The panel opinion, Brown v. Oneonta, 221 F.3d 329 (2nd Cir.2000), fails adequately to deal with a fundamental issue raised by the plaintiffs’ allegations. It does so in contravention of established precedents of the Supreme Court of the United States and of our court. Because of this failure, the panel is led, unnecessarily I think, to express views on a topic that is both complex and divisive. All this occurs in a context — police investigations in which race is a factor — that implicates some of the deepest and most searing questions in our society. I believe that review of the panel’s opinion is essential, and hence respectfully dissent from the denial of a rehearing in banc.
I
The facts in this case speak for themselves. They can be stated simply and are taken, for the most part, from the panel opinion.
Oneonta, a small town in upstate New York about sixty miles west of Albany, has about 10,000 full-time residents. In addition, some 7,500 students attend and reside at the State University of New York College at Oneonta (“SUCO”). The people in Oneonta are for the most part white. Fewer than three hundred blacks live in the town, and just two percent of [approximately 150 out of 7,500] students at SUCO are black.
On September 4, 1992, shortly before 2:00 a.m., someone broke into a house just outside Oneonta and attacked a seventy-seven-year-old woman. The woman told the police who responded to the scene that she could not identify her assailant’s face, but that he was wielding a knife; that he was a black man, based on her view of his hand and forearm; and that he was young, because of the speed with which he crossed her room. She also told the police that, as they struggled, the suspect had cut himself on the hand with the knife. A police canine unit tracked the assailant’s scent from the scene of the crime toward the SUCO campus, but lost the trail after several hundred yards.
The police immediately contacted SUCO and requested a list of its black male students. An official at SUCO supplied the list, and the police attempted to locate and question every black male student at SUCO. This endeavor produced no suspects. Then, over the next several days, the police conducted a “sweep” of Oneonta, stopping and questioning non-white persons on the streets and inspecting their hands for cuts. More than two hundred persons were questioned during that period, but no suspect was apprehended. Those persons whose names appeared on the SUCO list and those who were approached and questioned by the police, believing that they had been unlawfully singled out because of their race, decided to seek redress.
Id. at 334.
As the opinion goes on to note, despite the description given by the victim, “at least one woman, Sheryl Champen, was [allegedly] stopped by law enforcement of*780ficials during their sweep of Oneonta.” Id. at 338. In addition, though the panel does not mention it, the plaintiffs also alleged that:
During the “sweep,” which occurred over a five day period, the officials, without any basis for suspecting any individual approached except for his or her race, attempted to stop, question, and physically inspect the hands of any and every non-white person in and around the City of Oneonta. Second Amended Complaint, at page 4, lines 1-5.
Also not mentioned by the panel is that the breadth of the sweep allegedly was no accident. For, as is further asserted by the plaintiffs, “the objective, as defendant Chandler [the State Police Investigator in charge of the probe] told the newspapers Newsday and the Oneonta Daily Star, was to try ‘to examine the hands of all the black people in the community.’ ” Second Amended Complaint, at page 30, ¶ 100.
In other words, according to the plaintiffs’ allegations, on the basis of a victim’s statement that her assailant was a young black male, the police in Oneonta, instructed by Chandler, decided to stop and question (a) every male black student at SUCO, regardless of age; (b) every non-white person they could find in the City of Oneonta, regardless of age or sex; and (c) at least one black woman named Sheryl Champen, who has joined this suit as a plaintiff.
Because the district court dismissed the plaintiffs’ complaint on a 12(b)(6) motion, all of these facts, and any others that may be shown in support of the allegations, must be taken as true, Dangler v. New York City Off Track Betting Corp., 193 F.3d 130, 138 (2d Cir.1999), unless they are fanciful or delusionary, or, if instead of facts, they represent only legal conclusions. See Cooper v. Parsky, 140 F.3d 433, 440 (2d Cir.1998) (“[B]ald assertions and conclusions of law are insufficient” to survive dismissal.); cf. Contemporary Mission, Inc. v. United States Postal Serv., 648 F.2d 97, 107 n. 14 (2d Cir.1981) (permitting summary judgment if the plaintiff adduced only facts that were “fanciful, frivolous, gauzy, spurious, irrelevant....”).
II
The plaintiffs brought suit claiming that the defendants’ behavior contravened the Equal Protection Clause of the United States Constitution.1 The district court *781thereupon granted a 12(b)(6) motion to dismiss. But, as the panel acknowledges, it did so on an erroneous theory. The lower court believed, incorrectly, that in order to state an equal protection claim plaintiffs were required “... to plead the existence of a similarly situated group of non-minority individuals that were treated differently in the investigation of a crime.”2 Accordingly, after giving plaintiffs an opportunity to amend their pleadings to make such an allegation (which plaintiffs were unable to do), it dismissed their amended complaint.
As the panel properly notes, however,
There are several ways for a plaintiff to plead intentional discrimination that violates the Equal Protection Clause. A plaintiff could point to a law or policy that “expressly classifies persons on the basis of race.” [Hayden v. County of Nassau, 180 F.3d 42, 48 (2d Cir.1999)] (citing Adarand Constructors, Inc. v. Pena, 515 U.S. 200, 213, 227-29, 115 S.Ct. 2097, 132 L.Ed.2d 158 (1995)). Or, a plaintiff could identify a facially neutral law or policy that has been applied in an intentionally discriminatory manner. See Yick Wo v. Hopkins, 118 U.S. 356, 373-74, 6 S.Ct. 1064, 30 L.Ed. 220 (1886). A plaintiff could also allege that a facially neutral statute or policy has an adverse effect and that it was motivated by discriminatory animus. See Village of Arlington Heights v. Metropolitan Hous. Dev. Corp., 429 U.S. 252, 264-65, 97 S.Ct. 555, 50 L.Ed.2d 450 (1977); Johnson v. Wing, 178 F.3d 611, 615 (2d Cir.1999).
221 F.3d at 337. And, as the panel also says, plaintiffs are therefore correct “that it is not necessary to plead the existence of a similarly situated non-minority group when challenging a law or policy that contains an express, racial classification. These classifications are subject to strict judicial scrutiny, see Able v. United States, 155 F.3d 628, 631-32 (2d Cir.1998).” Id.
Accordingly, the issue before the panel was: Did the plaintiffs adequately plead that state officials imposed a constitutionally suspect classification? It is this question that the panel, in my view, fails adequately to answer. For, instead of considering, on the facts that we must take as true, whether the police created and acted upon a racial classification by setting aside all but the racial elements in the victim’s description, the panel examines the purely hypothetical question of whether, had the police acted on the victim’s description, such behavior would have imposed a racial classification. See id.
It follows that the panel’s answer to this question is irrelevant to deciding the controversy actually before us. That controversy, not addressed to any significant degree by the panel, remains the following: Is the state creating an express racial classification that can only be approved if it survives strict scrutiny when state officers (like the police) ignore essentially everything but the racial part of a victim’s description, and, acting solely on that racial element, stop and question all members of that race they can get hold of, even those who grossly fail to fit the victim’s description? The answer to that question, all but ignored by the panel, seems to me — both on the precedents and on plain logic — to be a resounding yes.3 See, e.g., *782Adarand Constructors, Inc. v. Pena, 515 U.S. 200, 235, 115 S.Ct. 2097, 132 L.Ed.2d 158 (1995) (“[W]e hold today that all racial classifications, imposed by whatever federal, state, or local governmental actor, must be analyzed by a reviewing court under strict scrutiny.”); Saenz v. Roe, 526 U.S. 489, 505, 119 S.Ct. 1518, 143 L.Ed.2d 689 (1999) (holding that where the right “to be treated equally” is at stake, “discriminatory classification is itself a penalty”); see also Hayden, 180 F.3d at 48; Able, 155 F.3d at 631-32.
If I am correct, then the only way in which it can be said that plaintiffs have not sufficiently pleaded a violation of the equal protection of laws is if they have failed to allege facts that, if proven, would show that the police of Oneonta discounted the non-racial elements in the victim’s description while focusing, instead, solely or predominantly on the fact that the perpetrator was black. It is, therefore, to the adequacy of the plaintiffs’ pleadings— again virtually unexamined by the panel opinion — to which I turn.
Ill
That the facts pleaded are adequate seems to me manifest. In addition to the allegation — acknowledged by the panel— that the police stopped at least one woman, the Second Amended Complaint states at page 4, lines 1-5:
During the “sweep,” which occurred over a five day period, the officials, without any basis for suspecting any individual approached except for his or her race, attempted to stop, question, and physically inspect the hands of any and every non-white person in and around the City of Oneonta.
And as the complaint says at page 30, ¶ 100, the objective of the sweep as described by the head state police investigator was “ ‘to examine the hands of all the black people in the community.’ ”
Furthermore, the Second Amended Complaint, on page 3, alleges that the “defendants law enforcement officials ... obtained ... a list containing the name ... of all the male African-American students of SUCO [and] then sought out, approached, questioned, seized, and/or searched every person on that list.” Like the prior allegations, this assertion forms a proper basis for proving that the defendants ignored the victim’s description and focused almost exclusively on its racial elements. (This will be so if plaintiffs demonstrate — what may well be the case — that some African-Anerican students at SUCO, a state college after all, were not young.)
*783If words mean anything, these statements in the pleading, which are unequivocal and non-conclusory, support precisely the claim that the police went beyond the victim’s description and created their own racial classification.4 The complaint, moreover, nowhere contradicts any of the foregoing factual assertions
The only way, therefore, that the plaintiffs’ allegations could be deemed inadequate would be if these pleadings, these facts, could be termed “bald” or “fanciful.” That they are neither is, however, demonstrated by the panel’s own account, “... the police conducted a ‘sweep’ of Oneonta, stopping and questioning non-white persons on the streets and inspecting their hands for cuts. More than two hundred persons were questioned during that period, but no suspect was apprehended” [while] “[f]ewer than three hundred blacks live in the town.”
Since it is extremely unlikely that out of “fewer than three hundred blacks” (apart from black SUCO students who had already been questioned), more than two hundred were young males, it is anything but fanciful or bald to suggest, as the complaint expressly claims, that the police questioned virtually all blacks they could find, and intentionally did so regardless of age and sex. Let me be clear: I am not saying that the complaint alleges this demographic data. It doesn’t need to. It alleges facts as to police behavior that, if proven, support a finding that the police went beyond the victim’s description and created their own racial category. That is enough to get by 12(b)(6),5 unless the factual allegations are so fanciful that the court can ignore them. What the conceded demographic data do is to make clear beyond peradventure that the plaintiffs’ assertions, far from being unlikely or fanciful, may well be true.
IV
How, then, can it be that allegations adequate to mandate the denial of a 12(b)(6) motion were deemed to be lacking? The only possibility that comes to mind is plainly wrong. The plaintiffs did not assert the specific legal theory on the basis of which the facts alleged would, if proven, constitute a violation of equal protection law (unless defendants’ actions survived strict scrutiny). That is, the plaintiffs did not articulate the following legal conclusion: “The facts alleged show that the police went beyond the victim’s description and therefore created a racial classification.” But, as is universally recognized, plaintiffs are required to plead facts not legal theories. It follows that a statement of a specific legal theory is in no way needed for a pleading to survive a 12(b)(6) dismissal motion. See Simonton v. Runyon, 232 F.3d 33, 36 (2d Cir.2000) (citing Marbury Mgt., Inc. v. Kohn, 629 F.2d 705, 711 n. 4 (2d Cir.1980)). As a result, the plaintiffs’ failure to articulate the precisely correct theory cannot justify the panel’s reaching out to propound its view of what the law would be if (contrary to the facts that must be taken as true) the police had followed the victim’s description.
The failure of the plaintiffs to articulate the race classification theory, however, does raise the possibility that, were we to go in banc, the plaintiffs might disavow that theory, and the in banc would then collapse. This failure, while it does not make the panel decision less egregious, *784could, for this reason, perhaps form the basis for a decision not to rehear that opinion in bane. But if this were the ground for declining to correct the panel’s errors in banc, it would become essential to remand the case to the district court, so that the plaintiffs could clarify their position by amending their complaint. Such a remand would, of course, make totally unnecessary any legal pronouncements on what the law would be if the police had followed the victim’s description.
This approach would have several advantages, apart from obviating the need for this dissent. First, it would demand of the plaintiffs that they make clear the link between the facts they alleged and the above mentioned legal theory (which is the one most proximately supported by these facts). This would helpfully tie the facts to the theory without risking the possibility of a misfired in banc. And second, it would recognize that (even on the contrary-to-fact assumption that the plaintiffs’ original pleadings were inadequate to support their equal protection claim) the plaintiffs nonetheless deserve an opportunity to make their pleadings good.
On our precedents, plaintiffs are regularly and properly given at least one chance to amend their complaint in response to a district court’s finding of inadequacy. See Branum v. Clark, 927 F.2d 698, 705 (2d Cir.1991) (court should not dismiss without granting leave to amend at least once when a liberal reading of the complaint gives any indication that a valid claim might be stated); Fed.R.Civ.P. 15(a) (leave to amend “shall be freely given when justice so requires”); Tarshis v. Riese Org., 211 F.3d 30, 39 (2d Cir.2000) (citing Leatherman v. Tarrant County Narcotics Intelligence and Coordination Unit, 507 U.S. 163, 167-69, 113 S.Ct. 1160, 122 L.Ed.2d 517 (1993)) (stating “the policy of liberally construing civil rights complaints”); Mian v. Donaldson, Lufkin & Jenrette Securities Corp., 7 F.3d 1085, 1087 (2d Cir.1993).
Here the plaintiffs were seemingly offered such an opportunity. But the “chance” that they were given was in fact totally useless. The district court said to the plaintiffs, “Amend to allege that others like situated were treated differently.” It did not say, “Your allegations are insufficient to support a finding that the police ignored large parts of the victim’s description and in doing so created a racial classification. Amend to allege facts that, if proven, would show this.” The district court was not, of course, under any obligation to point plaintiffs in the direction they needed to go. But it should not have told them both to go in the ivrong direction and that this was the only direction available.6 It follows that, even if the facts pleaded were deemed insufficient to assert a racial classification, the plaintiffs were owed the opportunity to assert additional facts that, if proven, would be sufficient.
Nevertheless, the panel declined to require that such an opportunity to replead be given.7 All the panel did was to allow the district court to consider permitting repleading. {See, footnote 9 of the panel opinion which states, “To the extent that this opinion clarifies equal protection law, the district court is free on remand to entertain a motion to replead. We express no opinion on the merits of any such motion.”) I will have more to say about the, not insignificant, effect of that footnote in a later part of this dissent. At the moment, however, it is enough to state that the failure to require that the plaintiffs be allowed to make a further amendment constitutes a serious deviation from correct practice that, by itself, would justify a rehearing in banc.
V
More broadly, two fundamental problems with the panel’s opinion justify in *785banc review. First, the panel errs in avoiding the critical issue that the plaintiffs’ factual allegations have raised — the creation in this case of a racial classification as a result of police deviation from the victim’s description. Second, that deficiency is compounded by the panel’s reaching out to decide the highly divisive, and, it seems to me, unripe, question of whether and when following a victim’s description is acceptable. Converting what would otherwise be dicta into what sounds like a statement of law is almost always undesirable. In the circumstances before us, it is especially unfortunate.
Why is this so? The first reason is that by doing this, the panel prematurely legitimates actions that — even if they might ultimately be deemed valid — are, as the panel itself recognized, extremely offensive to a much abused part of our population. See 221 F.3d at 339. However many heartfelt apologies the panel makes for doing so, this cannot help but hurt. If, as the plaintiffs alleged, the police did not merely follow the victim’s description in questioning every male black student and two thirds of all of the black residents of the City of Oneonta, I should have thought it wise for the court to welcome the opportunity these allegations gave it to avoid having to tell African-Americans that we are sorry, but you just have to put up with racially linked sweeps when victims — perhaps influenced by their own racial fears, or by our country’s long history of racial divisions — give an essentially racial description.8
But there are also other, structural, reasons why the panel’s, to me unnecessary, validation of the police sweep is particularly undesirable. The question of when, if ever, merely following a victim s description that is predominantly racial might violate equal protection norms is an extremely difficult one. A couple of examples will suggest why. Suppose an armed robbery occurs in which the victim cuts the arm of the robber. The robber, described by the victim in racial terms, runs into a crowded bar where there are only three others who could be so described. Is it wrong for the police to ask the four to show whether they have a cut on their arm? Of course not. But imagine, instead, that a passerby sees someone illegally swimming naked in a park pond and describes the swimmer to the police in racial terms, adding that the swimmer can readily be identified because he has a distinctive tattoo on his posterior. Can it possibly be acceptable for the police to ask every male in town who fits that racial description to strip, even if the police do so with utmost politeness and in full conformity with Fourth Amendment strictures? I would certainly think not.
In between these examples there are any number of permutations involving, among other things, (a) the seriousness of the crime; (b) the number of people in the racially defined group who are subject to questioning; (c) the significance and extent of %0%-racial attributes given by the victim in addition to the racial one; (d) the capacity of the victim to describe the perpetrator in non-racial (as well as in racial) terms; (e) the effort, if any, by the police to elicit from the victim such nonracial descriptions; (f) the intrusiveness of the questioning; and (g) the special indignity (arising from the existence of stereotypes) that may result from connecting *786those in a given racial group with a particular type of crime.
Given the complexity of this issue, courts should surely avoid premature judicial pronouncements of validity (or invalidity).9 In this respect, the revised panel opinion is an enormous improvement over the previous panel decision, Brown v. City of Oneonta, 195 F.3d 111 (2d Cir.1999), now happily vacated. Whereas the earlier version seemed to say that any kind of police behavior that followed a victim’s description was valid (unless expressly motivated by racial animus), the revised opinion seeks simply to approve police actions in this specific case. (It does so, of course, by improperly assuming that the police actually followed the victim’s description.) The fact that the crime here was a serious one, and that the victim’s description was not solely racial, makes the court’s statement, though still unnecessary, less harmful. But harmful it remains.
VI10
Determining where, if anywhere, a court should draw the line between acceptable and unacceptable police behavior in such a morass would be very hard, even in the best of circumstances. It is made yet more difficult by the fact that the legal categories courts currently have available are utterly unsuited to the task. We can leave to one side the properly forbidden instances of behavior motivated by racial animus, because we can assume that at least in most cases the police would be able to point out that their object was to catch the perpetrator and not to discriminate. Similarly, the question of whether the same action would have been taken if the racial description had been of a member of a more favored “racial” group, will (as the panel opinion demonstrates all too well) rarely be helpful in drawing appropriate lines. And so we are left only with the possibility that, perhaps sometimes, following a victim’s racial description can, without more, constitute a racial classification, and as such might be subject to strict scrutiny. But that possibility also turns out to be of little help.
The problem is that the strict scrutiny criteria developed by the Supreme Court are much too blunt. If an action is deemed a racial classification, it is very difficult, under the Supreme Court precedents, ever to justify it. And, were such justification made easier in cases of police following a victim’s description, the spillover to other racial classification contexts would be highly undesirable. In other words, were the requirements of strict scrutiny to be relaxed in the police/victim’s description area, it would be hard indeed to keep them from also being weakened in other areas in which racial classifications ought virtually never to be countenanced. If, instead, following victim racial descriptions by the police were not deemed to be, at least potentially, racial classifications, there would be no constitutional impediment on police sweeps to identify, say, even the racially described naked swimmer.
For these, and other similar reasons, courts should recognize severe limitations on their competence to deal with victim racial descriptions. But limitations do not mean impotence, they mean that courts *787ought to be reluctant to act alone. Rather, courts should encourage legislatures to develop guidelines for this area. Such legislative guidelines could make nuanced distinctions between what is needed and acceptable police behavior, and what is not.11 Courts could then both enforce those guidelines, and if a jurisdiction made distinctions that were inadequately sensitive, perhaps even strike some of them down.
My point is simply this: If courts give a blank check and broadly legitimate police behavior when it consists of blindly following victims’ descriptions that are predominantly racial, legislatures are very unlikely to step in. If, instead, courts try to define, on their own, what is acceptable and what is not, they will probably botch the job terribly. Neither approach is as likely to be as good as one that would derive from a dialogue that could be developed between courts and legislatures.12
By speaking at all when it did not need to, the panel both makes legislative intervention less likely and guides that intervention — were it to occur — prematurely and hence improperly. How much better it would have been, therefore, had the panel taken note of the fact that the plaintiffs’ allegations made this case one in which the police did not merely follow a victim’s description, but instead created their own racial category. The panel would then have reversed the district court while, at the same time, pointing out the difficulties that courts face with respect to situations in which the police do no more than follow victims’ descriptions. Had it done this, the panel would not only have not needed to tell African-Americans that,' sorry as we are, you must put up with demeaning treatment, it would also have furthered a legislative/judicial dialogue that, precisely because it would involve participation by, among others, African-Americans and the police, would have some hope of developing effective and non-hurtful ways of dealing with a very hard problem.13
*788VII
Blessedly,14 this case is far from over. In footnote 9, the panel opinion invites the district court to consider allowing replead-ing. If the proverbial wink in fact turns out to be as good as a nod, the district court will permit a new amended complaint (as I believe it must do). That revised complaint could tie the facts alleged to the specific theory that an equal protection violation here occurred because the police created their own racial classification by deviating in an impermissible way from the victim’s description.
It might then happen, if plaintiffs are able to support the facts they allege sufficiently to avoid summary judgment, that the case would proceed to trial. The result of such a trial could be that — except for its uncontroversial recognition that a prima facie violation of the Equal Protection of the law that is subject to strict scrutiny occurs when a racial classification is created by state action — everything the panel said would be of little significance. And the panel opinion would end up being no more than a minor footnote to a case decided on other grounds.15
CONCLUSION
It may be, therefore, that our failure to go in banc will ultimately not have serious *789consequences; ’tis a consummation devoutly to be wish’d. At the moment, however, we are faced with a panel opinion that, (a) though less bad than its prior version, now happily vacated, still makes unnecessary, and inevitably hurtful, remarks about when following victims’ descriptions involving race is constitutionally permissible; (b) does this by ignoring pleadings that are manifestly sufficient under our 12(b)(6) jurisprudence; and (c) additionally, does it by refusing to require the district court to permit further pleadings despite the fact that earlier repleadings were in response to an incorrect statement of the law by the district court. These errors, moreover, are egregious, and are made in a case that directly involves issues that most searingly divide our society. When such issues are incorrectly dealt with by a panel of our court, an in banc rehearing is, to my way of thinking, not only justified but essential. For that reason, I respectfully dissent from the denial of in banc review.
STRAUB, Circuit Judge, with whom CALABRESI, Circuit Judge, joins, dissenting from the denial of rehearing in banc:I respectfully dissent from the denial of rehearing in banc. This case presents “exceptionally] importan[t]” questions of constitutional law, Fed.R.App. P. 35(a)(2), concerning the manner in and degree to which police investigations that rely upon predominantly racial descriptions given by witnesses are to be scrutinized under the Fourteenth Amendment and the Ku Klux Act of 1871, 17 Stat. 13 (codified at 42 U.S.C. § 1983). The panel reaches a grave conclusion by holding that the police act constitutionally under the Fourteenth Amendment when, based on a witness’s predominantly racial description, they stop every young African American male in town to determine whether he can exclude himself from a vague class of potential suspects that has been defined in overwhelmingly racial terms.1 Even counsel for the defendant-appellees — who icon this appeal before the panel — seems to agree. See Bob Herbert, Breathing While Black, N.Y. Times, Nov. 4, 1999, at A29 (quoting statement by Attorney General of the State of New York that “[w]e won the case, but it makes your skin crawl”) (cited in Plaintiffs Appellants’ Second Petition for Rehearing with Rehearing En Banc at 11). Regardless of whether one agrees with the panel’s apparent reading of the complaint or Judge Calabresi’s rather different view, the legal questions presented by this case remain exceptionally important, and the panel’s conclusion remains quite severe. Indeed, as Judge Calabresi correctly notes, the panel’s reading of the complaint actually requires it to tackle a constitutional question even more complex (and perhaps, therefore, even more worthy of in banc review) than otherwise would have been necessary concerning the man*790ner in which police classifications that rely faithfully upon the descriptions of witnesses are to be reviewed under the Equal Protection Clause. See ante at 785 (Cala-bresi, J., dissenting).
It is far from self-evident that the panel’s summary disposition of that novel and complex constitutional question is correct. Regardless of the source of their descriptions of suspects, police departments and individual police officers acting under color of law have an independent constitutional obligation to ensure that their use of those descriptions comports with the requirements of the Fourteenth Amendment— even if the description they rely upon is faithful to what they have been told by a witness. After all, criminal investigations are conducted by the police, not by witnesses. The witness’s description in this case was given to the police as part of its process — “governmental in character,” Edmonson v. Leesville Concrete Co., Inc., 500 U.S. 614, 621-22, 111 S.Ct. 2077, 114 L.Ed.2d 660 (1991), and conducted under color of law2—of searching for criminal suspects. The racial classification that the witness’s description is alleged to embody could only become actionable under the Fourteenth Amendment on account of the police actually using it, which the plaintiffs allege that they did.
'Contrary to the suggestion of Chief Judge Walker, see ante at 773 (Walker, C.J., concurring), the recitation of that basic proposition in this opinion does not “propose” any rule beyond that which the panel itself suggests in its amended opinion — or at least beyond one of the two contradictory rules the panel suggests. The panel already has conceded, in its amended opinion, that police officers cannot wholly insulate themselves from equal protection review simply by claiming that the description they used in an investigation came, in the first instance, from a witness. See Brown v. Oneonta, 221 F.3d 329, 339 (2d Cir.2000) (suggesting that there may be circumstances in which the police, “when acting on a description of a suspect, violate the equal protection rights of non-suspects, whether or not the police only stop persons conforming to the description of the suspect given by the victim”). With that proposition I fully agree, and in this opinion I do not suggest, much less “propose,” anything more. That statement by the panel, however, contradicts a second proposition implicit in the panel’s original opinion, see Brown v. Oneonta, 195 F.3d 111, 119 (2d Cir.1999), and now made explicit in its amended opinion: that because the description “originated not with the state but with the victim,” Oneonta, 221 F.3d at 338, no race-based *791state action'took place at all in this case. Notwithstanding this dicta in the amended panel opinion, however, the initial source of the police’s description of a suspect bears no constitutional significance. It simply is not relevant, one way or another, to the question of whether state action takes place when the police act upon that description, and the panel’s disposition cannot plausibly rest on that basis. Cf. Palmore v. Sidoti, 466 U.S. 429, 433, 104 S.Ct. 1879, 80 L.Ed.2d 421 (1984) (“Private biases may be outside the reach of the law, but the law cannot, directly or indirectly, give them effect.”).
With no support for its suggestion that no state action took place, the panel is left to rely solely on the assertion that the police interrogations of the plaintiffs in this case involved no racial classification at all. See ante at 772 (Walker, C.J., concurring). The panel reaches that conclusion by noting that the plaintiffs “were not questioned solely on the basis of their race,” but based on a description that contained “also gender and age, as well as the possibility of a cut on the hand.” Oneonta, 221 F.3d at 337 (emphasis added). That proposition may appear to be more plausible than the panel’s suggestion concerning the presence or absence of state action, but still is not self-evidently correct. The fact that a predominantly racial description given by a witness includes other descriptors does not, by itself, make the description “race-neutral.” To be sure, as Judge Calabresi correctly notes, determining whether a witness’s predominantly racial description should be deemed to embody an “express” racial classification or a “race-neutral” classification is extremely difficult, and for that reason, the panel should have made an effort to avoid that constitutional question. See ante at 785-86 & n.9 (Calabresi, J., dissenting); Lyng v. Northwest Indian Cemetery Protective Ass’n, 485 U.S. 439, 445, 108 S.Ct. 1319, 99 L.Ed.2d 534 (1988); Ashwander v. Tennessee Valley Auth., 297 U.S. 288, 346-48, 56 S.Ct. 466, 80 L.Ed. 688 (1936) (Brandeis, J., concurring). The panel, however, treats it as no question at all. Clearly, the panel relied upon some criteria to determine that the presence of other descriptors (age, sex, and the “possibility of a cut on the hand”) was sufficient to render the witness’s description — though predominantly racial — “race-neutral.” What those criteria are, however, is anybody’s guess. Again, contrary to Chief Judge Walker’s intimation, see ante at 772 (Walker, C.J., concurring), I do not “propose” any rule suggesting what those criteria should be. Insofar as it fails to make explicit its own criteria, however, neither does the panel. And given the importance of the questions at issue in this case, that void in the panel opinion seems to demand in banc review, for this Court ought not allow such a severe conclusion to rest on such slender legal justification.
The judges of this Court obviously disagree sharply over the serious and difficult constitutional questions presented in this case, which appear to be of first impression in this Circuit and every other. For that reason alone, if not for any other, this case would seem to demand in banc reconsideration. Cf. Koehler v. Bank of Bermuda (New York) Ltd., 229 F.3d 187, 194 (2d Cir.2000) (Calabresi, J., dissenting from the denial of rehearing in banc); McCray v. Abrams, 756 F.2d 277, 279-80 (2d Cir.1985) (Van Graafeiland, J., dissenting from the denial of rehearing in banc). However, because I agree that the plaintiffs’ extraordinarily detailed, 94-page complaint more than sufficiently pleads facts in support of a claim that the police “created and acted upon a racial classification by setting aside all but the racial elements” in the witness’s description, ante at 781-82 (Cala-bresi, J., dissenting), I join Judge Calabre-si’s opinion. I also note that given the majority’s decision to deny rehearing in banc, I share the view expressed by Judges Sack and Katzmann, see ante at 779 (Sack and Katzmann, JJ., concurring), that the District Court should afford the plaintiffs an opportunity to amend their *792complaint on remand in light of our disposition.
. Some of the plaintiffs also alleged violations of the Fourth Amendment. I believe that the panel opinion's handling of the Fourth Amendment claims of Jamel Champen and Ricky Brown is in direct conflict with the law of two circuits, see Gardenhire v. Schubert, 205 F.3d 303, 314 (6th Cir.2000); McGann v. Northeast Ill. Reg. Commuter R.R., 8 F.3d 1174, 1186 (7th Cir.1993), and highlights an apparent conflict within our own circuit.
The issue raised by these cases concerns when a court and when a fact-finder should determine whether a reasonable person would conclude that he or she was free to leave while being questioned by the police. In United States v. Montilla, 928 F.2d 583, 588 (2d Cir.1991), we indicated that "freedom to leave” was generally a legal question to be decided by a court. In Posr v. Doherty, 944 F.2d 91, 99-100 (2d Cir.1991), we suggested, instead, that it was usually a fact issue for the jury. The revised panel opinion says that, actually, the question is a mixed one of law and fact. See 221 F.3d at 340 ("Whether a seizure occurred is a question of law to be reviewed de novo, while the factual findings underlying that determination are reviewed for clear error.”). Fair enough. But that statement does not explain whether the reasonableness of a questioned person's belief as to his or her freedom to leave is one of the underlying factual findings or is, instead, part of the legal conclusion.
. In an earlier, now vacated, opinion the current panel treated "reasonableness” as part of the legal conclusion, and found against the plaintiffs. In the present opinion, the panel still treats it as "law,” but finds for the plaintiffs. The new result may be more nearly correct on the facts but it may well be wrong on who is the proper decision maker. Treating reasonableness as a legal conclusion is not only in conflict with decisions of other circuits (see supra) but is also inconsistent with much other closely related law. See, e.g., United States v. Tehrani, 49 F.3d 54, 58 (2d Cir.1995) (the "point [at which] a permissible investigative detention ripens into an arrest is a question of fact”); Tenenbaum v. Williams, 193 F.3d 581, 605 (2d Cir.1999) ("[a] juiy could reasonably conclude” that a *781person of reasonable caution would have believed that the seizure at issue was not justified), cert. denied, 529 U.S. 1098, 120 S.Ct. 1832, 146 L.Ed.2d 776 (2000); see also DeMarco v. Sadiker, 952 F.Supp. 134, 140 (E.D.N.Y.1996) (relying on Posr in a different, but not unrelated, situation). At the very least, greater clarity on this point, both in terms of our circuit's own law and in terms of our law’s relation to that of other circuits is sorely needed.
. It is not surprising that the district court erred in this way, given the fact that the parties had not argued the theory of the case that I discuss below.
. I do not, of course, mean to suggest that any seeming police deviation from a victim's de*782scription constitutes the creation of a racial classification. If the police question those people in a racially defined group who range from, say, 5'10" to 6'2", on the basis of a victim's description of a 6' tall man “of that race,” such behavior could hardly be termed a deviation. And, there is no reason for the police to be tied to elements in a victim’s description that are readily alterable within the relevant time frame. Thus if the description includes a red bandana, which can quickly be ditched, it would be absurd to limit the police's questioning to those with red bandanas. Other attributes (trousers, facial hair, even color of hair) are also readily alterable, but only if the time period is long enough. As a result, whether the police focused on the racial elements in the victim's description and ignored the others, thereby creating a racial classification, frequently will depend on the time period and circumstances involved.
Similarly, and more important, even seemingly permanent attributes can be disguised. And if there were evidence that the perpetrator had (or was likely to) disguise his sex, or age, these attributes could also be discounted by the police. But "race” too can be disguised. Thus, unless the police are operating on the basis of stereotypes, or unless there is actual evidence to the contrary, there is little reason for the police to assume that the perpetrator disguised his or her gender or age, rather than race, and as a result, proceed to question only those who fit the racial description.
In any event, no such issues are plausibly before us in this case. And, even if they were, it would be extremely unlikely that they would justify a 12(b)(6) dismissal since the reasons justifying the deviation would generally be best examined on the basis of affidavits or other factual submissions.
. See McLain v. Real Estate Bd. of New Orleans, Inc., 444 U.S. 232, 246, 100 S.Ct. 502, 62 L.Ed.2d 441 (1980); Conley v. Gibson, 355 U.S. 41, 45-46, 78 S.Ct. 99, 2 L.Ed.2d 80 (1957) (holding that “a complaint should not be dismissed for failure to state a claim unless it appears beyond doubt that the plaintiff can prove no set of facts in support of his claim which would entitle him to relief”) (emphasis added); Charles W. v. Maul, 214 F.3d 350, 357 (2d Cir.2000) (citing Conley).
. I am not, of course, assuming that the facts alleged will turn out to be sufficient to convince a jury, or even to avoid summary judgment. But they are more than enough to survive dismissal on 12(b)(6).
. Again, what the district court did was quite understandable. See supra note 2.
. Contrast, in this respect, the concurring opinion of Judges Sack and Katzmann.
. There can be little doubt that what descriptions are given, among many possible ones, reflect a country’s underlying biases. Thus in Italy after the Fascist racial laws were passed in 1938, people, for the first time, came to be described in all sorts of police situations as "Jewish looking.” Similarly, at the time of the Palsgraf case, Palsgraf v. Long Island R.R. Co., 248 N.Y. 339, 162 N.E. 99 (1928), police stated that some of the parties involved were "probably Italians.” "Bomb Blast Injures 13 in Station Crowd,” N.Y. Times, Aug. 25, 1924. Today, we don’t generally describe people that way. But we do say "black," "Hispanic,” etc. That this is so does not necessarily mean courts can forbid the police from acting on such descriptions. It should make us reluctant, however, prematurely to approve of such actions.
. I would argue, just on the basis of what it told the district court in footnote 9, that the panel was obligated to avoid making these pronouncements. The footnote makes it possible that the case belore us will ultimately be concluded without regard to any rules dealing with police behavior that merely follows victims’ descriptions. This being so, there was no need in the current appeal to make any pronouncements on that issue. This is especially true since the pronouncements involve constitutional questions that, under our precepts, courts are to avoid deciding unless it is necessary to do so. Lyng v. Northwest Indian Cemetery Protective Ass’n, 485 U.S. 439, 445, 108 S.Ct. 1319, 99 L.Ed.2d 534 (1988); Horne v. Coughlin, 191 F.3d 244, 246 (2d Cir.1999).
. Judges Parker and Sotomayor, while joining all the other parts of the opinion, do not join this section.
. Legislatures could, for example, if they deemed it appropriate, require the police to request that a victim or witness, who had given a solely racial description, answer questions seeking to elicit significant non-racial attributes of the perpetrator. Defining such detailed requirements — Miranda, to the contrary notwithstanding — is not generally something courts do well, but it fits squarely within the competence of legislatures.
. For these purposes, when I say legislatures I presuppose involvement of the relevant executives as well.
. In answer to all this, Chief Judge Walker and Judge Jacobs seek to make four points. I believe that all of them are incorrect.
They first suggest that the Fourth Amendment is sufficient to protect citizens from discriminatory police behavior and, hence, that applying any equal protection review to police investigations is superfluous. Relatedly, they claim that even highly limited equal protection review would have dire consequences. But any argument that the Fourth Amendment could possibly suffice depends, I think, on two propositions, never mentioned, let alone accepted, in the concurring opinions: that where there is a seizure (like a Terry stop) one of the things that must influence whether that seizure is reasonable under the Fourth Amendment is the presence or absence of racial considerations with respect to the seizure; and further, that such a determination of reasonableness in cases of racial classification is at least open to the application of strict scrutiny criteria akin to those of equal protection. Unfortunately, these assumptions appear to be precluded by the Supreme Court’s decision in Whren v. United States, 517 U.S. 806, 116 S.Ct. 1769, 135 L.Ed.2d 89 (1996), which not only makes the motive of the officers involved irrelevant but suggests that the Equal Protection Clause and not the Fourth Amendment is the appropriate basis for objection when seizures "based on considerations such as race” occur. Id. at 813, 116 S.Ct. 1769.
In any event, even with these assumptions, the Fourth Amendment seems to me manifestly inadequate to deal with the underlying problem. There are, for example, in a society with deep racial divisions, any number of police intrusions on citizens that do not amount to Terry stops or to other forms of searches and seizures cognizable under the Fourth Amendment, but that are, nonetheless, immensely hurtful. Some of these may be necessary in order to control serious crimes. Other intrusions, however, because they are not needed to apprehend the perpetrators of such crimes, or because they deal only with *788trivial violations, cannot be countenanced under our Constitution. See supra at 781-82 n. 3, 785-86. And yet, because no searches or seizures are involved, the Fourth Amendment cannot preclude them. As a result, excluding even a minimal consideration oí equal protection when reviewing police behavior in such cases, far from protecting society from dire consequences, treats every conceivable interest of law and order, however insignificant, as if it were necessarily more important than any interest in not being categorized on the basis of race. And that seems to me clearly untenable.
Second, they state that the police in this case could not possibly have ignored all but the racial elements in the victim’s description because to do so would have been stupid. But that is precisely what the plaintiffs alleged in their complaint (and adduced evidence to support). And — quite apart from the fact that racial stereotyping leading to blatantly stupid behavior is far from unknown in our society, even, I expect, among the police — for the concurrers to disbelieve these allegations is to violate the most elementary rules of decision with respect to 12(b)(6) dismissals.
Third, Chief Judge Walker and Judge Jacobs claim that the equal protection theory discussed in this opinion is both unheard of and academic. But, in fact, as we all know, the suspect nature of racial classifications is thoroughly grounded in the Supreme Court's and in this court's precedents. See, e.g., Whren, 517 U.S. at 813, 116 S.Ct. 1769 ("[T]he constitutional basis for objecting to intentionally discriminatory application of laws is the Equal Protection Clause, not the Fourth Amendment."). Moreover, the application of racial classification jurisprudence to the particular facts of this case can be found fully in what was argued to us, most cogently in the amicus brief in support of the plaintiffs. See Memorandum of Amici NAACP Legal Defense & Educational Fund, Inc., New York Civil Liberties Union, and Center for Constitutional Rights in Support of Petition for Rehearing with Rehearing Bn Banc. Since it is conceded that theories do not have to be stated in the complaint, and — given that the facts alleged may, if proven, demonstrate a violation of equal protection under the theory presented by amici — it seems clear that that theory should, at least, be considered by the district court at the 12(b)(6) stage.
Finally, the concurrers assert that a racial classification is not necessarily created every time the police question someone who does not fit the victim’s description of the perpetrator. After all, the police may only be searching for witnesses and not suspects. Of course. But one must also ask why, in the case before us — according to the allegations that we are required to take as true — all of these so-called witnesses were black, and all were asked to show whether they had the incriminating cut on their wrist.!
. In this respect, I note that a majority of the twelve currently active judges of our court do not wish this case to be treated as ended. Five judges (Judges Kearse, Parker, Straub, Sotomayor, in addition to me) have voted in favor of a rehearing in banc, and two judges (Judges Sack and Katzmann), while voting against a rehearing in banc, have stated unequivocally that they believe the district court should permit the plaintiffs to amend their complaint. See opinion of Judges Sack and Katzmann concurring in the denial of the rehearing in banc, ante at 779.
. See note 7 supra.
. Chief Judge Walker observes that “there have been times and places in this country in which the police have tolerated crime in African American communities.” Ante at 771 (Walker, C.J., concurring). That statement is true enough. It is equally true, however, that there have been, and continue to be, times and places in this country in which victims have been neither accurate nor innocent in their use of race to describe criminal suspects — witness, for example, the Boston Police Department's wrongful arrest of an innocent man in 1989 for the murder of Carol Stuart, based upon a witness's fabricated description of the suspect as being "a black male in his late 20s or early 30s, about 5 feet 10 inches tall and weighing 150 to 160 pounds,” who spoke "in a raspy, 'sing-song' tone.” Sally Jacobs & Diego Ribadeneira, No Wallet, So Killer Opened Fire, Boston Globe, Oct. 26, 1989, at 1; see also Peter J. Howe, From Nightmare to Reality, A City is Reeling, Boston Globe, Jan. 7, 1990, at 1.
No less than when they rely upon racial classifications of their own making, the police impose tremendous social costs upon people of color when they act primarily upon the race-based suspicion of victims and other witnesses. Indeed, at least one commentator has gone so far as to describe the costs of using race-based suspicion in police investigations as imposing a "racial tax” upon people of color. See Randall Kennedy, Race, Crime, and the Law 169-63 (1997). The panel pays insufficient heed to these social costs.
. In this respect, the panel appears to err by requiring the plaintiffs to point to a "law or policy that expressly classifies persons on the basis of race” — which, for the panel, seems to refer only to "established profilefs]” or "regular policies] based on racial stereotypes” — in order to make out a viable claim under the Equal Protection Clause that an express racial classification was used. Brown v. Oneonta, 221 F.3d 329, 337 (2d Cir.2000) (emphasis added and internal quotation marks omitted). "Established profilefs]” and "regular policies] based on racial stereotypes” do not exhaust the universe of possible Equal Protection Clause'violations by the police. To the contrary, "any person ... has the right to demand that any governmental actor subject to the Constitution justify any racial classification subjecting that person to unequal treatment under the strictest judicial scrutiny.” Adarand Constructors, Inc. v. Pena, 515 U.S. 200, 224, 115 S.Ct. 2097, 132 L.Ed.2d 158 (1995) (emphasis added).
It is true that claims under section 1983 against government officials in their official capacities (or against the City of Oneonta itself) must show that the entity’s "policy or custom ... played a part in the violation of federal law.” Hafer v. Melo, 502 U.S. 21, 25, 112 S.Ct. 358, 116 L.Ed.2d 301 (1991). However, the Oneonta defendants also have been sued under section 1983 in their individual capacities — and the plaintiffs need not show any connection to government "policy or custom” to support these individual capacity claims. See id. at 25, 112 S.Ct. 358. Regardless, therefore, of whether police "policy or custom” played any role in this case, dismissal of the plaintiffs’ complaint based on the absence of any "policy or custom” seems inappropriate, given the plaintiffs’ explicit assertion of these individual capacity claims.