan, who was advised by counsel, reached an agreement with the Navy, the terms of which were embodied in a “statement of understanding” signed by Steffan. Steffan acknowledged in the statement that based upon the recommendation of the Academic Board, the Superintendent of the Academy would recommend his discharge. Steffan had been given a choice: either submit a “qualified resignation” or litigate and risk recommendation of a discharge. The official transcript of a midshipman who submits a “qualified resignation” reads “Resigned” rather than “Discharged” as the cause of separation. But the qualified resignation itself includes an ac-knowledgement by the midshipman that he will be recommended for discharge by the Superintendent if he does not resign. Had Steffan chosen to appeal — presumably to the Secretary of the Navy — and had the Secretary decided that discharge was in order, Steffan’s transcript would have revealed “Discharged” as the reason for his termination. Steffan chose the first option and resigned from the Academy. The statement of understanding provided that by choosing to submit his resignation Steffan forfeited “his right to show cause to higher authority why he should not be disenrolled from the Naval Academy.” The Secretary of the Navy accepted Steffan’s resignation on May 28,1987. Subsequently, the Naval Investigative Service terminated its uncompleted investigation into possible conduct-related criminal and regulatory violations by Steffan.
Roughly a year and a half after submitting his resignation, Steffan wrote the Secretary of the Navy seeking to withdraw his resignation and resume his studies at the Academy. The Superintendent of the Academy “strongly” recommended to the Secretary that he deny the request. The Superintendent’s letter noted that Steffan had made an informed *684decision to resign following the conclusion of all the hearings to which he was entitled under Academy regulations. As for the merits of Steffan’s request, the Superintendent pointed out that Steffan’s admission that he was a homosexual constituted a basis for separation under the Academy regulations, and that the DOD Directives provided that “homosexuality” was incompatible with military service. The Secretary disapproved Steffan’s request to withdraw his resignation “in accordance with the recommendation of the Superintendent.”
Following that denial, Steffan brought suit in district court. Perhaps because of uncertainty as to whether his discharge was based on the Academy regulations or the DOD Directives, Steffan’s complaint sought a declaration generally that “the regulations pursuant to which the Naval Academy acted are unconstitutional on their face and as applied to the Plaintiff herein.” Steffan also sought an order “enjoining Defendants from prohibiting [him] from graduating and receiving his diploma from the Academy” and “from denying [him] his commission in the United States Navy.” The district court entered summary judgment in favor of the government, and Steffan appealed.2
Steffan, whose brief focuses almost entirely on the DOD Directives, argues that the military regulations lack a rational basis because they are simply an attempt to eater to the prejudices of members of the military and because they “punish” homosexuals simply on the basis of their “status” and “thoughts” rather than on the basis of conduct. Steffan concedes — and this concession frames the dispute — that the military may discharge those who engage in homosexual conduct whether on or off duty. The government contends that the regulations are a rational attempt to exclude from the military individuals who engage in, or demonstrate a propensity to engage in, homosexual conduct. The government also asserts that admission into the military of those who engage in such conduct would undermine unit cohesion. And the government defends the regulations as an attempt to protect the privacy of service members.
The record is, to say the least, confusing as to which regulation was applied to Steffan. At various times different bodies within the military hierarchy relied on either the Academy regulation or the Directives, or both. In light of this ambiguity, and because Steffan’s complaint clearly sought invalidation of any regulation on which the Naval Academy relied, we "will consider the constitutionality of both the Academy regulations and the DOD Directives, as if each alone had provided the basis for Steffan’s discharge.
II. Naval Academy Regulations
The familiar parameters of rational basis review were recently reiterated by the Supreme Court in Heller v. Doe, — U.S. —, 113 S.Ct. 2637,125 L.Ed.2d 257 (1993).3 “[R]ational-basis review in equal protection analysis ‘is not a license for courts to judge the wisdom, fairness, or logic of legislative choices.’” Id. at —, 113 S.Ct. at 2642 (citations omitted). The government, “moreover, has no obligation to produce evidence to sustain the rationality of a [regulatory] classification.” Id. Because “a classification neither involving fundamental rights nor proceeding along suspect lines is accorded a strong presumption of validity,” id., “ ‘[t]he burden is on the one attacking the [governmental] arrangement to negative every conceivable basis which might support it,’ whether or not the basis has a foundation in the record.” Id. at—, 113 S.Ct. at 2643 (citations omitted). This presumption of rational*685ity does not apply merely to congressional or state legislative schemes, but extends to administrative regulatory action as well, such as the military regulations at issue here. See Pacific States Box & Basket Co. v. White, 296 U.S. 176, 186, 56 S.Ct. 159, 163-64, 80 L.Ed. 138 (1935). The classification “is not subject to courtroom factfinding and may be based on rational speculation unsupported by evidence or empirical data.” FCC v. Beach Communications, Inc., — U.S. —, —, 113 S.Ct. 2096, 2098, 124 L.Ed.2d 211 (1993). The dissent is quite mistaken in asserting that under rational basis review the government’s position is weakened if it does not produce evidence to support (“demonstrate”) its regulatory proposition. See dissent at 709. It is hard to imagine a more deferential standard than rational basis, but when judging the rationality of a regulation in the military context, we owe even more special deference to the “considered professional judgment” of “appropriate military officials.” Goldman v. Weinberger, 475 U.S. 503, 509, 106 S.Ct. 1310, 1314, 89 L.Ed.2d 478 (1986).
Under this line of precedent we are required to ask two questions of the regulations. First, are they directed at the achievement of a legitimate governmental purpose? Second, do they rationally further that purpose? The first of these questions is not even in dispute in this case. As we have noted, 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.4 Counsel at oral argument further admitted, in connection with a discussion focused on the DOD Directives, that the military could ban even those who reveal an “intention” to engage in such conduct. It is common ground, then, that the regulations would be serving a legitimate purpose by excluding those who engage in homosexual conduct or who intend to do so.5
The dispute between the parties is thus limited to the question whether the regulations (focusing now on the Academy regulations), by requiring the discharge of those midshipmen who describe themselves as homosexual — whether or not the Academy has information establishing that an individual has engaged in homosexual conduct or intends to do so — are rational. Steffan first argues that there is no necessary factual connection between such self-description and such conduct. But Steffan relies primarily on a more subtle and novel argument. Even if the government could rationally, as a factual matter, draw a connection between the statement and the conduct, other legal considerations prevent the government from so doing. The military may not, according to Steffan, “punish” homosexuals solely on the basis of their “status.” Nor may the military presume that self-declared homosexuals will actually engage in homosexual conduct, for such conduct is illegal under the Code of Military Justice. (Sodomy is prohibited under 10 U.S.C. § 925 (1988).) Such a presumption — that someone will actually break the law — is inconsistent, he argues, with our legal traditions.
We consider first whether the Academy regulation has a rational factual basis. The appropriate question, it seems to us, is whether banning those who admit to being homosexual rationally furthers the end of banning those who are engaging in homosexual conduct or are likely to do so. The Academy can treat someone who intends to pursue homosexual conduct in the same manner as someone who engages in that conduct, because such an intent is a precursor to the *686proscribed conduct and makes subsequent homosexual conduct more likely than not. And the military may reasonably assume that when a member states that he is a homosexual, that member means that he either engages or is likely to engage in homosexual conduct. The inference seems particularly valid in this case because Steffan made no attempt to clarify what he meant by the term. He did not specify (nor was he asked by the Board) whether he had engaged in homosexual conduct in the past, whether he was presently engaged in homosexual conduct, whether he intended to engage in homosexual conduct in the future, or whether all three wei’e true. Indeed, as we noted, he had previously invoked his right to remain silent when questioned on these subjects. Nor did Steffan ever indicate that his answer to the Board referred to homosexual orientation as a concept implicating only wants or thoughts unrelated to conduct — a meaning that he now suggests was a possible interpretation of the term and which the dissent' embraces. He left it to the Board to draw what he apparently thought were the ordinary inferences the term homosexual suggests. These ordinary inferences are reflected in the Academy regulations and were the apparent bases for the Board’s conclusions.6 The dissent’s deconstruction of Stef-fan’s terse response overlooks the obvious point that Steffan assumed that the Board would fully understand what he meant.
Admittedly, it is conceivable that someone would describe himself as a homosexual based on his orientation or tendencies (and, perhaps, past conduct), notwithstanding the absence of any ongoing conduct or the probability of engaging in such conduct. That there may be exceptions to the assumption on which the regulation is premised is irrelevant, however, so long as the classification (the regulation) in the run of cases furthers its purpose, and we readily conclude
that it does. As then-judge Kennedy pointed out in Beller v. Middendorf, 632 F.2d 788 (9th Cir.1980), cert. denied, 454 U.S. 855, 102 S.Ct. 304, 70 L.Ed.2d 150 (1981):
Nearly any statute which classifies people may be irrational as applied in particular cases. Discharge of the particular plaintiffs before us would be rational, under minimal scrutiny, not because their particular cases present the dangers which justify Navy policy, but instead because the general policy of discharging all homosexuals is rational.
Id. at 808 n. 20 (citation omitted). The rule of law presupposes the creation of categories.
The military thus may rely on presumptions that avoid the administratively costly need to adduce proof of conduct or intent, so long as there is a rational basis for believing that the presumption furthers that end. And the military certainly furthers its policy of discharging those members who either engage in, or are likely to engage in, homosexual conduct when it discharges those who state that they are homosexual. The special deference we owe the military’s judgment necessarily affects the scope of the court’s inquiry into the rationality of the military’s policy. Compare dissent at 709. Whether a certain course of conduct is rational does not depend solely upon the degree of correlation that exists between a surface characteristic and a corresponding hidden trait. For the question whether the degree of correlation justifies the action taken — i.e., whether it is rational — necessarily depends on one’s assessment of the magnitude of the problem the action seeks to avoid. The military is entitled to deference with respect to its estimation of the effect of homosexual conduct on military discipline and therefore to the degree of correlation that is tolerable. Particularly in light of this deference, we think the class of self-de*687scribed homosexuals is sufficiently close to the class of those who engage or intend to engage in homosexual conduct for the military’s policy to survive rational basis review.
Because removing from the military all those who admit to being homosexual furthers the military’s coneededly legitimate purpose of excluding from service those who engage in homosexual conduct, Steffan’s argument at bottom must be based on the notion that the classification drawn by the military is impermissibly over-inclusive — that the military may not presume that all admitted homosexuals will engage in homosexual conduct because some homosexuals would not. However,
courts are compelled under rational-basis review to accept a legislature’s generalizations even when there is an imperfect fit between means and ends. A classification does not fail rational-basis review because it “is not made with mathematical nicety or because in practice it results in some inequality.” “The problems of government are practical ones and may justify, if they do not require, rough accommodations— illogical, it may be, and unscientific.”
Heller, — U.S. —, 113 S.Ct. at 2643 (iquoting Dandridge v. Williams, 397 U.S. 471, 485, 90 S.Ct. 1153, 1161, 25 L.Ed.2d 491 (1970) and Metropolis Theatre Co. v. Chicago, 228 U.S. 61, 69-70, 33 S.Ct. 441, 443, 57 L.Ed. 730 (1913)).7
Steffan seeks to end-run this analysis by arguing that a prohibition triggered simply by an admission of homosexuality is one based on “status” rather than conduct, and therefore is legally impermissible regardless of its rational relationship, as a factual matter, to the military’s objective. As the panel that initially decided this case put the point, “America’s hallmark has been to judge people by what they do, and not by who they are.” Steffan v. Aspin, 8 F.3d 57, 70 (D.C.Cir.1993), vacated and rehearing en banc granted (D.C.Cir. Jan. 7, 1994). In our view, however, Steffan’s attempt to invoke a rule against “punishment” based on “status” is unavailing, because it derives from a misunderstanding of constitutional law.
It is true that4 the Constitution forbids criminal punishments based on a person’s qualities — we assume that this is what is meant by “status” — rather than on his or her conduct. See Robinson v. California, 370 U.S. 660, 82 S.Ct. 1417, 8 L.Ed.2d 758 (1962). Yet, this proposition has never meant that employment decisions — which is what this ease is about — cannot be made on such a basis. One cannot be put in jail for having been born blind (although a blind person who drives a truck and kills someone could be jailed for his act). But it obviously would be constitutional for the military to prohibit blind people from serving in the armed forces, even though congenital blindness is certainly a sort of “status.” The logic of Steffan’s argument and of the original panel’s decision — that “America’s hallmark” prohibits “punishment” (which term is meant to encompass discharge decisions) based on a person’s “status” — would mean that the military acts unconstitutionally if it refuses to enlist blind individuals.
It is asserted that one does not choose to be homosexual and that therefore *688it is unfair for the military to make distinctions on that basis. But whether or not one’s homosexuality is genetically predetermined, one’s height certainly is. Steffan conceded at oral argument that the Navy’s maximum height restrictions are constitutional because they rationally further a legitimate naval purpose. That concession amounts to an admission that employment decisions based on a person’s characteristics are subject to the same analysis as decisions based on a person’s conduct. Both are tested to see whether they rationally further a legitimate purpose.
The controversy before us is quite analogous to Massachusetts Bd. of Retirement v. Murgia, 427 U.S. 307, 96 S.Ct. 2562, 49 L.Ed.2d 520 (1976) (per curiam), in which the Supreme Court upheld a mandatory retirement age of 50 for police officers on the grounds that the classification rationally furthered the government’s purpose of excluding those who lacked the physical conditioning to be officers. See id. at 314-17, 96 S.Ct. at 2567-69. In other words, the Court upheld a classification based on “status”- — -after all, a classification based on age turns on how old someone is, not on what he can do — that was aimed prophylactically at preventing the risk of unsatisfactory conduct. The connection between homosexuality and homosexual conduct is at least as strong (indeed, it seems much stronger) as the relationship upheld in Murgia between age — a paradigmatic “status” — and unsatisfactory job performance. The dissent would distinguish Murgia and our other examples on the grounds that a self-declared homosexual can control his sexual drives, whereas no one can hold back the process of aging. But people can and do manage to keep in superb physical condition even in advanced age. Murgia upheld the mandatory retirement age for police officers as reasonable not because every officer becomes infirm at 50, but rather because it was permissible to require retirement in every case despite the possible alternative of using medical examinations to indicate which officers over 50 were still fit to serve. The state’s method was over-inclusive, but it was not irrational. Id. at 314, 96 S.Ct. at 2567. Similarly, in this case, the possible existence of some self-identified homosexuals who do not and would not act on their desires in a military or civilian setting does not render irrational a regulation that reaches the class as a whole. Just as age can be used as a rough proxy for diminishing physical capacities, here we think that a statement that one is a homosexual can rationally be used by the Navy as a proxy for homosexual conduct— past, present, or future.
The government’s right to rely on a classification based on identifiable characteristics as a proxy for conduct was also sanctioned in New York City Transit Auth. v. Beazer, 440 U.S. 568, 99 S.Ct. 1355, 59 L.Ed.2d 587 (1979), in which the Court found that the Transit Authority’s policy of barring employment to drug addicts was rational as applied to methadone users enrolled in a treatment program. Given the “legitimate inference that ... a degree of uncertainty [regarding future drug use while still in the program] persists,” id. at 591, 99 S.Ct. at 1368, the government was entitled to assume that an addict might engage in unacceptable conduct while on the job. The dissent argues that the presumption in Beazer was based on indications of “past conduct,” since one would not be in a methadone program unless one had previously taken drugs. Dissent at 715. But surely the logic of the opinion indicates that the case would not have been decided differently if the Transit Authority had excluded as well those who stated that they were heroin addicts. Indeed, a statement can be a more reliable predictor of future (or present) behavior than evidence of past conduct. The unqualified statement “I am a homosexual” might well be more indicative of future homosexual conduct than a determination that someone had in the past engaged in such conduct, perhaps under unusual conditions or in brief experimentation. In either case, a correlation with future conduct is certainly to be expected; the only issue is the degree of correlation. As we have said, we are persuaded that in this case the correlation is more than sufficient to justify the government’s policy.8
*689To be sure, it would not pass even rational basis review for the military to reject service members because of characteristics— such as race or religion or the lack of inherited wealth — that have absolutely no bearing on their military service.9 Homosexuality, by contrast, is not irrelevant to homosexual conduct. And once Steffan concedes that the military may constitutionally seek to prevent the latter, his analogy to a hypothetical exclusion of those of a particular race or religion fails.
Nevertheless, Steffan, in order to make his point, would have us see homosexual status — which is all that he should be thought to have acknowledged — as conceptually unrelated to homosexual conduct. Although there may well be individuals who could, in some sense, be described as homosexuals based strictly on an inchoate orientation, certainly in the great majority of cases those terms are coterminous.10 Homosexuality, like all forms of sexual orientation, is tied closely to sexual conduct. As the Seventh Circuit reasoned:
It is true that actual lesbian conduct has not been admitted by plaintiff on any particular occasion, and the Army has offered no evidence of such conduct. Judge Gordon found no reason to believe that the lesbian admission meant that plaintiff was likely to commit homosexual acts. We see it differently. Plaintiffs lesbian acknowl-edgement, if not an admission of its practice, at least can rationally and reasonably be viewed as reliable evidence of a desire and propensity to engage in homosexual conduct. Such an assumption cannot be said to be without individual exceptions, but it is compelling evidence that plaintiff has in the past and is likely to again engage in such conduct. To this extent, therefore, the regulation does not classify plaintiff based merely upon her status as a lesbian, but upon reasonable inferences about her probable conduct in the past and in the future. The Army need not shut its eyes to the practical realities of this situation, nor be compelled to engage in the sleuthing of soldiers’ personal relationships for evidence of homosexual conduct in order to enforce its ban on homosexual acts, a ban not challenged here.... The Army need not try to fine tune a regulation to fit a particular lesbian’s subjective thoughts and propensities.
Ben-Shalom v. Marsh, 881 F.2d 454, 464 (7th Cir.1989), cert. denied, 494 U.S. 1004, 110 S.Ct. 1296, 108 L.Ed.2d 473 (1990) (emphasis added).
The dissent insists that homosexual self identification and homosexual conduct are not coterminous, or at least have not been proved “in this case or any other case” to be coterminous. Dissent at 709. As we have already noted, however, the dissent’s reasoning — that the government’s failure to produce evidence (“demonstrate”) that its inference is “rooted in reality” undermines its position — is predicated on an incorrect view of constitutional law, for which the dissent miscites Heller, see id., and relies on recent district court decisions challenging the military’s ban on homosexuals which simply represent an undisciplined rebellion against the governing constitutional doctrine. Id. at 711-12. It is not the government’s burden to establish the .exact degree of correlation between those who describe themselves as homosexual and those who engage in homosex*690ual conduct — let alone to show that the concepts are coterminous. Indeed, “the theory of rational basis review ... does not require the [government] to place any evidence in the record.” Heller, — U.S. at —, 113 S.Ct. at 2642 (emphasis added). Neither appellant nor the dissent, it must be emphasized, actually denies that there is a correlation (we are, after all, speaking of those who openly identify themselves as homosexuals, not of those who might simply experience what might be interpreted as a fleeting homosexual impulse). The dissent does, however, quote approvingly from a district judge who asserted, inter alia, that “there is almost no correlation between an individual’s sexual ‘orientation’ and his or her sexual conduct.” Dissent at 711-12. Even defining “orientation” in the broadest possible terms, we think this assertion is preposterous.
The government’s understanding of what is meant when an individual identifies himself or herself without qualification as a homosexual .is identical to the view of Judge Reinhardt, perhaps the federal judiciary’s most vocal proponent of constitutional protections for homosexuals in military or civilian life. See, e.g., Reinhardt, The Court and the Closet: Why Should Federal Judges Have to Hide Homosexuality?, Washington Post, Oct. 31, 1993, at C3. In his dissent in Watkins v. United States Army, 847 F.2d 1329, 1353 (9th Cir.1988), in which he sharply criticized Bowers v. Hardwick, 478 U.S. 186, 106 S.Ct. 2841, 92 L.Ed.2d 140 (1986), he said:
Even if we define the class as those who have a ‘homosexual orientation,’ its members will consist principally of active, practicing homosexuals....
... To pretend that homosexuality or heterosexuality is unrelated to sexual conduct borders on the absurd. What distinguishes the class of homosexuals from the class of heterosexuals is not some vague ‘range of emotions,’ but the nature of the member’s sexual proclivities or interests ....
... Whether the group is defined by status or by conduct, its composition is essentially the same.
Id. at 1360-61 & n. 19.11
We need not endorse Judge Reinhardt’s unequivocal position or his harsh criticism of the reasoning that is now relied upon by our dissenting colleagues. It is sufficient to recognize that the government’s presumption, as embodied in the Academy regulations, is certainly rational given that the human sexual drive is enormously powerful and that an open declaration that one is a homosexual is a rather reliable indication as to the direction of one’s drive.
The dissent would employ the military’s new policy, adopted in 1993, (which of course is not formally before the court) as an indication that the military has implicitly conceded that the Academy regulations (and former DOD Directives) were irrational. That proposition is a non sequitur. In fight of the extremely deferential nature of rational basis review, there would always be a range of policy choices that would meet that standard. A shift from one of those choices to another hardly suggests that the government believes the former was unconstitutional. In any event, under the new policy, Steffan’s statement — which we again emphasize is what this ease is about — would be taken to mean just what the Academy Board apparently thought it meant. The new Directives provide that a “statement by a Service member that he or she is a homosexual ... creates a rebuttable presumption that the Service member engages in homosexual acts or has a propensity or intent to do so.” DOD Directive 1332.14.H.l.b.(2) (Dec. 22, 1993). To be sure, under the new policy, the government explicitly disavows any concern with a service member’s totally private homosexual “orientation,” meaning his “sexual attrac*691tions”; but however that language might be applied in future cases, it obviously would have no relevance to a service member who, like Steffan, has disclosed that he is a homosexual.
Certainly, individuals like Steffan who identify themselves as homosexual in a military setting — where a declaration of homosexuality is grounds for discharge — convey the impression that they are not in doubt as to the direction of their sexual drive. The inference drawn by the government in this sort of case is thus even stronger than it might be in civilian life, where it is more conceivable that an individual would experiment with such an identification. The dissent asserts that the fear of discharge would prevent a self-identified homosexual from actually engaging in homosexual conduct, see dissent at 712, but its reasoning overlooks the point that such fears, if present, would presumably also have discouraged the initial statement, particularly if a person were unsure of his or her identification and its relationship to the military’s definition. Given that the military’s response is the same in each case — discharge—it is unclear why the dissent thinks the deterrent would affect only the later decision.
Even if the assumption that declared homosexuals will engage in homosexual conduct is reasonable in certain contexts, Steffán maintains that it is nevertheless impermissible for the military to act on that assumption; it implies that service members will engage in criminal misconduct — violate the Uniform Code of Military Justice. Steffan argues that such an assumption flies in the face of core traditions of American jurisprudence. Although Steffan’s argument has a certain superficial attractiveness, it seems to us that upon close examination it is more clever than real. First of all, the Academy regulations reach all homosexual conduct — a category of actions that may include conduct that is not illegal under the Code, which proscribes sodomy. More important, we think that when a service member declares or openly admits that he is a homosexual without any explanation, the Academy may rationally take that statement, at least for purposes unrelated to criminal enforcement, as highly likely to be an admission of homosexual conduct or intent. In a discharge proceeding, the Navy need not conduct an inquisition to test whether a particular midshipman possesses an idiosyncratic view of the term. When an individual’s statement can reasonably be taken to evidence a propensity to engage in certain conduct, the military may certainly take that individual at his word.
The authority Steffan presents to support his point does not bear on his case. Steffan principally relies on Jacobson v. United States, — U.S. —, 112 S.Ct. 1535, 118 L.Ed.2d 174 (1992), in which the Supreme Court overturned petitioner’s conviction on charges of receiving pornography through the mail, holding that government agents “may not originate a criminal design” or “implant in an innocent person’s mind the disposition to commit a criminal act.” Id. at —, 112 S.Ct. at 1540. But the government’s obligation to prove predisposition in an entrapment prosecution has no relevance to whether the military may assume that a serviceman identifying himself as a homosexual is likely to engage in homosexual conduct.12 This is not a criminal case; Steffan was not charged with misconduct, and therefore the constitutional protections we accord criminal defendants are not applicable.13 *692Steffan claims that, in contrast to its assumption about homosexuals’ behavior, the military does not make a similar assumption with respect to heterosexuals and that therefore the inference directed against homosexuals reflects impermissible bias. Acts of sodomy by heterosexuals are also misconduct under the Code of Military Justice, and yet the military does not presume that heterosexuals will engage in that practice.14 It is not clear just what military procedure Stef-fan asserts is based on this allegedly contrary assumption about heterosexuals. In any event, as the government responds, to criminalize one form of sexual conduct between heterosexuals is not the same as prohibiting all sexual conduct between homosexuals; the latter puts a much greater restraint on sexual drives — one that the military reasonably believes is difficult, if not impossible, to maintain. The military presumes, in parallel fashion, that homosexuals are as likely to engage in homosexual conduct as heterosexuals are likely to engage in heterosexual conduct. Homosexuals and heterosexuals are, however, differently situated in that heterosexuals have a permissible outlet for then-particular sexual desires whereas homosexuals in the military do not. The temptations facing heterosexuals, moreover, are less compelling than those that homosexuals would encounter, because men and women are quartered separately. They are separated because the military rationally assumes that heterosexuals, like homosexuals, are likely to act in accordance with their sexual drives whether or not such actions would be misconduct. (Under the dissent’s logic — that it is irrational to infer sexual conduct from indicia of sexual tendencies — even separating men and women could be thought unconstitutional.15) The military obviously could not eliminate the difficulties of quartering homosexuals with persons of the same sex by totally segregating homosexuals. Besides the troubling implications of such a separation, putting all homosexuals together would not diminish their mutual sexual attractions. The military’s concerns, then, do not stem from an irrational bias or, as the dissent suggests, “naked stereotypes,” dissent at 708, 712; rather, heterosexuals and homosexuals are treated differently because the means at the military’s disposal for dealing with the natural phenomenon of sexual attraction differ for the two.
We have said that this is not a criminal case. It is also not a First Amendment case. Steffan was not discharged from the Navy because he expressed sympathy for homosexuals or because he openly opposed the Navy’s policy of banning homosexuals, nor has he claimed a First Amendment violation. (To be sure, even the First Amendment must yield at times to the exigencies of military life. Goldman v. Weinberger, 475 U.S. 503, 507-09, 106 S.Ct. 1310, 1313-14, 89 L.Ed.2d 478 (1986); Brown v. Glines, 444 U.S. 348, 354, 100 S.Ct. 594, 599, 62 L.Ed.2d 540 (1980) (iquoting Parker v. Levy, 417 U.S. 733, 759, 94 S.Ct. 2547, 2563, 41 L.Ed.2d 439 (1974))). Steffan’s “thoughts” were not put in issue by the Naval Academy. He was not discharged for imagining the demise of the President or even the Superintendent of the Academy. See dissent at 713-15. Thus, most of the cases relied on by the dissent, both the Supreme Court and circuit opinions, are utterly inapposite.16
We recognize that the government’s policy — that homosexuals (using the ordinary meaning) may not serve in the armed *693forces — is quite controversial. The issue is politically devisive. We think, however, that Steffan’s claim that the government cannot rationally infer that one who states he or she is a homosexual is a practicing homosexual, or is at least likely to engage in homosexual acts, is so strained a constitutional argument as to amount to a basic attack on the policy itself.17 And we think that the language used by the dissent — its invocation of “discrimination,” dissent at 715 n. 20, “fundamental due process,” id., and “fundamental impediments deeply rooted in our constitutional jurisprudence,” id. at 709 —represents a rhetorical effort to bréale out of the narrow constraints of the court’s role on rational basis review.
III. DOD Directives
A.
The DOD Directives use somewhat different wording to define a homosexual than does the Academy regulation:
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), 32 C.F.R. Pt. 41, App. A (1991) (superseded). Steffan’s lawsuit takes aim primarily at the Directives because the phrase “desires to engage in” is claimed to extend the definition of homosexual into the realm of a service member’s private thoughts. The DOD regulations are unconstitutional, Steffan argues, because they allow for the expulsion of service members on the basis of their inner feelings alone. They are irrationally overbroad — and thus fail rational basis review — because they define the class of excludable persons to reach those who merely harbor homosexual impulses, without requiring any indications that such impulses are likely to be reflected in admittedly impermissible homosexual conduct. Steffan’s complaint purports to challenge the Directives as applied and on then-face (although his brief is ambiguous on the point).
As we have noted, however, Steffan’s counsel agreed at oral argument that the Directives constitutionally could be applied to a service member who stated that he was a homosexual and who meant by the statement that he actually engaged in homosexual conduct. This concession, that some situations exist to which the Directives may constitutionally be applied, renders Steffan’s facial challenge defective. “[Ujnder our constitutional system courts are not roving commissions assigned to .pass judgment on the validity of the Nation’s laws.” Broadrick v. Oklahoma, 413 U.S. 601, 610-11, 93 S.Ct. 2908, 2915, 37 L.Ed.2d 830 (1973). For this reason, “[constitutional judgments ... are justified only out of the necessity of adjudicating rights in particular cases.” Id. at 611, 93 S.Ct. at 2915. Accordingly, outside the First Amendment context, see, e.g., Gooding v. Wilson, 405 U.S. 518, 520-22, 92 S.Ct. 1103, 1105-06, 31 L.Ed.2d 408 (1972), a “facial challenge to a legislative Act ... must establish that no set of circumstances exists under which the Act would be valid.” United States v. Salerno, 481 U.S. 739, 745, 107 S.Ct. 2095, 2100, 95 L.Ed.2d 697 (1987) (emphasis added). By his own admission, Steffan cannot meet this requirement.
Where, as here, a statute or regulation has some concededly constitutional applications, a successful challenger must demonstrate that the statute is unconstitutional as “applied to the particular facts of [his] case.” Id. at 745 n. 3,107 S.Ct. at 2100 n. 3. To this end, we are told, see Appellant’s Reply Br. at 10, that when Steffan answered affirmatively to the question whether he was a homosexual he was obliged to reveal, and therefore must be.deemed to have revealed, his innermost desires. Steffan, however, has not at any stage in this litigation sought to show why and how the DOD Directives were actually applied to him in an unconstitutional manner. He never disputed that he was a homosexual; indeed, he forthrightly admitted it at his hearing and declined to add anything to his response. (It will be recalled that he had actually invoked his right to refuse to discuss the matter with Naval investigators.) He did *694not point to that part of the DOD Directives’ definition of homosexual that includes those persons who “desire[ ] to engage in ... homosexual conduct” as the reason for his answer. He never gave any explanation as to what he meant, or did not mean, when he stated that he was a homosexual. And he never claimed (even when he wrote to the Secretary of the Navy a year and a half after his resignation) that it was only the “desires” component of the Directives — the alleged unconstitutionally overbroad aspect — that enmeshed him in the definition.
Normally, a case such as this would arise as an appeal from a military adjudication under the Administrative Procedure Act. The APA provides a right to review agency action only to those “adversely affected or aggrieved” by such action, see 5 U.S.C. § 702 (1988), and in order to establish a grievance resulting from the application of a regulatory term, a party would first have to show whether and how that term was applied to him. This is all the more so when the petition for review is based on a claim of unconstitutionality. Continental Air Lines, Inc. v. Department of Transp., 848 F.2d 1444, 1455-56 (D.C.Cir.1988) (constitutional challenges to congressional statute must be raised in agency proceedings to be preserved for review by appellate court); Meredith Corp. v. FCC, 809 F.2d 863, 872-74 (D.C.Cir. 1987) (remanding for agency initially to address constitutional challenges to its own policy). Although this case comes to us as a collateral challenge based on the Constitution rather than the APA, that route does not alter the desirability of an agency elucidation as to the meaning of that part of the Directive that Steffan claims was crucial to his discharge. Where on the spectrum between “intentions” and mere “fantasies” does the military’s interpretation of the word “desires” fall? Because the only administrative bodies that can decide that question were never pressed by Steffan to do so — indeed were never even alerted to the issue — we cannot know.
Only on appeal (almost seven years after his resignation) did Steffan fix upon the “desires” language in the definition as the root of his legal difficulty. If the Navy’s procedures (which have not been brought before us) require internal appeal before resorting to collateral challenge, see Darby v. Cisneros, —: U.S. —, —, 113 S.Ct. 2539, 2548, 125 L.Ed.2d 113 (1993), it might well have been thought that Steffan had failed to exhaust his administrative remedies by agreeing to the disposition of his case at the Academy. The Navy did raise this precise argument in a recent Ninth Circuit case, Meinhold v. United States Dept. of Defense, 34 F.3d 1469 (9th Cir.1994), in which a naval officer was dismissed under a ruling of a Discharge Board (the lowest level in the administrative proceedings) for having stated on national television, “Yes, I am in fact gay.” The Navy pointed out that further administrative proceedings appealing the decision to the Board for Correction of Naval Records would provide a record for judicial review and would clarify confusion as to the regulatory definition of “homosexual,” which the parties disputed. The Ninth Circuit, rather astoundingly in our view, rejected the exhaustion argument on the grounds that the Discharge Board had been advised that a service member is conclusively deemed to be a homosexual if he makes a statement to that effect (which, of course, is not what the DOD Directives themselves said, see DOD Directive 1332.20.H.l.c.(2), (providing for separation on the basis of statement “unless there is a further finding that the member is not a homosexual”)), and that the Navy had not indicated that the disposition of the case, would be any different on appeal. Meinhold, 34 F.3d at 1477. It should be obvious that this sort of reasoning would destroy the exhaustion-of-remedies doctrine.18
*695In this case, although the government did in fact assert an exhaustion defense of some sort, it was rejected by the district court and the issue was not raised on appeal. It will be recalled also that the documents Steffan signed, with the advice of counsel, to facilitate his resignation specifically acknowledged that he would risk discharge if he pressed his appeal and for that reason explicitly waived his right to seek review. The government apparently initially asserted a variation of a waiver defense as well, which was also rejected and not appealed. As neither the waiver nor exhaustion arguments are before us, it would not be appropriate for us to consider them sua sponte. Cf. National Wildlife Fed’n v. Burford, 835 F.2d 305, 316-18 (D.C.Cir.1987) (discussing district court’s discretion to enforce exhaustion claims raised against plaintiffs by nongovernmental third parties).
Still, in making an as-applied challenge, it is Steffan’s burden — and the government does make this point — to show exactly how the Directives were applied against him illegally. Steffan seeks to meet this obligation by admitting that he entertains homosexual desires and arguing that such admission alone would be sufficient under the DOD Directives to cause his termination. The difficulty with Steffan’s inventive position, however, is that even under his theory of the case he would only have a constitutional claim if his statement had meant that he harbored homosexual desires and yet had neither engaged in or intended to engage in homosexual conduct. Otherwise, the desires portion of the definition would not have been crucial to his discharge.
Steffan insists that the Navy never alleged that he engaged in homosexual acts or intended to do so. That is true but, it seems to us, quite beside the point. After all, the Navy never alleged that Steffan had homosexual “desires.” Steffan openly admitted his homosexuality, and under the circumstances there was no reason for the Navy to proceed further. If Steffan had wished to explain that his admission was based only on his desires — and not his conduct or intentions — he could have done so and joined the issue. The government would have been obliged, in that event, either to rest on his explanation (more naturally, he would have been at least asked what he meant by “desires”) or to pursue further evidence of intent or conduct. In that manner, one could subsequently determine whether the Navy had relied on only the desires portion of the definition and therefore whether it had actually been applied to Steffan.
Indeed, the government in its brief before this court, while conceding that the word “desires” may be ambiguous “in isolation,” tells us that the Department of Defense interprets the term as “conduct related,” ie., bordering on intent, and referring to the actual “prospect of future acts.” If the Directives were interpreted in this way — notwithstanding their different language — to mean essentially the same as the Academy regulations, then, as we have already held, it would pass constitutional muster. If, instead, the word “desires” were interpreted to extend the definition of homosexual to persons not likely to engage in homosexual conduct, a different question would be presented; but Steffan has not indicated, in his framing of the case, that he is affected by that possible (if far-fetched) definition.
The dissent nevertheless argues that “nothing Steffan could have said before the Performance Board would have been remotely relevant to his predicament.” Dissent at 716-17. We simply do not understand how our colleagues can so conclude. Had Steffan responded “yes” to the question whether he was a homosexual but also said that he meant only that he entertained homosexual thoughts, it is not at all clear to us what the Board would have done. The Board might well have determined that under those circumstances it should enter a “further finding that the member is not a homosexual,” as the Directives contemplate. Or, the Board might then have sought to determine the parameters of the term “desires.” Does the term mean something close to an actual intent to engage in homosexual conduct or, as the original panel in this case concluded, mere “inclinations” and “fantasies.” Steffan v. Aspin, 8 F.3d 57, 66 (D.C.Cir.1993), vacated and rehearing en banc granted (D.C.Cir. Jan. 7, 1994). We must bear in mind that *696although Steffan has implicitly conceded that he harbors homosexual desires, he has never explained how he defines that term.
Steffan was, moreover, amply represented by counsel (compare dissent at 717). If he believed that he was obliged to answer the question affirmatively solely because of the “desires” portion of the DOD Directives, it surely would have occurred to his counsel that the appropriate predicate for a constitutional challenge would include an explanation of what Steffan meant — and did not mean— by his answer.
As the record stands, Steffan has made no effort either in the Academy proceeding, in his subsequent letter to the Secretary of the Navy, or even in his suit in federal court to demonstrate how the Directives’ allegedly overbroad definition of homosexual was applied to him. Putting the proceedings before the Board aside, it seems to us that Steffan has no conceivable basis, once he came to federal court with a collateral challenge, for not stating in his allegations exactly how the “desires” portion of the Directives injured him. Steffan seeks to blur the distinction between an as-applied and a facial challenge; he wishes to attack the Directives at the point he believes they are weakest legally without shouldering the burden requisite to such an attack — a demonstration that the supposedly irrationally overbroad definition had a real impact on him. This he cannot do.
To be sure, as the dissent emphasizes, dissent at 717-18, this case has been before this court before. In Steffan v. Cheney, 920 F.2d 74 (D.C.Cir.1990) (per curiam), we reversed the district court’s dismissal of the case as a sanction for Steffan’s refusal to answer deposition questions about whether he had engaged in homosexual conduct during or after his tenure as a midshipman. Steffan asserted his Fifth Amendment privilege and also objected that the questions were not relevant. The district court disagreed on the theory that Steffan’s conduct, even subsequent to dismissal, would be germane to the question of remedy — that is, whether, if he had been illegally dismissed, he could still be ordered commissioned. We held that the government was obliged to defend the discharge based on the administrative record before the Board at the time of its action and disagreed with the district court as to its remedial rationale. Id. at 76.
We decidedly did not say that Steffan’s conduct prior to his dismissal was categorically irrelevant to the case; we expressly envisioned the possibility that conduct-related issues might be “relevant” on some “other ground.” Id. Neither we nor the district court even considered at that point whether Steffan had an obligation in making an as-applied challenge to indicate that his statement embracing homosexual status was not linked to homosexual conduct, or the intent to engage in that conduct.
The dissent’s assertion that “We also specifically rejected the government’s argument that because of a ‘rebuttable presumption’ or ‘celibate homosexual exception’ in the regulations, Steffan was presumptively caught by the conduct or intent prongs of the regulation until he demonstrated otherwise,” dissent at 718, is simply not accurate, and therefore the dissent’s contention that we have reversed the prior panel decision is incorrect. Our entire treatment of the question was contained in the following footnote:
The Government now argues that Steffan’s admission of homosexuality raised a “re-buttable regulatory presumption that he had a prediliction [sic] to commit, and had committed, homosexual acts.” This argument, not raised in the district court, finds no support in the record.
Id. at 76 n. * (emphasis added). As we contemplated, the government, on remand, did present the argument in the district court, which is why the next time the case appeared before this court, see Steffan v. Aspin, 8 F.3d 57, 64-65 (D.C.Cir.1993), vacated and reh’g en banc granted (D.C.Cir. Jan 7, 1994), the panel treated the issue on the merits and did not assert that it was foreclosed, as the dissenters now argue, by the law of the case doctrine. Dissent at 718 & n. 25 (citing Steffan v. Aspin, 8 F.3d at 64-65).19
*697B.
Alternatively, we think Steffan lacks standing to bring his particular challenge to the DOD Directives. Steffan may well have Article III standing, which requires only injury in fact (his termination from the Academy) that is fairly traceable to the Directives (assuming, for purpose of this section of our opinion, that the Directives rather than the Academy regulations were the basis for his termination). But a litigant must still pass through the prudential standing barrier, and Steffan does not. Prudential standing is of course, like Article III standing, a jurisdictional concept. Normally we would focus on that issue first before considering whether Steffan brought a viable claim. But in this case the two issues are quite intertwined; it is only after a careful dissection of Steffan’s claim that it also becomes apparent that he lacks standing to litigate the issue he wishes us to decide.20
Absent certain limited exceptions — such as in the First Amendment context, see, e.g., Gooding v. Wilson, 405 U.S. 518, 520-22, 92 S.Ct. 1103, 1105-06, 31 L.Ed.2d 408 (1972), where some impediment to the assertion by a rightholder of his or her own rights exists, see Powers v. Ohio, 499 U.S. 400, 410-11, 111 S.Ct. 1364, 1370-71, 113 L.Ed.2d 411 (1991), or where litigants have been allowed to assert the rights of third parties in order to protect a threatened relationship, see Wulff v. Singleton, 428 U.S. 106, 112-16, 96 S.Ct. 2868, 2873-75, 49 L.Ed.2d 826 (1976); Craig v. Boren, 429 U.S. 190, 195, 97 S.Ct. 451, 455-56, 50 L.Ed.2d 397 (1976); see also Fair Employment Council v. BMC Marketing Corp., 28 F.3d 1268, 1277-78 (1994) — prudential standing notions mandate that a plaintiffs suit seek to vindicate his own legal rights or interests, not those of some absent third party. See Warth v. Seldin, 422 U.S. 490, 499-500, 95 S.Ct. 2197, 2205-06, 45 L.Ed.2d 343 (1975); Nordlinger v. Hahn, — U.S. —,—, 112 S.Ct. 2326, 2332, 120 L.Ed.2d 1 (1992). This requirement means that there must be a connection between the injury suffered and the legal right or theory asserted. Prudential standing, then, again like Article III standing, focuses in part on causation. But unlike Article III, which requires a plaintiff merely to show a connection between his injury and the governmental “action” he challenges, see Valley Forge Christian College v. Americans United For Separation of Church and State, Inc., 454 U.S. 464, 472, 102 S.Ct. 752, 758-59, 70 L.Ed.2d 700 (1982), prudential standing requires the existence of a further link between the injury and the legal right asserted by the plaintiff. For if the alleged wrongfulness of the injurious action challenged, that is, the asserted violation of a legal right, was not causally related to the injury suffered— if, in other words, it was not the illegal aspect of the action challenged that harmed the plaintiff — then the suit in question would not be one to vindicate that plaintiffs own rights. One cannot vindicate what has never been threatened.
The legal right or interest on which Stef-fan rests his claim for relief is a right against discharge (which he terms “punishment”) on the basis of his thoughts or “desires” — what he sometimes calls his “status.” That claim only vindicates his own rights or interests if Steffan’s discharge was, as we have noted, due solely to his desires. If, on the other hand, Steffan had by his statement meant that he had actually engaged or intended to engage in homosexual conduct, so as to have qualified as a homosexual under the Directives irrespective of the “desires” clause, then his challenge to that clause would vindicate no cognizable interest of his.
In order for Steffan’s claim for relief to be based on the assertion of his own right or interest, then, the legal wrong he ultimately *698alleges — the inclusion of the term “desires” in the Directives — must have led to his discharge. In turn, for Steffan’s discharge to have been caused by the inclusion of the “desires” clause in the Directives, each of the following five facts would have to be true: Steffan did not engage or intend to engage in homosexual conduct; Steffan read the Directives (or learned about them) and determined that they defined the term “homosexual” to include one who “desires to engage in homosexual conduct”; Steffan understood “desires” to mean a mental state sufficiently removed from “intent” as to approach simple “thoughts” and thereby create a constitutional problem (Steffan, after all, conceded that if the verb “desires” was virtually synonymous with “intent” the Directives would be constitutional21); when Steffan answered “yes” to the question “are you a homosexual,” he meant by that answer only that he had such a “desire” to engage in homosexual conduct; and, finally, the military defined “desires” to mean something similar to that understanding. It is evident that Steffan has failed adequately to allege the existence of any of them.
Taking the last point first, Steffan, as we have already explained, never obliged the Navy to give the term meaning by applying it in an adjudication. Instead, he asks us to assume a meaning, one that is sufficiently expansive to raise a constitutional issue. Perhaps, if the word “desires” had an obvious interpretation in this context Steffan’s claim might be defensible, but as oral argument made clear, that is hardly so. The possible meanings stretch from a publicly expressed passion (which would seem to be practically indistinguishable from intent) to a fleeting private imagining. Steffan does not simply ask us to speculate as to the Navy’s understanding, which would itself be inappropriate — he would have us provoke, rather than avoid, a possibly difficult constitutional issue.
Turning to Steffan — his experiences, his knowledge, and his state of mind when he answered the fateful question affirmatively— the crucial gap in his allegations is the absence of any claim that as of the time he answered he had not engaged or intended to engage in homosexual conduct. If he had, the military would have been entitled even under his constitutional theory to terminate him from the Academy. For under such circumstances, his desires would not be independent of conduct and therefore of no particular relevance to the military — or to him— when he answered the question. Steffan’s causal argument implicitly is that he, not the Board, interpreted the Directives to oblige him to answer the question “Are you a homosexual?” affirmatively, but that claim is not plausible or logical unless at the time Stef-fan’s only reason for answering yes was his desires. Understandably, Steffan has never suggested that he meant only that he harbored homosexual desires and that by desires he referred to something so removed from an intention as to constitute a pure thought. For that matter, he does not claim that he had even heard of the definition of homosexual in the DOD Directives. In sum, Steffan has not even attempted to trace his injury to the legal infirmity in the regulations he would have us address.
Steffan would have us rule on a delicate question of law in what is, in truth, a hypothetical case. For all the record shows, and for all the pleadings reveal, Steffan declared himself to be a “homosexual” because he in fact engaged in homosexual conduct. As applied to such facts, the regulations would be, by Steffan’s own admission, constitutional. Were the military to discharge someone on the basis of a statement that the member was “homosexual,” and were that member to indicate that he had neither engaged nor intended to engage in homosexual conduct, that person would be an appropriate plaintiff to bring a case focusing on the constitutionality of the “desires to engage in” clause in the Directives. This is not such a case.22
*699For the foregoing reasons, the judgment of the district court is affirmed.
So ordered.
. The Academy regulations found in the record postdate Steffan’s separation from the Academy. The parties agree, however, that these regulations are identical, so far as relevant to this appeal, to the version existing at the time of the events giving rise to Steffan’s suit.
. A panel of this court reversed the district court decision in Steffan v. Aspin, 8 F.3d 57 (D.C.Cir. 1993). The full court then vacated the panel’s judgment and ordered that the case be reheard en banc.
. We dismiss Steffan's oblique suggestion, made only in a sketchy footnote and apparently abandoned during oral argument, that heightened scrutiny should be applied because homosexuals constitute a "suspect class” under the Supreme Court's test for identifying such classes. See Padula v. Webster, 822 F.2d 97, 103-04 (D.C.Cir. 1987). As we explained in Padula, if the government can criminalize homosexual conduct, a group that is defined by reference to that conduct cannot constitute a "suspect class.” See id. Indeed, Steffan as much as concedes the point by agreeing that the military can ban those who engage in homosexual conduct.
. See Dronenburg v. Zech, 741 F.2d 1388, 1398 (D.C.Cir.1984) (Navy’s “policy requiring discharge for homosexual conduct is a rational means of achieving ... legitimate interests”); Belter v. Middendorf, 632 F.2d 788, 812 (9th Cir.1980) (upholding “the Navy's blanket rule requiring discharge of all who have engaged in homosexual conduct”), cert. denied, 452 U.S. 905, 101 S.Ct. 3030, 69 L.Ed.2d 405 (1981).
. The regulations state that homosexuality "limit[s] a midshipman’s aptitude and potential for commissioned service" (emphasis added), which might suggest that it would be particularly troublesome for an officer to be a declared homosexual. But the government at oral argument expressly denied that the regulations were crafted with any specific concern for officers. In light of Goldman's admonition that we owe special deference to the "considered professional judgment" of the military officials, we do not think it open to us to draw any distinction between officers and enlisted members.
. The Board discharged Steffan solely on the basis of his admission of homosexuality. Thus, the dissent’s claim that a Naval Investigative Service inquiry into Steffan’s conduct led to “no evidence” of homosexual activity is misleading. See dissent at 701 n. 1. Steffan actually refused to answer conduct-related questions; for purposes of discharge, the Navy presumably would have been justified in drawing adverse inferences as to Steffan’s conduct from this refusal. Baxter v. Palmigiano, 425 U.S. 308, 318, 96 S.Ct. 1551, 1557-58, 47 L.Ed.2d 810 (1976); Lefkowitz v. Cunningham, 431 U.S. 801, 808 n. 5, 97 S.Ct. 2132, 2137 n. 5, 53 L.Ed.2d 1 (1977) (precedent ”permit[s] an inference to be drawn in a civil case from a party's refusal to testify"). But that was not the ground the Navy relied on.
. In Meinhold v. United States Dept. of Defense, 34 F.3d 1469 (9th Cir.1994), the court addressed the “desires” portion of the DOD Directive in a case'involving a serviceman who said on national television, "Yes, I am in fact gay.” A discharge panel was convened to consider Meinhold’s statement and as far as we can determine Mein-hold did not appear before the Board. The Ninth Circuit construed the "desires” language to mean something akin to intent, and therefore concluded that separation could be based on a statement identifying oneself as a homosexual only when it was accompanied by evidence of conduct or intent. Finding that Meinhold’s televised announcement failed to provide any such evidence, compare infra at 694 n. 18 (discussing the appropriateness of the Meinhold court’s factual finding), the court determined that Mein-hold’s discharge was illegal under the Navy's own regulations.
The Ninth Circuit accepted Meinhold’s characterization that the class of persons at issue was those "who say they are gay but have not acted in accordance with their propensity in the past.” In our view, however, the proper characterization of the class is persons who say they are gay, but as to whom the military has no additional evidence as to their conduct. The Meinhold court also did not consider the rationality of treating all persons who identify themselves as homosexuals as likely violators of the prohibition on homosexual conduct.
. The dissent would also distinguish Beazer on the basis of the district court's findings that some *689of the participants in the methadone treatment program exhibited physical symptoms that impaired their capacity for employment. Dissent at 715. The Supreme Court did not, however, rely on this basis in upholding the policy; indeed, the Court recognized that some participants in the program, if “examined individually, satisfied the Transit Authority's employment criteria,” but • nevertheless held that "it is of no constitutional significance that the degree of rationality is not as great with respect to certain ill-defined sub-parts of the classification as it is with respect to the classification as a whole.” Id., at 593, 99 S.Ct. at 1370 (citation omitted).
. Classifications based on race or religion, of course, would trigger strict scrutiny.
. The dissent mistakenly asserts that our view on this issue is contradicted by the DOD Directives and the government's position in Selland v. Aspin, 832 F.Supp. 12, 13 (D.D.C.1993). Dissent at 710. We repeatedly emphasize that the military’s practice of discharging service members who state that they are homosexual, based on an inference of future conduct, is permissible because the inference is rational in the run of cases. Contrary to the dissent's assertions, the existence of possible individual exceptions to the rule does not affect our equal protection analysis.
. Lambda, the gay rights organization representing Steffan, appeared as amicus in Bowers. Arguing against the constitutionality of criminalizing homosexual sodomy, it asserted that the “regulation of same sex behavior constitutes the total prohibition of an entire way of life” because homosexuality is inexorably intertwined with "homosexual conduct.” See Amicus Curiae Brief on Behalf of Respondents by Lambda Legal Defense and Education Fund, Inc. at 23, n. 28, Bowers v. Hardwick, 478 U.S. 186, 106 S.Ct. 2841, 92 L.Ed.2d 140 (No. 85-149) (emphasis added).
. The dissent’s reliance on Robinson v. California, supra, is also misplaced. The Court’s holding in that case that a state statute criminalizing the "illness” of narcotics addiction was cruel and unusual punishment bears little legal or factual similarity to this case. The Robinson Court had no occasion to consider the rationality of an employment decision based on statements from which the probability of future conduct can easily be inferred.
. The other cases relied on by Steffan are similarly inapposite. Noto v. United States, 367 U.S. 290, 81 S.Ct. 1517, 6 L.Ed.2d 836 (1961), involved the sufficiency of evidence necessary to convict a Communist Party member under the Smith Act for advocating the violent overthrow of the government. United States v. Robel, 389 U.S. 258, 88 S.Ct. 419, 19 L.Ed.2d 508 (1967), held that a criminal statute prohibiting employment of Communist Party members at certain defense facilities violated the guarantee of freedom of association. Both these First Amendment cases require the government to allege particular evidence of guilt against individual members of political organizations in criminal prosecutions; they do not affect our analysis of wheth*692er the military's employment restrictions in this case are rational under the Equal Protection Clause.
.As we have said, however, the Navy’s presumption of homosexual conduct does not necessarily imply illegal acts, i.e., sodomy. Although the dissent implies that there may be significant distinctions, see dissent at 705-06 n. 8, Steffan, for his part, never distinguished in his concession between sodomy and other homosexual conduct.
. As a sex-based classification, separate quartering for men and women would be reviewed under the more stringent intermediate-level equal protection standard.
. While the Fourth Circuit did supplement its First Amendment analysis in Gay Alliance of Students v. Matthews, 544 F.2d 162, 167 (4th Cir.1976), with the brief statement that withholding recognition from a gay students’ group also denied equal protection, it did so under a "tailored to serve a substantial government interest standard,” not rational basis review.
. Because we find the Academy regulations rationally justified on these grounds, we see no need to discuss other possible rationales presented by the government or that we might conceive.
. Paradoxically, the court proceeded to advance its own interpretation of the "desires” language of the DOD Directives, concluding that — in order to avoid a supposed but not clearly identified constitutional issue — it must mean that a statement of homosexuality “mandates separation only when that statement should be interpreted as portraying a concrete, expressed desire to commit homosexual acts." Then, in an extraordinary reversal of the role of an appellate court, the Ninth Circuit determined, without remand to the Department or the district court, that as a matter of fact Meinhold's statement manifested no such "concrete" desire under the circumstances (presumably because made on national television). Id. at 1479.
. Our conclusion does not depend on the existence of a so-called “celibate homosexual exception." On this record, we cannot determine whether the government would have formulated *697such a concept in applying the DOD Directives, or what it would mean. Although the government has used the term in its briefs, our opinion does not rely upon it, and we therefore need not reach Steffan's "due process” argument that he could not have known about that "exception.” It is indisputable that the Directives provide that the Board could determine that a member was not a homosexual even if he stated that he was a homosexual, and it was surely open to anyone to force the government to elaborate upon the meaning of “desires.”
. Standing can be raised at any point in a case proceeding and, as a jurisdictional matter, may be raised, sua sponte, by the court.
. In fairness to Steffan, his counsel vigorously disputed that the term “desires” could have such a meaning. But see Meinhold, 34 F.3d at 1479 (interpreting "desires” as used in the Directives to have essentially the same meaning as intent).
. The dissenters disagree with our conclusions on the constitutionality of the regulations and insist that Steffan was wrongly discharged. They presumably also dissent from the implicit denial of relief. And yet they do not decide the question *699on which the government sought rehearing, i.e., whether separation-of-powers principles permit the court to order relief in such a case — a question which, despite their characterization, is hardly “hypothetical” if one concludes that Stef-fan was unconstitutionally dismissed. See dissent at 720 n. 27.
. The government at no time put forth the majority's novel argument that Steffan had failed to exhaust his administrative remedies because the BCNR or some other agency could have granted relief by finding that some individuals who state “I am a homosexual" do not fit the Navy’s definition of "homosexual.” Maj. op. at 695. To the contrary, the government made clear that if Stef-fan were reinstated to the Academy, he would then “undoubtedly” be discharged. They wrote: "Officers who admit that they fit within the Navy's regulatory definition of homosexuality are inevitably discharged_ The regulation requires discharge in all cases.” Steffan, 733 F.Supp. 115, 120 (D.D.C.1989) (citing Defendants' Reply Brief at 21 n. 8).
The Ninth Circuit, in Meinhold v. U.S. Dep’t of Defense, 34 F.3d 1469 (9th Cir.1994), recently rejected the argument that "further review could have made a difference as the parties differed over the regulatory definition of 'homosexual.' " The court held that further administrative review would have been futile because there was "nothing in the record [to] suggest[ ] that a disposition by the BCNR could have differed [from the disposition of the discharge board].” Id. at 1474. The same is, of course, true in this case.