concurring in the judgment.
I agree with the majority that the judgment of the Minnesota Supreme Court should be reversed. However, our agreement ends there.
This ease could easily be decided within the contours of established First Amendment law by holding, as petitioner argues, that the St. Paul ordinance is fatally overbroad because it criminalizes not only unprotected expression but expression protected by the First Amendment. See Part II, infra. Instead, “find[ing] it unnecessary” to consider the questions upon which we granted review,1 ante, at 381, the *398Court holds the ordinance facially unconstitutional on a ground that was never presented to the Minnesota Supreme Court, a ground that has not been briefed by the parties before this Court, a ground that requires serious departures from the teaching of prior cases and is inconsistent with the plurality opinion in Burson v. Freeman, 504 U. S. 191 (1992), which was joined by two of the five Justices in the majority in the present case.
This Court ordinarily is not so eager to abandon its precedents. Twice within the past month, the Court has declined to overturn longstanding but controversial decisions on questions of constitutional law. See Allied-Signal, Inc. v. Director, Division of Taxation, 504 U. S. 768 (1992); Quill Corp. v. North Dakota, 504 U. S. 298 (1992). In each case, we had the benefit of full briefing on the critical issue, so that the parties and amici had the opportunity to apprise us of the impact of a change in the law. And in each case, the Court declined to abandon its precedents, invoking the principle of stare decisis. Allied-Signal, Inc., supra, at 783-786; Quill Corp., supra, at 317-318.
But in the present ease, the majority casts aside long-established First Amendment doctrine without the benefit of briefing and adopts an untried theory. This is hardly a judicious way of proceeding, and the Court’s reasoning in reaching its result is transparently wrong.
*399I
A
This Court’s decisions have plainly stated that expression falling within certain limited categories so lacks the values the First Amendment was designed to protect that the Constitution affords no protection to that expression. Chaplinsky v. New Hampshire, 315 U. S. 568 (1942), made the point in the clearest possible terms:
“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.” Id., at 571-572.
See also Bose Corp. v. Consumers Union of United States, Inc., 466 U. S. 485, 504 (1984) (citing Chaplinsky).
Thus, as the majority concedes, see ante, at 383-384, this Court has long held certain discrete categories of expression to be proscribable on the basis of their content. For instance, the Court has held that the individual who falsely shouts “fire” in a crowded theater may not claim the protection of the First Amendment. Schenck v. United States, 249 U. S. 47, 52 (1919). The Court has concluded that neither child pornography nor obscenity is protected by the First Amendment. New York v. Ferber, 458 U. S. 747, 764 (1982); Miller v. California, 413 U. S. 15, 20 (1973); Roth v. United States, 354 U. S. 476, 484-485 (1957). And the Court has observed that, “[ljeaving aside the special considerations when public officials [and public figures] are the target, a libelous publication is not protected by the Constitution.” Ferber, supra, at 763 (citations omitted).
*400All of these categories are content based. But the Court has held that the First Amendment does not apply to them because their expressive content is worthless or of de mini-mis value to society. Chaplinsky, supra, at 571-572. We have not departed from this principle, emphasizing repeatedly that, “within the confines of [these] given classification[s], the evil to be restricted so overwhelmingly outweighs the expressive interests, if any, at stake, that no process of case-by-case adjudication is required.” Ferber, supra, at 763-764; Bigelow v. Virginia, 421 U. S. 809, 819 (1975). This categorical approach has provided a principled and narrowly focused means for distinguishing between expression that the government may regulate freely and that which it may regulate on the basis of content only upon a showing of compelling need.2
Today, however, the Court announces that earlier Courts did not mean their repeated statements that certain categories of expression are “not within the area of constitutionally protected speech.” Roth, supra, at 483. See ante, at 383, citing Beauharnais v. Illinois, 343 U. S. 250, 266 (1952); Chaplinsky, supra, at 571-572; Bose Corp., supra, at 504; Sable Communications of Cal., Inc. v. FCC, 492 U. S. 115, 124 (1989). The present Court submits that such clear statements “must be taken in context” and are not “literally true.” Ante, at 383.
To the contrary, those statements meant precisely what they said: The categorical approach is a firmly entrenched part of our First Amendment jurisprudence. Indeed, the Court in Roth reviewed the guarantees of freedom of expression in effect at the time of the ratification of the Constitution and concluded, “In light of this history, it is apparent that the unconditional phrasing of the First Amendment was *401not intended to protect every utterance.” 354 U. S., at 482-483.
In its decision today, the Court points to “[njothing ... in this Court’s precedents warranting] disregard of this longstanding tradition.” Burson, 504 U. S., at 216 (Scalia, J., concurring in judgment); Allied-Signal, Inc., supra, at 783. Nevertheless, the majority holds that the First Amendment protects those narrow categories of expression long held to be undeserving of First Amendment protection — at least to the extent that lawmakers may not regulate some fighting words more strictly than others because of their content. The Court announces that such content-based distinctions violate the First Amendment because “[t]he government may not regulate use based on hostility — or favoritism — towards the underlying message expressed.” Ante, at 386. Should the government want to criminalize certain fighting words, the Court now requires it to criminalize all fighting words.
To borrow a phrase: “Such a simplistic, all-or-nothing-at-all approach to First Amendment protection'is at odds with common sense and with our jurisprudence as well.” Ante, at 384. It is inconsistent to hold that the government may proscribe an entire category of speech because the content of that speech is evil, Ferber, supra, at 763-764; but that the government may not treat a subset of that category differently without violating the First Amendment; the content of the subset is by definition worthless and undeserving of constitutional protection.
The majority’s observation that fighting words are “quite expressive indeed,” ante, at 385, is no answer. Fighting words are not a means of exchanging views, rallying supporters, or registering a protest; they are directed against individuals to provoke violence or to inflict injury. Chaplinsky, 315 U. S., at 572. Therefore, a ban on all fighting words or on a subset of the fighting words category would restrict only the social evil of hate speech, without creating the danger of driving viewpoints from the marketplace. See ante, at 387.
*402Therefore, the Court’s insistence on inventing its brand of First Amendment underinclusiveness puzzles me.3 The overbreadth doctrine has the redeeming virtue of attempting to avoid the chilling of protected expression, Broadrick v. Oklahoma, 413 U. S. 601, 612 (1973); Osborne v. Ohio, 495 U. S. 103,112, n. 8 (1990); Brockett v. Spokane Arcades, Inc., 472 U. S. 491, 503 (1985); Ferber, supra, at 772, but the Court’s new “underbreadth” creation serves no desirable function. Instead, it permits, indeed invites, the continuation of expressive conduct that in this case is evil and worthless in First Amendment terms, see Ferber, supra, at 763-764; Chaplinsky, supra, at 571-572, until the city of St. Paul cures the underbreadth by adding to its ordinance a catchall phrase such as “and all other fighting words that may constitutionally be subject to this ordinance.”
Any contribution of this holding to First Amendment jurisprudence is surely a negative one, since it necessarily signals that expressions of violence, such as the message of intimidation and racial hatred conveyed by burning a cross on someone’s lawn, are of sufficient value to outweigh the social interest in order and morality that has traditionally placed such fighting words outside the First Amendment.4 Indeed, by characterizing fighting words as a form of “debate,” ante, at 392, the majority legitimates hate speech as a form of public discussion.
*403Furthermore, the Court obscures the line between speech that could be regulated freely on the basis of content (i. e., the narrow categories of expression falling outside the First Amendment) and that which could be regulated on the basis of content only upon a showing of a compelling state interest (i e., all remaining expression). By placing fighting words, which the Court has long held to be valueless, on at least equal constitutional footing with political discourse and other forms of speech that we have deemed to have the greatest social value, the majority devalues the latter category. See Burson v. Freeman, supra, at 196; Eu v. San Francisco Cty. Democratic Central Comm., 489 U. S. 214, 222-223 (1989).
B
In a second break with precedent, the Court refuses to sustain the ordinance even though it would survive under the strict scrutiny applicable to other protected expression. Assuming, arguendo, that the St. Paul ordinance is a content-based regulation of protected expression, it nevertheless would pass First Amendment review under settled law upon a showing that the regulation “'is necessary to serve a compelling state interest and is narrowly drawn to achieve that end/ ” Simon & Schuster, Inc. v. Members of N. Y. State Crime Victims Bd., 502 U. S. 105, 118 (1991) (quoting Arkansas Writers’ Project, Inc. v. Ragland, 481 U. S. 221, 231 (1987)). St. Paul has urged that its ordinance, in the words of the majority, “helps to ensure the basic human rights of members of groups that have historically been subjected to discrimination . . . .” Ante, at 395. The Court expressly concedes that this interest is compelling and is promoted by the ordinance. Ibid. Nevertheless, the Court treats strict scrutiny analysis as irrelevant to the constitutionality of the legislation:
“The dispositive question ... is whether content discrimination is reasonably necessary to achieve St. Paul’s compelling interests; it plainly is not. An ordinance not *404limited to the favored topics, for example, would have precisely the same beneficial effect.” Ante, at 395-396.
Under the majority’s view, a narrowly drawn, content-based ordinance could never pass constitutional muster if the object of that legislation could be accomplished by banning a wider category of speech. This appears to be a general renunciation of strict scrutiny review, a fundamental tool of First Amendment analysis.5
This abandonment of the doctrine is inexplicable in light of our decision in Burson v. Freeman, 504 U. S. 191 (1992), which was handed down just a month ago.6 In Burson, seven of the eight participating Members of the Court agreed that the strict scrutiny standard applied in a ease involving a First Amendment challenge to a content-based statute. See id., at 198 (plurality opinion); id., at 217 (Stevens, J., *405dissenting).7 The statute at issue prohibited the solicitation of Votes and the display or distribution of campaign materials within 100 feet of the entrance to a polling place. The plurality concluded that the legislation survived strict scrutiny because the State had asserted a compelling interest in regulating electioneering near polling places and because the statute at issue was narrowly tailored to accomplish that goal. Id., at 208-210.
Significantly, the statute in Burson did not proscribe all speech near polling places; it restricted only political speech. Id., at 197. The Burson plurality, which included The ChieF Justice and Justice Kennedy, concluded that the distinction between types of speech required application of strict scrutiny, but it squarely rejected the proposition that the legislation failed First Amendment review because it could have been drafted in broader, content-neutral terms:
“States adopt laws to address the problems that confront them. The First Amendment does not require States to regulate for problems that do not exist .” Id., at 207 (emphasis added).
This reasoning is in direct conflict with the majority’s analysis in the present case, which leaves two options to lawmakers attempting to regulate expressions of violence: (1) enact a sweeping prohibition on an entire class of speech (thereby requiring “regulation] for problems that do not exist”); or (2) not legislate at all.
Had the analysis adopted by the majority in the present case been applied in Burson, the challenged election law would have failed constitutional review, for its content-based distinction between political and nonpolitical speech could not have been characterized as “reasonably necessary,” ante, *406at 395, to achieve the State’s interest in regulating polling place premises.8
As with its rejection of the Court’s categorical analysis, the majority offers no reasoned basis for discarding our firmly established strict scrutiny analysis at this time. The majority appears to believe that its doctrinal revisionism is necessary to prevent our elected lawmakers from prohibiting libel against members of one political party but not another and from enacting similarly preposterous laws. Ante, at 884. The majority is misguided.
Although the First Amendment does not apply to categories of unprotected speech, such as fighting words, the Equal Protection Clause requires that the regulation of unprotected speech be rationally related to a legitimate government interest. A defamation statute that drew distinctions on the basis of political affiliation or “an ordinance prohibiting only those legally obscene works that contain criticism of the city government,” ibid., would unquestionably fail rational-basis review.9
*407Turning to the St. Paul ordinance and assuming, arguendo, as the majority does, that the ordinance is not constitutionally overbroad (but see Part II, infra), there is no question that it would pass equal protection review. The ordinance proscribes a subset of “fighting words,” those that injure “on the basis of race, color, creed, religion or gender.” This selective regulation reflects the city’s judgment that harms based on race, color, creed, religion, or gender are more pressing public concerns than the harms caused by other fighting words. In light of our Nation’s long and painful experience with discrimination, this determination is plainly reasonable. Indeed, as the majority concedes, the interest is compelling. Ante, at 395.
C
The Court has patched up its argument with an apparently nonexhaustive list of ad hoc exceptions, in what can be viewed either as an attempt to confine the effects of its decision to the facts of this case, see post, at 415 (Blackmun, J., concurring in judgment), or as an effort to anticipate some of the questions that will arise from its radical revision of First Amendment law.
For instance, if the majority were to give general application to the rule on which it decides this case, today’s decision would call into question the constitutionality of the statute making it illegal to threaten the life of the President. 18 U.S.C. §871. See Watts v. United States, 394 U.S. 705 (1969) (per curiam). Surely, this statute, by singling out certain threats, incorporates a content-based distinction; it indicates that the Government especially disfavors threats against the President as opposed to threats against all oth*408ers.10 See ante, at 391. But because the Government could prohibit all threats and not just those directed against the President, under the Court’s theory, the compelling reasons justifying the enactment of special legislation to safeguard the President would be irrelevant, and the statute would fail First Amendment review.
To save the statute, the majority has engrafted the following exception onto its newly announced First Amendment rule: Content-based distinctions may be drawn within an unprotected category of speech if the basis for the distinctions is “the very reason the entire class of speech at issue is proscribable.” Ante, at 388. Thus, the argument goes, the statute making it illegal to threaten the life of the President is constitutional, “since the reasons why threats 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 person of the President.” Ibid.
The exception swallows the majority’s rule. Certainly, it should apply to the St. Paul ordinance, since “the reasons why [fighting words] are outside the First Amendment. . . have special force when applied to [groups that have historically been subjected to discrimination].”
To avoid the result of its own analysis, the Court suggests that fighting words are simply a mode of communication, rather than a content-based category, and that the St. Paul ordinance has not singled out a particularly objectionable mode of communication. Ante, at 386,393. Again, the majority confuses the issue. A prohibition on fighting words is not a time, place, or manner restriction; it is a ban on a class of speech that conveys an overriding message of personal injury and imminent violence, Chaplinsky, 315 U. S., at 572, a message that is at its ugliest when directed against groups *409that have long been the targets of discrimination. Accordingly, the ordinance falls within the first exception to the majority’s theory.
As its second exception, the Court posits that certain content-based regulations will survive under the new regime if the regulated subclass “happens to be associated with particular ‘secondary effects’ of the speech . . .,” ante, at 389, which the majority treats as encompassing instances in which “words can... violate laws directed not against speech but against conduct. . . ,” ibid.11 Again, there is a simple explanation for the Court’s eagerness to craft an exception to its new First Amendment rule: Under the general rule the Court applies in this case, Title VII hostile work environment claims would suddenly be unconstitutional.
Title VII of the Civil Rights Act of 1964 makes it unlawful to discriminate “because of [an] individual’s race, color, religion, sex, or national origin,” 42 U. S. C. § 2000e-2(a)(l), and the regulations covering hostile workplace claims forbid “sexual harassment,” which includes “[u]nwelcome sexual advances, requests for sexual favors, and other verbal or physical conduct of a sexual nature” that create “an intimidating, hostile, or offensive working environment,” 29 CFR § 1604.11(a) (1991). The regulation does not prohibit workplace harassment generally; it focuses on what the majority would characterize as the “disfavored topi[e]” of sexual harassment. Ante, at 391. In this way, Title VII is similar to the St. Paul ordinance that the majority condemns because it “impose[sJ special prohibitions on those speakers who express views on disfavored subjects.” Ibid. Under the broad principle the Court uses to decide the present ease, *410hostile work environment claims based on sexual harassment should fail First Amendment review; because a general ban on harassment in the workplace would cover the problem of sexual harassment, any attempt to proscribe the subcategory of sexually harassing expression would violate the First Amendment.
Hence, the majority’s second exception, which the Court indicates would insulate a Title VII hostile work environment claim from an underinclusiveness challenge because “sexually derogatory ‘fighting words’... may produce a violation of Title VII’s general prohibition against sexual discrimination in employment practices.” Ante, at 389. But application of this exception to a hostile work environment claim does not hold up under close examination.
First, the hostile work environment regulation is not keyed to the presence or absence of an economic quid pro quo, Meritor Savings Bank, F. S. B. v. Vinson, 477 U. S. 57, 65 (1986), but to the impact of the speech on the victimized worker. Consequently, the regulation would no more fall within a secondary effects exception than does the St. Paul ordinance. Ante, at 394. Second, the majority’s focus on the statute’s general prohibition on discrimination glosses over the language of the specific regulation governing hostile working environment, which reaches beyond any “incidental” effect on speech. United States v. O’Brien, 391 U. S. 367, 376 (1968). If the relationship between the broader statute and specific regulation is sufficent to bring the Title VII regulation within O’Brien, then all St. Paul need do to bring its ordinance within this exception is to add some prefatory language concerning discrimination generally.
As to the third exception to the Court’s theory for deciding this case, the majority concocts a catchall exclusion to protect against unforeseen problems, a concern that is heightened here given the lack of briefing on the majority’s deci-sional theory. This final exception would apply in cases in which “there is no realistic possibility that official suppression of ideas is afoot.” Ante, at 390. As I have demon*411strated, this case does not concern the official suppression of ideas. See supra, at 401. The majority discards this notion out of hand. Ante, at 395.
As I see it, the Court’s theory does not work and will do nothing more than confuse the law. Its selection of this ease to rewrite First Amendment law is particularly inexplicable, because the whole problem could have been avoided by deciding this case under settled First Amendment principles.
HH ] — I
Although I disagree with the Court s analysis, I do agree with its conclusion: The St. Paul ordinance is unconstitutional. However, I would decide the case on overbreadth grounds.
We have emphasized time and again that overbreadth doctrine is an exception to the established principle that “a person to whom a statute may constitutionally be applied will not be heard to challenge that statute on the ground that it may conceivably be applied unconstitutionally to others, in other situations not before the Court.” Broadrick v. Oklahoma, 413 U. S., at 610; Brockett v. Spokane Arcades, Inc., 472 U. S., at 503-504. A defendant being prosecuted for speech or expressive conduct may challenge the law on its face if it reaches protected expression, even when that person’s activities are not protected by the First Amendment. This is because “the possible harm to society in permitting some unprotected speech to go unpunished is outweighed by the possibility that protected speech of others may be muted.” Broadrick, supra, at 612; Osborne v. Ohio, 495 U. S., at 112, n. 8; New York v. Ferber, 458 U. S., at 768-769; Schaumburg v. Citizens for a Better Environment, 444 U. S. 620, 634 (1980); Gooding v. Wilson, 405 U. S. 518, 521 (1972).
However, we have consistently held that, because over-breadth analysis is “strong medicine,” it may be invoked to strike an entire statute only when the overbreadth of the statute is not only “real, but substantial as well, judged in relation to the statute’s plainly legitimate sweep,” Broad*412rick, 413 U. S., at 615, and when the statute is not susceptible to limitation or partial invalidation, id., at 613; Board of Airport Comm’rs of Los Angeles v. Jews for Jesus, Inc., 482 U. S. 569, 574 (1987). “When a federal court is dealing with a federal statute challenged as overbroad, it should ... construe the statute to avoid constitutional problems, if the statute is subject to a limiting construction.” Ferber, 458 U. S., at 769, n. 24. Of course, “[a] state court is also free to deal with a state statute in the same way.” Ibid. See, e.g., Osborne, 495 U. S., at 113-114.
Petitioner contends that the St. Paul ordinance is not susceptible to a narrowing construction and that the ordinance therefore should be considered as written, and not as construed by the Minnesota Supreme Court. Petitioner is wrong. Where a state court has interpreted a provision of state law, we cannot ignore that interpretation, even if it is not one that we would have reached if we were construing the statute in the first instance. Ibid.; Kolender v. Lawson, 461 U. S. 352, 355 (1983); Hoffman Estates v. Flipside, Hoffman Estates, Inc., 455 U. S. 489, 494, n. 5 (1982).12
Of course, the mere presence of a state court interpretation does not insulate a statute from overbreadth review. We have stricken legislation when the construction supplied by the state court failed to cure the overbreadth problem. *413See, e. g., Lewis v. New Orleans, 415 U. S. 130, 132-133 (1974); Gooding, supra, at 524-525. But in such cases, we have looked to the statute as construed in determining whether it contravened the First Amendment. Here, the Minnesota Supreme Court has provided an authoritative construction of the St. Paul antibias ordinance. Consideration of petitioner’s overbreadth claim must be based on that interpretation.
I agree with petitioner that the ordinance is invalid on its face. Although the ordinance as construed reaches categories of speech that are constitutionally unprotected, it also criminalizes a substantial amount of expression that — however repugnant — is shielded by the First Amendment.
In attempting to narrow the scope of the St. Paul antibias ordinance, the Minnesota Supreme Court relied upon two of the categories of speech and expressive conduct that fall outside the First Amendment’s protective sphere: words that incite “imminent lawless action,” Brandenburg v. Ohio, 395 U. S. 444, 449 (1969), and “fighting” words, Chaplinsky v. New Hampshire, 315 U. S., at 571-572. The Minnesota Supreme Court erred in its application of the Chaplinsky fighting words test and consequently interpreted the St. Paul ordinance in a fashion that rendered the ordinance facially overbroad.
In construing the St. Paul ordinance, the Minnesota Supreme Court drew upon the definition of fighting words that appears in Chaplinsky — words “which by their very utterance inflict injury or tend to incite an immediate breach of the peace.” Id., at 572. However, the Minnesota court was far from clear in identifying the “injúr[ies]” inflicted by the expression that St. Paul sought to regulate. Indeed, the Minnesota court emphasized (tracking the language of the ordinance) that “the ordinance censors only those displays that one knows or should know will create anger, alarm or resentment based on racial, ethnic, gender or religious bias.” In re Welfare of R. A. V., 464 N. W. 2d 507, 510 (1991). I *414therefore understand the court to have ruled that St. Paul may constitutionally prohibit expression that “by its very utterance” causes “anger, alarm or resentment.”
Our fighting words cases have made clear, however, that such generalized reactions are not sufficient to strip expression of its constitutional protection. The mere fact that expressive activity causes hurt feelings, offense, or resentment does not render the expression unprotected. See United States v. Eichman, 496 U. S. 310, 319 (1990); Texas v. Johnson, 491 U. S. 397, 409, 414 (1989); Hustler Magazine, Inc. v. Falwell, 485 U. S. 46, 55-56 (1988); FCC v. Pacifica Foundation, 438 U. S. 726, 745 (1978); Hess v. Indiana, 414 U. S. 105, 107-108 (1973); Cohen v. California, 403 U. S. 15, 20 (1971); Street v. New York, 394 U. S. 576, 592 (1969); Terminiello v. Chicago, 337 U. S. 1 (1949).
In the First Amendment context, “[e]rrminal statutes must be scrutinized with particular care; those that make unlawful a substantial amount of constitutionally protected conduct may be held facially invalid even if they also have legitimate application.” Houston v. Hill, 482 U. S. 451,459 (1987) (citation omitted). The St. Paul antibias ordinance is such a law. Although the ordinance reaches conduct that is unprotected, it also makes criminal expressive conduct that causes only hurt feelings, offense, or resentment, and is protected by the First Amendment. Cf. Lewis, supra, at 132.13 The ordinance is therefore fatally overbroad and invalid on its face.
*415HH hH
Today, the Court has disregarded two established principles of First Amendment law without providing a coherent replacement theory. Its decision is an arid, doctrinaire interpretation, driven by the frequently irresistible impulse of judges to tinker with the First Amendment. The decision is mischievous at best and will surely confuse the lower courts. I join the judgment, but not the folly of the opinion.
The Court granted certiorari to review the following questions:
“1. May a local government enact a content-based, 'hate-crime’ ordinance prohibiting the display of symbols, including a Nazi swastika or a burning cross, on public or private property, which one knows or has reason to know arouses anger, alarm, or resentment in others on the basis of race, color, creed, religion, or gender without violating overbreadth and vagueness principles of the First Amendment to the United States Constitution?
“2. Can the constitutionality of such a vague and substantially over-broad content-based restraint of expression be saved by a limiting construction, like that used to save the vague and overbroad content-neutral laws, restricting its application to 'fighting words’ or ‘imminent lawless action?’” Pet. for Cert. i.
It has long been the rule of this Court that “[o]nly the questions set forth in the petition, or fairly included therein, will be considered by the Court.” This Court's Rule 14.1(a). This Rule has served to focus the issues presented for review. But the majority reads the Rule so expansively that any First Amendment theory would appear to be “fairly included” within the questions quoted above.
Contrary to the impression the majority attempts to create through its selective quotation of petitioner’s briefs, see ante, at 381-382, n. 3, petitioner did not present to this Court or the Minnesota Supreme Court anything approximating the novel theory the majority adopts today. Most certainly petitioner did not “reiterat[e]” such a claim at argument; he responded to a question from the bench, Tr. of Oral Arg. 8. Previously, this Court has shown the restraint to refrain from deciding cases on the basis *398of its own theories when they have not been pressed or passed upon by a state court of last resort. See, e. g., Illinois v. Gates, 462 U. S. 213, 217-224 (1983).
Given this threshold issue, it is my view that the Court lacks jurisdiction to decide the case on the majority rationale. Cf. Arkansas Electric Cooperative Corp. v. Arkansas Pub. Serv. Comm’n, 461 U. S. 375, 382, n. 6 (1983). Certainly the preliminary jurisdictional and prudential concerns are sufficiently weighty that we would never have granted certiorari had petitioner sought review of a question based on the majority’s decisional theory.
“In each of these areas, the limits of the unprotected category, as well as the unprotected character of particular communications, have been determined by the judicial evaluation of special facts that have been deemed to have constitutional significance.” Bose Corp. v. Consumers Union of United States, Inc., 466 U. S. 485, 504-505 (1984).
The assortment of exceptions the Court attaches to its rule belies the majority’s claim, see ante, at 387, that its new theory is truly concerned with content discrimination. See Part I-C, infra (discussing the exceptions).
This does not suggest, of course, that cross burning is always unprotected. Burning a cross at a political rally would almost certainly be protected expression. Cf. Brandenburg v. Ohio, 395 U. S. 444, 445 (1969). But in such a context, the cross burning could not be characterized as a “direct personal insult or an invitation to exchange fisticuffs,” Texas v. Johnson, 491 U. S. 397, 409 (1989), to which the fighting words doctrine, see Part II, infra, applies.
The majority relies on Boos v. Barry, 485 U. S. 312 (1988), in arguing that the availability of content-neutral alternatives “‘undercuts] significantly’” a claim that content-based legislation is ‘“necessary to serve the asserted [compelling] interest.’” Ante, at 395 (quoting Boos, supra, at 329, and Burson v. Freeman, 504 U. S. 191, 199 (1992) (plurality opinion)). Boos does not support the majority’s analysis. In Boos, Congress already had decided that the challenged legislation was not necessary, and the Court pointedly deferred to this choice. 485 U. S., at 329. St. Paul lawmakers have made no such legislative choice.
Moreover, in. Boos, the Court held that the challenged statute was not narrowly tailored because a less restrictive alternative was available. Ibid. But the Court’s analysis today turns Boos inside-out by substituting the majority’s policy judgment that a more restrictive alternative could adequately serve the compelling need identified by St. Paul lawmakers. The result would be: (a) a statute that was not tailored to fit the need identified by the government; and (b) a greater restriction on fighting words, even though the Court clearly believes that fighting words have protected expressive content. Ante, at 384-385.
Earlier this Term, seven of the eight participating Members of the Court agreed that strict scrutiny analysis applied in Simon & Schuster, Inc. v. Members of N. Y. State Crime Victims Bd., 502 U. S. 105 (1991), in which we struck down New York’s “Son of Sam” law, which required “that an accused or convicted criminal’s income from works describing his crime be deposited in an escrow account.” Id., at 108.
The Burson dissenters did not complain that the plurality erred in applying strict scrutiny; they objected that the plurality was not sufficiently rigorous in its review. 504 U. S., at 225-226 (Stevens, J., dissenting).
Justice Scalia concurred in the judgment in Burson, reasoning that the statute, “though content based, is constitutional [as] a reasonable, viewpoint-neutral regulation of a nonpublic forum.” Id., at 214. However, nothing in his reasoning in the present case suggests that a content-based ban on fighting words would be constitutional were that ban limited to nonpublic fora. Taken together, the two opinions suggest that, in some settings, political speech, to which “the First Amendment ‘has its fullest and most urgent application/” is entitled to less constitutional protection than fighting words. Eu v. San Francisco Cty. Democratic Central Comm., 489 U. S. 214, 223 (1989) (quoting Monitor Patriot Co. v. Roy, 401 U. S. 265, 272 (1971)).
The majority is mistaken in stating that a ban on obscene works critical of government would fail equal protection review only because the ban would violate the First Amendment. Ante, at 384-385, n. 4. While decisions such as Police Dept. of Chicago v. Mosley, 408 U. S. 92 (1972), recognize that First Amendment principles may be relevant to an equal protection claim challenging distinctions that impact on protected expression, id., at 95-99, there is no basis for linking First and Fourteenth Amendment analysis in a case involving unprotected expression. Certainly, one *407need not resort to First Amendment principles to conclude that the sort of improbable legislation the majority hypothesizes is based on senseless distinctions.
Indeed, such a law is content based in and of itself because it distinguishes between threatening and nonthreatening speech.
The consequences of the majority’s conflation of the rarely used secondary effects standard and the O’Brien test for conduct incorporating “speech” and “nonspeech” elements, see generally United States v. O’Brien, 391 U. S. 367, 376-377 (1968), present another question that I fear will haunt us and the lower courts in the aftermath of the majority’s opinion.
Petitioner can derive no support from our statement in Virginia v. American Booksellers Assn., Inc., 484 U. S. 383, 397 (1988), that “the statute must be ‘readily susceptible’ to the limitation; we will not rewrite a state law to conform it to constitutional requirements.” In American Booksellers, no state court had construed the language in dispute. In that instance, we certified a question to the state court so that it would have an opportunity to provide a narrowing interpretation. Ibid. In Erznoznik v. Jacksonville, 422 U. S. 205, 216 (1975), the other case upon which petitioner principally relies, we observed not only that the ordinance at issue was not “by its plain terms . . . easily susceptible of a narrowing construction,” but that the state courts had made no effort to restrict the scope of the statute when it was challenged on overbreadth grounds.
Although the First Amendment protects offensive speech, Johnson v. Texas, 491 U. S., at 414, it does not require us to be subjected to such expression at all times, in all settings. We have held that such expression may be proscribed when it intrudes upon a “captive audience.” Frisby v. Schultz, 487 U. S. 474, 484-485 (1988); FCC v. Pacifica Foundation, 438 U. S. 726, 748-749 (1978). And expression may be limited when it merges into conduct. United States v. O’Brien, 391 U. S. 367 (1968); cf. Meritor Savings Bank, F. S. B. v. Vinson, 477 U. S. 57, 66 (1986). However, because of the manner in which the Minnesota Supreme Court construed the St. Paul ordinance, those issues are not before us in this case.