Justice Stevens,
with whom Justice White and Justice Blackmun join as to Part I,concurring in the
judgment.
Conduct that creates special risks or causes special harms may be prohibited by special rules. Lighting a fire near an ammunition dump or a gasoline storage tank is especially dangerous; such behavior may be punished more severely than burning trash in a vacant lot. Threatening someone because of her race or religious beliefs may cause particularly severe trauma or touch off a riot, and threatening a high public official may cause substantial social disruption; such threats may be punished more severely than threats against someone based on, say, his support of a particular athletic team. There are legitimate, reasonable, and neutral justifications for such special rules.
This case involves the constitutionality of one such ordinance. Because the regulated conduct has some communicative content — a message of racial, religious, or gender hostility — the ordinance raises two quite different First Amendment questions. Is the ordinance “overbroad” be*417cause it prohibits too much speech? If not, is it “under-broad” because it does not prohibit enough speech?
In answering these questions, my colleagues today wrestle with two broad principles: first, that certain “categories of expression [including ‘fighting words’] are ‘not within the area of constitutionally protected speech,”’ ante, at 400 (White, J., concurring in judgment); and second, that “[e]ontent-based regulations [of expression] are presumptively invalid,” ante, at 382 (majority opinion). Although in past opinions the Court has repeated both of these maxims, it has — quite rightly — adhered to neither with the absolutism suggested by my colleagues. Thus, while I agree that the St. Paul ordinance is unconstitutionally overbroad for the reasons stated in Part II of Justice White’s opinion, I write separately to suggest how the allure of absolute principles has skewed the analysis of both the majority and Justice White’s opinions.
I
Fifty years ago, the Court articulated a categorical approach to First Amendment jurisprudence.
“There are certain well-defined and narrowly limited classes of speech, the prevention and punishment of which have never been thought to raise any Constitutional problem.... It has been well observed that such utterances are no essential part of any exposition of ideas, and are of such slight social value as a step to truth that any benefit that may be derived from them is clearly outweighed by the social interest in order and morality.” Chaplinsky v. New Hampshire, 315 U. S. 568, 571-572 (1942).
We have, as Justice White observes, often described such categories of expression as “not within the area of constitutionally protected speech.” Roth v. United States, 354 U. S. 476, 483 (1957).
*418The Court today revises this categorical approach. It is not, the Court rules, that certain “categories” of expression ' are “unprotected,” but rather that certain “elements” of expression are wholly “proseribable.” To the Court, an expressive act, like a chemical compound, consists of more than one element. Although the act may be regulated because it contains a proseribable element, it may not be regulated on the basis of another (nonproscribable) element it also contains. Thus, obscene antigovernment speech may be regulated because it is obscene,-but not because it is antigovernment. Ante, at 384. It is this revision of the categorical approach that allows the Court to assume that the St. Paul ordinance proscribes only fighting words, while at the same time concluding that the ordinance is invalid because it imposes a content-based regulation on expressive activity.
As an initial matter, the Court’s revision of the categorical approach seems to me something of an adventure in a doctrinal wonderland, for the concept of “obscene antigovernment” speech is fantastical. The category of the obscene is very narrow;’to be obscene, expression must be found by the trier of fact to “appea[l] to the prurient interest,... depie[t] or deserib[e], in a patently offensive way, sexual conduct, [and], taken as a whole, lac[k] serious literary, artistic, 'political, or scientific value.” Miller v. California, 413 U. S. 15, 24 (1973) (emphasis added). “Obscene antigovernment” speech, then, is a contradiction in terms: If expression is anti-government, it does not “lac[k] serious... political... value” and cannot be obscene.
The Court attempts to bolster its argument by likening its novel analysis to that applied to restrictions on the time, place, or manner of expression or on expressive conduct. It is true that loud speech in favor of the Republican Party can be regulated because it is loud, but not because it is pro-Republican; and it is true that the public burning of the American flag can be regulated because it involves public burning and not because it involves the flag. But these anal*419ogies are inapposite. In each of these examples, the two elements (e. g., loudness and pro-Repuhlican orientation) can coexist; in the case of “obscene antigovernment” speech, however, the presence of one element (“obscenity”) by definition means the absence of the other. To my mind, it is unwise and unsound to craft a new doctrine based on such highly speculative hypotheticals.
I am, however, even more troubled by the second step of the Court’s analysis — namely, its conclusion that the St. Paul ordinance is an unconstitutional content-based regulation of speech. Drawing on broadly worded dicta, the Court establishes a near-absolute ban on content-based regulations of expression and holds that the First Amendment prohibits the regulation of fighting words by subject matter. Thus, while the Court rejects the “all-or-nothing-at-all” nature of the categorical approach, ante, at 384, it promptly embraces an absolutism of its own: Within a particular “proscribable” category of expression, the Court holds, a government must either proscribe all speech or no speech at all.1 This aspect of the Court’s ruling, fundamentally misunderstands the role and constitutional status of content-based regulations on speech, conflicts with the very nature of First Amendment jurisprudence, and disrupts well-settled principles of First Amendment law.
*420Although the Court has, on occasion, declared that content-based regulations of speech are "never permitted,” Police Dept. of Chicago v. Mosley, 408 U. S. 92, 99 (1972), such claims are overstated. Indeed, in Mosley itself, the Court indicated that Chicago's selective proscription of non-labor picketing was not per se unconstitutional, but rather could be upheld if the city demonstrated that nonlabor picketing was “clearly more disruptive than [labor] picketing.” Id., at 100. Contrary to the broad dicta in Mosley and elsewhere, our decisions demonstrate that content-based distinctions, far from being presumptively invalid, are an inevitable and indispensable aspect of a coherent understanding of the First Amendment.
This is true at every level of First Amendment law. In broadest terms, our entire First Amendment jurisprudence creates a regime based on the content of speech. The scope of the First Amendment is determined by the content of expressive activity: Although the First Amendment broadly protects “speech,” it does not protect the right to “fix prices, breach contracts, make false warranties, place bets with bookies, threaten, [or] extort.” Schauer, Categories and the First Amendment: A Play in Three Acts, 34 Vand. L. Rev. 265, 270 (1981). Whether an agreement among competitors is a violation of the Sherman Act or protected activity under the Noerr-Pennington doctrine2 hinges upon the content of the agreement. Similarly, “the line between permissible advocacy and impermissible ineitation to crime or violence depends, not merely on the setting in which the speech occurs, but also on exactly what the speaker had to say.” Young v. American Mini Theatres, Inc., 427 U. S. 50, 66 (1976) (plurality opinion); see also Musser v. Utah, 333 U. S. 95, 100-103 (1948) (Rutledge, J., dissenting).
*421Likewise, whether speech falls within one of the categories of “unprotected” or “proscribable” expression is determined, in part, by its content. Whether a magazine is obscene, a gesture a fighting word, or a photograph child pornography is determined, in part, by its content. Even within categories of protected expression, the First Amendment status of speech is fixed by its content. New York Times Co. v. Sullivan, 376 U. S. 254 (1964), and Dun & Bradstreet, Inc. v. Greenmoss Builders, Inc., 472 U. S. 749 (1985), establish that the level of protection given to speech depends upon its subject matter: Speech about public officials or matters of public concern receives greater protection than speech about other topics. It can, therefore, scarcely be said that the regulation of expressive activity cannot be predicated on its content: Much of our First Amendment jurisprudence is premised on the assumption that content makes a difference.
Consistent with this general premise, we have frequently upheld content-based regulations of speech. For example, in Young v. American Mini Theatres, the Court upheld zoning ordinances that regulated movie theaters based on the content of the films shown. In FCC v. Pacifica Foundation, 438 U. S. 726 (1978) (plurality opinion), we upheld a restriction on the broadcast of specific indecent words. In Lehman v. Shaker Heights, 418 U. S. 298 (1974) (plurality opinion), we upheld a city law that permitted commercial advertising, but prohibited political advertising, on city buses. In Broadrick v. Oklahoma, 413 U. S. 601 (1973), we upheld a state law that restricted the speech of state employees, but only as concerned partisan political matters. We have long recognized the power of the Federal Trade Commission to regulate misleading advertising and labeling, see, e.g., Jacob Siegel Co. v. FTC, 327 U.S. 608 (1946), and the National Labor Relations Board’s power to regulate an employer’s election-related speech on the basis of its content, see, e. g., NLRB v. Gissel Packing Co., 395 U. S. 575, 616-618 (1969). *422It is also beyond question that the Government may choose to limit advertisements for cigarettes, see 15 U. S. C. §§ 1331-1340,3 but not for cigars; choose to regulate airline advertising, see Morales v. Trans World Airlines, Inc., 504 U. S. 374 (1992), but not bus advertising; or choose to monitor solicitation by lawyers, see Ohralik v. Ohio State Bar Assn., 436 U. S. 447 (1978), but not by doctors.
All of these cases involved the selective regulation of speech based on content — precisely the sort of regulation the Court invalidates today. Such selective regulations are unavoidably content based, but they are not, in my opinion, “presumptively invalid.” As these many decisions and examples demonstrate, the prohibition on content-based regulations is not nearly as total as the Mosley dictum suggests.
Disregarding this vast body of ease law, the Court today goes beyond even the overstatement in Mosley and applies the prohibition on content-based regulation to speech that the Court had until today considered wholly “unprotected” by the First Amendment — namely, fighting words. This new absolutism in the prohibition of content-based regulations severely contorts the fabric of settled First Amendment law.
Our First Amendment decisions have created a rough hierarchy in the constitutional protection of speech. Core political speech occupies the highest, most protected position; commercial speech and nonobscene, sexually explicit speech are regarded as a sort of second-class expression; obscenity and fighting words receive the least protection of all. Assuming that the Court is correct that this last class of speech is not wholly “unprotected,” it certainly does not follow that fighting words and obscenity receive the same sort of protection afforded core political speech. Yet in ruling that proseribable speech cannot be regulated based on subject *423matter, the Court does just that.4 Perversely, this gives fighting words greater protection than is afforded commercial speech. If Congress can prohibit false advertising directed at airline passengers without also prohibiting false advertising directed at bus passengers and if a city can prohibit political advertisements in its buses while allowing other advertisements, it is ironic to hold that a city cannot regulate fighting words based on “race, color, creed, religion or gender” while leaving unregulated fighting words based on “union membership ... or homosexuality.” Ante, at 391. The Court today turns First Amendment law on its head: Communication that was once entirely unprotected (and that still can be wholly proscribed) is now entitled to greater protection than commercial speech — and possibly greater protection than core political speech. See Burson v. Freeman, 504 U. S. 191, 195, 196 (1992).
Perhaps because the Court recognizes these perversities, it quickly offers some ad hoe limitations on its newly extended prohibition on content-based regulations. First, the Court states that a content-based regulation is valid “[w]hen the basis for the content discrimination consists entirely of the very reason the entire class of speech ... is proscribable.” Ante, at 388. In a pivotal passage, the Court writes:
“[T]he Federal Government can criminalize only those threats of violence that are directed against the President, see 18 U. S. C. § 871 — since the reasons why *424threats of violence are outside the First Amendment (protecting individuals from the fear of violence, from the disruption that fear engenders, and from the possibility that the threatened violence will occur) have special force when applied to the . . . President.” Ibid.
As I understand this opaque passage, Congress may choose from the set of unprotected speech (all threats) to proscribe only a subset (threats against the President) because those threats are particularly likely to cause “fear of violence,” “disruption,” and actual “violence.”
Precisely this same reasoning, however, compels the conclusion that St. Paul’s ordinance is constitutional. Just as Congress may determine that threats against the President entail more severe consequences than other threats, so St. Paul’s City Council may determine that threats based on the target’s race, religion, or gender cause more severe harm to both the target and to society than other threats. This latter judgment — that harms caused by racial, religious, and gender-based invective are qualitatively different from that caused by other fighting words — seems to me eminéntly reasonable and realistic.
Next, the Court recognizes that a State may regulate advertising in one industry but not another because “the risk of fraud (one of the characteristics... that justifies depriving [commercial speech] of full First Amendment protection...)” in the regulated industry is “greater” than in other industries. Ibid. Again, the same reasoning demonstrates the constitutionality of St. Paul’s ordinance. “[Q]ne of the characteristics that justifies” the constitutional status of fighting words is that such words “by their very utterance inflict injury or tend to incite an immediate breach of the peace.” Chaplinsky, 315 U. S., at 572. Certainly a legislature that may determine that the risk of fraud is greater in the legal *425trade than in the medical trade may determine that the risk of injury or breach of peace created by race-based threats is greater than that created by other threats.
Similarly, it is impossible to reconcile the Court’s analysis of the St. Paul ordinance with its recognition that “a prohibition of fighting words that are directed at certain persons or groups ... would be facially valid.” Ante, at 392 (emphasis deleted). A selective proscription of unprotected expression designed to protect “certain persons or groups” (for example, a law proscribing threats directed at the elderly) would be constitutional if it were based on a legitimate determination that the harm created by the regulated expression differs from that created by the unregulated expression (that is, if the elderly are more severely injured by threats than are the nonelderly). Such selective protection is no different from a law prohibiting minors (and only minors) from obtaining obscene publications. See Ginsberg v. New York, 390 U. S. 629 (1968). St. Paul has determined — reasonably in my judgment — that fighting-word injuries “based on race, color, creed, religion or gender” are qualitatively different and more severe than fighting-word injuries based on other characteristics. Whether the selective proscription of pro-seribable speech is defined by the protected target (“certain persons or groups”) or the basis of the harm (injuries “based on race, color, creed, religion or gender”) makes no constitutional difference: What matters is whether the legislature’s selection is based on a legitimate, neutral, and reasonable distinction.
In sum, the central premise of the Court’s ruling — that “[cjontent-based regulations are presumptively invalid”— has simplistic appeal, but lacks support in our First Amendment jurisprudence. To make matters worse, the Court today extends this overstated claim to reach categories of hitherto unprotected speech and, in doing so, wreaks havoc in an area of settled law. Finally, although the Court reeog-*426nizes exceptions to its new principle, those exceptions undermine its very conclusion that the St. Paul ordinance is unconstitutional. Stated directly, the majority’s position cannot withstand scrutiny.
II
Although I agree with much of Justice White’s analysis, I do not join Part I-A of his opinion because I have reservations about the “categorical approach” to the First Amendment. These concerns, which I have noted on other occasions, see, e. g., New York v. Ferber, 458 U. S. 747, 778 (1982) (opinion concurring in judgment), lead me to find Justice White’s response to the Court’s analysis unsatisfying.
Admittedly, the categorical approach to the First Amendment has some appeal: Either expression is protected or it is not — the categories create safe harbors for governments and speakers alike. But this approach sacrifices subtlety for clarity and is, I am convinced, ultimately unsound. As an initial matter, the concept of “categories” fits poorly with the complex reality of expression. New dividing lines in First Amendment law are straight and unwavering, and efforts at categorization inevitably give rise oiily to fuzzy boundaries. Our definitions of “obscenity,” see, e. g., Marks v. United States, 430 U. S. 188, 198 (1977) (Stevens, J., concurring in part and dissenting in part), and “public forum,” see, e. g., United States Postal Service v. Council of Greenburgh Civic Assns., 453 U. S. 114, 126-131 (1981); id., at 136-140 (Brennan, J., concurring in judgment); id., at 147-151 (Marshall, J., dissenting); id., at 152-154 (Stevens, J., dissenting) (all debating the definition of “public forum”), illustrate this all too well. The quest for doctrinal certainty through the definition of categories and subcategories is, in my opinion, destined to fail.
Moreover, the categorical approach does not take seriously the importance of context. The meaning of any expression and the legitimacy of its regulation can only be determined *427in context.5 Whether, for example, a picture or a sentence is obscene cannot be judged in the abstract, but rather only-in the context of its setting, its use, and its audience. Similarly, although legislatures may freely regulate most nonob-seene child pornography, such pornography that is part of “a serious work of art, a documentary on behavioral problems, or a medical or psychiatric teaching device” may be entitled to constitutional protection; the “question whether a specific act of communication is protected by the First Amendment always requires some consideration of both its content and its context.” Ferber, 458 U. S., at 778 (Stevens, J., concurring in judgment); see also Smith v. United States, 431 U. S. 291, 311-321 (1977) (Stevens, J., dissenting). The categorical approach sweeps too broadly when it declares that all such expression is beyond the protection of the First Amendment.
Perhaps sensing the limits of such an all-or-nothing approach, the Court has applied its analysis less categorically than its doctrinal statements suggest. The Court has recognized intermediate categories of speech (for example, for indecent nonobscene speech and commercial speech) and geographic categories of speech (public fora, limited public fora, nonpublic fora) entitled to varying levels of protection. The Court has also stringently delimited the categories of unprotected speech. While we once declared that “[ljibelous utterances [are] not... within the area of constitutionally protected speech,” Beauharnais v. Illinois, 343 U. S. 250, 266 (1952), our rulings in New York Times Co. v. Sullivan, 376 U. S. 254 (1964); Gertz v. Robert Welch, Inc., 418 U. S. 323 (1974), and Dun & Bradstreet, Inc. v. Greenmoss Builders, Inc., 472 U. S. 749 (1985), have substantially qualified this *428broad claim. Similarly, we have consistently construed the “fighting words” exception set forth in Chaplinsky narrowly. See, e. g., Houston v. Hill, 482 U. S. 451 (1987); Lewis v. New Orleans, 415 U. S. 130 (1974); Cohen v. California, 403 U. S. 15 (1971). In the case of commercial speech, our ruling that “the Constitution imposes no . . . restraint on government [regulation] as respects purely commercial advertising,” Valentine v. Chrestensen, 316 U. S. 52, 54 (1942), was expressly repudiated in Virginia Bd. of Pharmacy v. Virginia Citizens Consumer Council, Inc., 425 U. S. 748 (1976). In short, the history of the categorical approach is largely the history of narrowing the categories of unprotected speech.
This evolution, I believe, indicates that the categorical approach is unworkable and the quest for absolute categories of “protected” and “unprotected” speech ultimately futile. My analysis of the faults and limits of this approach persuades me that the categorical approach presented in Part I-A of Justice White’s opinion is not an adequate response to the novel “underbreadth” analysis the Court sets forth today.
Ill
As the foregoing suggests, I disagree with both the Court’s and part of Justice White’s analysis of the constitutionality of the St. Paul ordinance. Unlike the Court, I do not believe that all content-based regulations are equally infirm and presumptively invalid; unlike Justice White, I do not believe that fighting words are wholly unprotected by the First Amendment. To the contrary, I believe our decisions establish a more complex and subtle analysis, one that considers the content and context of the regulated speech, and the nature and scope of the restriction on speech. Applying this analysis and assuming, arguendo, (as the Court does) that the St. Paul ordinance is not overbroad, I conclude that such a selective, subject-matter regulation on proscribable speech is constitutional.
*429Not all content-based regulations are alike; our decisions clearly recognize that some content-based restrictions raise more constitutional questions than others. Although the Court’s analysis of content-based regulations cannot be reduced to a simple formula, we have considered a number of factors in determining the validity of such regulations.
First, as suggested above, the scope of protection provided expressive activity depends in part upon its content and character. We have long recognized that when government regulates political speech or “the expression of editorial opinion on matters of public importance,” FCC v. League of Women Voters of Cal., 468 U. S. 364, 375-376 (1984), “First Amendment protectio[n] is "at its zenith/ ” Meyer v. Grant, 486 U. S. 414, 425 (1988). In comparison, we have recognized that “commercial speech receives a limited form of First Amendment protection,” Posadas de Puerto Rico Associates v. Tourism Co. of Puerto Rico, 478 U. S. 328, 340 (1986), and that “society’s interest in protecting [sexually explicit films] is of a wholly different, and lesser, magnitude than [its] interest in untrammeled political debate,” Young v. American Mini Theatres, 427 U. S., at 70; see also FCC v. Pacifica Foundation, 438 U. S. 726 (1978). The character of expressive activity also weighs in our consideration of its constitutional status. As we have frequently noted, “[t]he government generally has a freer hand in restricting expressive conduct than it has in restricting the written or spoken word.” Texas v. Johnson, 491 U. S. 397, 406 (1989); see also United States v. O’Brien, 391 U. S. 367 (1968).
The protection afforded expression turns as well on the context of the regulated speech. We have noted, for example, that “[a]ny assessment of the precise scope of employer expression, of course, must be made in the context of its labor relations setting . .. [and] must take into account the economic dependence of the employees on their employers.” NLRB v. Gissel Packing Co., 395 U. S., at 617. Similarly, the distinctive character of a university environment, see *430Widmar v. Vincent, 454 U. S. 263, 277-280 (1981) (Stevens, J., concurring in judgment), or a secondary school environment, see Hazelwood School Dist. v. Kuhlmeier, 484 U. S. 260 (1988), influences our First Amendment analysis. The same is true of the presence of a “‘captive audience[, one] there as a matter of necessity, not of choice.’ ” Lehman v. Shaker Heights, 418 U. S., at 302 (citation omitted).6 Perhaps the most familiar embodiment of the relevance of context is our “fora” jurisprudence, differentiating the levels of protection afforded speech in different locations.
The nature of a contested restriction of speech also informs our evaluation of its constitutionality. Thus, for example, “[a]ny system of prior restraints of expression comes to this Court bearing a heavy presumption against its constitutional validity.” Bantam Books, Inc. v. Sullivan, 372 U. S. 58, 70 (1963). More particularly to the matter of content-based regulations, we have implicitly distinguished between restrictions on expression based on subject matter and restrictions based on viewpoint, indicating that the latter are particularly pernicious. “If there is a bedrock principle underlying the First Amendment, it is that the Government may not prohibit the expression of an idea simply because society finds the idea itself offensive or disagreeable.” Texas v. Johnson, 491 U. S., at 414. “Viewpoint discrimination is censorship in its purest form,” Perry Ed. Assn. v. Perry Local Educators’ Assn., 460 U. S. 37, 62 (1983) (Brennan, J., dissenting), and requires particular scrutiny, in part because such regulation often indicates a legislative effort to skew public debate on an issue, see, e. g., Schacht v. United States, 398 U. S. 58, 63 (1970). “Especially where... the legislature’s suppression of speech suggests an attempt *431to give one side of a debatable public question an advantage in expressing its views to the people, the First Amendment is plainly offended.” First Nat. Bank of Boston v. Bellotti, 435 U. S. 765, 785-786 (1978). Thus, although a regulation that on its face regulates speech by subject matter may in some instances effectively suppress particular viewpoints, see, e. g., Consolidated Edison Co. of N. Y. v. Public Serv. Comm’n of N. Y., 447 U. S. 530, 546-547 (1980) (Stevens, J., concurring in judgment), in general, viewpoint-based restrictions on expression require greater scrutiny than subject-matter-based restrictions.7
Finally, in considering the validity of content-based regulations we have also looked more broadly at the scope of the restrictions. For example, in Young v. American Mini Theatres, 427 U. S., at 71, we found significant the fact that “what [was] ultimately at stake [was] nothing more than a limitation on the place where adult films may be exhibited.” Similarly, in FCC v. Pacifica Foundation, the Court emphasized two dimensions of the limited scope of the FCC ruling. First, the ruling concerned only broadcast material which presents particular problems because it “confronts the citizen ... in the privacy of the home”; second, the ruling was not a complete ban on the use of selected offensive words, but rather merely a limitation on the times such speech could be broadcast. 438 U. S., at 748-750.
All of these factors play some role in our evaluation of content-based regulations on expression. Such a multifaceted analysis cannot be conflated into two dimensions. Whatever the allure of absolute doctrines, it is just too simple to declare expression “protected” or “unprotected” or to proclaim a regulation “content based” or “content neutral.”
*432In applying this analysis to the St. Paul ordinance, I assume, arguendo — as the Court does — that the ordinance regulates only fighting words and therefore is wot overbroad. Looking to the content and character of the regulated activity, two things are clear. First, by hypothesis the ordinance bars only low-value speech, namely, fighting words. By definition such expression constitutes “no essential part of any exposition of ideas, and [is] of such slight social value as a step to truth that any benefit that may be derived from [it] is clearly outweighed by the social interest in order and morality.” Chaplinsky, 315 U. S., at 572. Second, the ordinance regulates “expressive conduct [rather] than ... the written or spoken word.” Texas v. Johnson, 491 U. S., at 406.
Looking to the context of the regulated activity, it is again significant that the ordinance (by hypothesis) regulates only fighting words. Whether words are fighting words is determined in part by their context. Fighting words are not words that merely cause offense; fighting words must be directed at individuals so as to “by their very utterance inflict injury.” By hypothesis, then, the St. Paul ordinance restricts speech in confrontational and potentially violent situations. The case at hand is illustrative. The cross burning in this case — directed as it was to a single African-American family trapped in their home — was nothing more than a crude form of physical intimidation. That this cross burning sends a message of racial hostility does not automatically endow it with complete constitutional protection.8
*433Significantly, the St. Paul ordinance regulates speech not on the basis of its subject matter or the viewpoint expressed, but rather on the basis of the harm the speech causes. In this regard, the Court fundamentally misreads the St. Paul ordinance. The Court describes the St. Paul ordinance as regulating expression “addressed to one of [several] specified disfavored topics,” ante, at 391 (emphasis supplied), as policing “disfavored subjects,” ibid, (emphasis supplied), and as “prohibiting].. . speech solely on the basis of the subjects the speech addresses,” ante, at 381 (emphasis supplied). Contrary to the Court’s suggestion, the ordinance regulates only a subcategory of expression that causes injuries based on “race, color, creed, religion or gender,” not a subcategory that involves discussions that concern those characteristics.9 The ordinance,- as construed by the Court, criminalizes expression that “one knows ... [by its very utterance inflicts injury on] others on the basis of race, color, creed, religion or *434gender.” In this regard, the ordinance resembles the child pornography law at issue in Ferber,. which in effect singled out child pornography because those publications caused far greater harms than pornography involving adults.
Moreover, even if the St. Paul ordinance did regulate fighting words based on its subject matter, such a regulation would, in my opinion, be constitutional. As noted above, subject-matter-based regulations on commercial speech are widespread and largely unproblematic. As we have long recognized, subject-matter regulations generally do not raise the same concerns of government censorship and the distortion of public discourse presented by viewpoint regulations. Thus, in upholding subject-matter regulations we have carefully noted that viewpoint-based discrimination was not implicated. See Young v. American Mini Theatres, 427 U. S., at 67 (emphasizing “the need for absolute neutrality by the government,” and observing that the contested statute was not animated by “hostility for the point of view" of the theaters); FCC v. Pacifica Foundation, 438 U. S., at 745-746 (stressing that “government must remain neutral in the marketplace of ideas”); see also FCC v. League of Women’s Voters of Cal., 468 U. S., at 412-417 (Stevens, J., dissenting); Metromedia, Inc. v. San Diego, 453 U. S. 490, 554-555 (1981) (Stevens, J., dissenting in part). Indeed, some subject-matter restrictions are a functional necessity in contemporary governance: “The First Amendment does not require States to regulate for problems that do not exist.” Burson v. Freeman, 504 U. S., at 207.
Contrary to the suggestion of the majority, the St. Paul ordinance does not regulate expression based on viewpoint. The Court contends that the ordinance requires proponents of racial intolerance to “follow the Marquis of Queensberry rules” while allowing advocates of racial tolerance to “fight freestyle.” The law does no such thing.
*435The Court writes:
“One could hold up a sign saying, for example, that all ‘anti-Catholic bigots’ are misbegotten; but not that all ‘papists’ are, for that would insult and provoke violence ‘on the basis of religion.’ ” Ante, at 391-392.
This may be true, but it hardly proves the Court’s point. The Court’s reasoning is asymmetrical. The response to a sign saying that “all [religious] bigots are misbegotten” is a sign saying that “all advocates of religious tolerance are misbegotten.” Assuming such signs could be fighting words (which seems to me extremely unlikely), neither sign would be banned by the ordinance for the attacks were not “based on . . . religion” but rather on one’s beliefs about tolerance. Conversely (and again assuming such signs are fighting words), just as the ordinance would prohibit a Muslim from hoisting a sign claiming that all Catholics were misbegotten, so the ordinance would bar a Catholic from hoisting a similar sign attacking Muslims.
The St. Paul ordinance is evenhanded. In a battle between advocates of tolerance and advocates of intolerance, the ordinance does not prevént either side from hurling fighting words at the other on the basis of their conflicting ideas, but it does bar both sides from hurling such words on the basis of the target’s “race, color, creed, religion or gender.” To extend the Court’s pugilistic metaphor, the St. Paul ordinance simply bans punches “below the belt”— by either party. It does not, therefore, favor one side of any debate.10
*436Finally, it is noteworthy that the St. Paul ordinance is, as construed by the Court today, quite narrow. The St. Paul ordinance does not ban all “hate speech,” nor does it ban, say, all cross burnings or all swastika displays. Rather it only bans a subcategory of the already narrow category of fighting words. Such a limited ordinance leaves open and protected a vast range of expression on the subjects of racial, religious, and gender equality. As construed by the Court today, the ordinance certainly does not “ 'rais[e] the specter that the Government may effectively drive certain ideas or viewpoints from the marketplace.’” Ante, at 387. Petitioner is free to burn a cross to announce a rally or to express his views about racial supremacy, he may do so on private property or public land, at day or at night, so long as the burning is not so threatening and so directed at an individual as to “by its very [execution] inflict injury.” Such a limited proscription scarcely offends the First Amendment.
In sum, the St. Paul ordinance (as construed by the Court) regulates expressive activity that is wholly proseribable and does so not on the basis of viewpoint, but rather in recognition of the different harms caused by such activity. Taken together, these several considerations persuade me that the St. Paul ordinance is not an unconstitutional content-based regulation of speech. Thus, were the ordinance not over-broad, I would vote to uphold it.
The Court disputes this characterization because it has crafted two exceptions, one for “certain media or markets” and the other for content discrimination based upon “the very reason that the entire class of speech at issue is proscribable.” Ante, at 388. These exceptions are, at best, ill defined. The Court does not tell us whether, with respect to the former, fighting words such as cross burning could be proscribed only in certain neighborhoods where the threat of violence is particularly severe, or whether, with respect to the second category, fighting words that create a particular risk of harm (such as a race riot) would be proscribable. The hypothetical and illusory category of these two exceptions persuades me that either my description of the Court’s analysis is accurate or that the Court does not in fact mean much of what it says in its opinion.
See Mine Workers v. Pennington, 381 U. S. 657 (1965); Eastern Railroad Presidents Conference v. Noerr Motor Freight, Inc., 365 U. S. 127 (1961).
See also Packer Corp. v. Utah, 285 U. S. 105 (1932) (Brandéis, J.) (upholding a statute that prohibited the advertisement of cigarettes on billboards and streetcar placards).
The Court states that the prohibition on content-based regulations “applies differently in the context of proseribable speech” than in the context of other speech, ante, at 387, but its analysis belies that claim. The Court strikes down the St. Paul ordinance because it regulates fighting words based on subject matter, despite the fact that, as demonstrated above, we have long upheld regulations of commercial speech based on subject matter. The Court’s self-description is inapt: By prohibiting the regulation of fighting words based on its subject matter, the Court provides the same protection to fighting words as is currently provided to core political speech.
“A word,” as Justice Holmes has noted, “is not a crystal, transparent and unchanged, it is the skin of a living thought and may vary greatly in color and content according to the circumstances and the time in which it is used.” Towns v. Eisner, 245 U. S. 418, 425 (1918); see also Jacobellis v. Ohio, 378 U. S. 184, 201 (1964) (Warren, C. J., dissenting).
Cf. In re Chase, 468 F. 2d 128, 139-140 (CA71972) (Stevens, J., dissenting) (arguing that defendant who, for reasons of religious belief, refused to rise and stand as the trial judge entered the courtroom was not subject to contempt proceedings because he was not present in the courtroom “as a matter of choice”).
Although the Court has sometimes suggested that subject-matter-based and viewpoint-based regulations are equally problematic, see, e. g., Consolidated Edison Co. of N. Y. v. Public Serv. Comm’n of N. Y., 447 U. S., at 537, our decisions belie such claims.
The Court makes much of St. Paul’s description of the ordinance as regulating “a message.” Ante, at 393. As always, however, St. Paul’s argument must be read in context:
“Finally, we ask the Court to reflect on the‘content’ of the ‘expressive conduct’ represented by a ‘burning cross.’ It is no less than the first step in an act of racial violence. It was and unfortunately still is the equivalent of [the] waving of a knife before the thrust, the pointing of a gun before it is fired, the lighting of the match before the arson, the hanging of the noose before the lynching. It is not a political statement, or even *433a cowardly statement of hatred. It is the first step in an act of assault. It can be no more protected than holding a gun to a victimas] head. It is perhaps the ultimate expression of ‘fighting words.’ ” App. to Brief for Petitioner C-6.
The Court contends that this distinction is “wordplay,” reasoning that “[w]hat makes [the harms caused by race-based threats] distinct from [the harms] produced by other fighting words is ... the fact that [the former are] caused by a distinctive idea.” Ante, at 392-393 (emphasis added). In this way, the Court concludes that regulating speech based on the injury it causes is no different from regulating speech based on its subject matter. This analysis fundamentally miscomprehends the role of “race, color, creed, religion [and] gender” in contemporary American society. One need look no further than the recent social unrest in the Nation’s cities to see that race-based threats may cause more harm to society and to individuals than other threats. Just as the statute prohibiting threats against the President is justifiable because of the place of the President in our social and political order, so a statute prohibiting race-based threats is justifiable because of the place of race in our social and political order. Although it is regrettable that race occupies such a place and is so incendiary an issue, until the Nation matures beyond that condition, laws such as St. Paul’s ordinance will remain reasonable and justifiable.
Cf. FCC v. League of Women Voters of Cal., 468 U. S. 364, 418 (1984) (Stevens, J., dissenting) (“In this case... the regulation applies ... to a defined class of... licensees [who] represent heterogenous points of view. There is simply no sensible basis for considering this regulation a viewpoint restriction — or ... to condemn it as ‘content-based’ — -because it applies equally to station owners of all shades of opinion”).