dissenting from rejection of suggestion for rehearing en bane:
I dissent from the denial of rehearing en bane for the reasons set forth in my dissent above [p. 306], and for the reasons set forth in Judge Norris’s opinion.
WILLIAM A. NORRIS, Circuit Judge,respecting the denial of rehearing en banc:
I
In rejecting the facial challenge to the Seattle ordinance on First Amendment grounds, Roulette v. City of Seattle, 78 F.3d 1425 (9th Cir.1996), makes new law by departing from the standard that conduct needs only “a significant expressive element” to merit First Amendment protection. Arcara v. Cloud Books, 478 U.S. 697, 702, 706, 106 S.Ct. 3172, 3175, 3177, 92 L.Ed.2d 568 (1986). Roulette replaces the Arcara standard with a more stringent standard for making facial challenges: A statute directed at conduct cannot be facially challenged unless the conduct is “patently expressive or communicative” or “integral to, or commonly associated with, expression.” Roulette, 78 F.3d at 1427-28.
Interestingly, the Roulette panel makes its new First Amendment test applicable to facial challenges only. It expressly disclaims any intent to make its new test applicable to as-applied First Amendment claims. Thus, Roulette reserves the right of plaintiffs to bring individual as-applied challenges to the Seattle ordinance, even after denying their right to bring a facial challenge: “Of course, nothing we say today forecloses the possibility of mounting a successful as applied challenge.” Id. at 1429 n. 10. In other words, under Roulette, conduct may be sufficiently expressive to mount an as-applied attack on a statute restricting it, but not sufficiently expressive to mount a facial overbreadth attack. As far as I can tell, this dichotomy is unprecedented in First Amendment jurisprudence.
None of the cases cited in Roulette supports the dichotomy it creates between facial and as-applied challenges. Most of the cited cases do not discuss the expressive potential *312of conduct because they involve behavior that is indisputably expressive.1 The cases that do address whether conduct is sufficiently expressive to come under the protective umbrella of the First Amendment do not distinguish between facial and as-applied claims in their analyses. To the contrary, when discussing the expressive potential of the conduct at issue, cases involving facial attacks borrow their reasoning freely from eases involving as-applied claims.2 The overlapping of facial and as-applied cases in the analyses shows that the threshold standard for deciding whether the conduct in question is sufficiently communicative to bring First Amendment analysis into play does not vary depending on whether the plaintiff brings a facial or an as-applied claim.3
The real differences between facial over-breadth and as-applied analyses do not emerge until after a court determines whether the behavior targeted by a statute has sufficient communicative content to trigger the First Amendment. In other words, the threshold inquiry common to both facial overbreadth and as-applied challenges is whether the behavior has “a significant expressive element.” Only after the threshold inquiry do the two analyses diverge. In an as-applied challenge, there is a narrow focus on the particular plaintiffs behavior and whether the statute is constitutional as applied to her. In a facial overbreadth challenge, there is a broad focus on the entire range of behavior affected by the statute, and whether the unconstitutional applications of a statute are substantial in relation to the statute’s legitimate effect. Broadrick v. Oklahoma, 413 U.S. 601, 612, 93 S.Ct. 2908, 2916, 37 L.Ed.2d 830 (1973).
Under proper overbreadth analysis, if a court determines that a statute regulates behavior with “a significant expressive element,” it must then ask whether there are a substantial number of instances in which the statute will violate the First Amendment. See Broadrick, 413 U.S. at 601, 93 S.Ct. at 2910; Yniguez v. Arizonans for Official English, 69 F.3d 920, 932 (9th Cir.1995), cert. granted, — U.S. -, 116 S.Ct. 1316, 134 L.Ed.2d 469 (1996). To determine whether the statute’s prohibitions on expressive conduct are unconstitutional, a court may have to ask whether the prohibitions are valid time, place, and manner restrictions. See Members of the City Council v. Taxpayers for Vincent, 466 U.S. 789, 803-17, 104 S.Ct. 2118, 2127-35, 80 L.Ed.2d 772 (1984). Thus, even if conduct hit by a statute is deemed expressive by the threshold inquiry, a facial challenge will fail unless the parties show that the statute hits a substantial amount of protected expression and that its restrictions on this expression violate the First Amendment.
*313II
The Seattle ordinance at issue in this case prohibits any person from sitting or lying down on a public sidewalk, or upon an object placed on a public sidewalk, between 7:00 a.m. and 9:00 p.m. in commercial areas of the city. Seattle Mun. Code § 15.48.040. The plaintiffs are political activists, social service providers, a deputy registrar of voters, a street musician, and homeless people. Based on the false dichotomy it creates, the Roulette panel holds that no one can bring a facial overbreadth challenge to the Seattle ordinance because sitting or lying on the sidewalk is not sufficiently expressive to merit First Amendment protection. Sitting or lying on the sidewalk, however, is sufficiently expressive to invoke the First Amendment if a plaintiff brings an as-applied challenge. Roulette, 78 F.3d at 1428 nn. 7 & 10.
The Roulette panel holds that a facial challenge to a statute directed at conduct must fail unless the conduct is “patently expressive or communicative” or “integral to, or commonly associated with, expression.” Id. at 1427-28. Since a majority of the panel believes that sitting or lying on the sidewalk is not “integral to, or commonly associated with, expression,” id. at 1428, it holds that plaintiffs cannot bring a facial challenge to the Seattle ordinance. Thus, the panel never reaches the question of how much expressive conduct the ordinance hits, nor whether the ordinance’s restrictions on expression are reasonable time, place, and manner restrictions.
The test the panel creates is impermissi-bly subjective. In applying the new test, the panel interprets “patently” and “integral to, or commonly associated with,” as a license to make a completely subjective judgment about the expressive nature of sitting. The panel simply pronounces ex cathedra that sitting or lying on the sidewalk is not integral to, or commonly associated with, expression. There is no discussion of the possible communicative power of sitting, such as the possibility that a beggar’s message is dramatized by sitting, instead of standing or walking. Rather, we have only the words “sitting” and “lying” and the imperious conclusion that these activities, when judged in a vacuum, are not sufficiently expressive to make the ordinance vulnerable to a facial overbreadth challenge.
It surely cannot be the law that such imperious, subjective reasoning of judges can dictate whether sitting by a speaker, artist, musician, or solicitor is sufficiently expressive to permit a facial First Amendment overbreadth challenge. For example, while the panel belittles the expressive power of sitting, an entire genre of fourteenth century paintings, the Madonna of Humility, is defined by the expressive nature of sitting. As noted by the prominent art historian Millard Meiss, “[T]he humility of the Virgin resided primarily in the single fact that she was seated on the ground.” Millard Meiss, Painting in Florence and Siena after the Black Death: The Arts, Religion, and Society in the Midr-Fourteenth Century 132 n. 1 (1951). It is settled law that the First Amendment protects a person’s right to choose how to express herself, including her right to decide the manner in which she communicates. Cohen v. California, 403 U.S. 15, 91 S.Ct. 1780, 29 L.Ed.2d 284 (1971). First Amendment inquiry into the expressiveness of conduct must proceed from the speaker’s perspective. The panel errs because it proceeds from its own subjective point of view.
The correct test for whether conduct is expressive enough to implicate the First Amendment looks at the circumstances surrounding the conduct. See, e.g., id. at 15, 91 S.Ct. at 1783. Conduct triggers the First Amendment when the actor intends to convey a particularized message, and the likelihood is great under the circumstances that the message will be understood by those who view it. Spence v. Washington, 418 U.S. 405, 410-11, 94 S.Ct. 2727, 2730, 41 L.Ed.2d 842 (1974). Conduct needs only a “significant expressive element” or “at least the semblance of expressive activity” to invoke First Amendment analysis. Arcara, 478 U.S. at 702, 706, 106 S.Ct. at 3175, 3177. Thus, “[t]he fact that sitting can possibly be expressive,” Roulette, 78 F.3d at 1427, is enough to trigger First Amendment facial overbreadth analysis. Even conduct that is “[not] necessarily expressive ... [nor] ordinarily expressive” cannot be summarily dismissed as insufficiently expressive to trigger *314overbreadth analysis. Arcara, 478 U.S. at 702, 106 S.Ct. at 3175.
By viewing sitting in a vacuum, the panel majority reaches a foreordained conclusion, since an action devoid of context, though “possibly expressive,” can easily be characterized as pure conduct. In examining the medium divorced from the message, the panel ignores the power of the medium to enhance the message. Essentially the panel holds that an ordinance which on its face is aimed only at conduct cannot be subject to a First Amendment facial overbreadth challenge, even if it hits a great deal of protected expression in the process. This theory is unprecedented in First Amendment jurisprudence.
The panel insists that its newly minted test is perfectly reasonable, because even without sitting, “homeless people remain free to beg_ [vjoter registrars may solicit appli-cations_ [mjembers of the Freedom Socialist Party may doggedly pursue petition signatures and donations,” etc. Roulette, 78 F.3d at 1428. Again, the panel confuses the analysis. The availability of alternative means of expression may be significant in First Amendment analysis, but as part of a time, place, and manner inquiry. See Ward v. Rock Against Racism, 491 U.S. 781, 791, 109 S.Ct. 2746, 2753-54, 105 L.Ed.2d 661 (1989); One World One Family Now v. City & County of Honolulu, 76 F.3d 1009, 1012 (9th Cir.1996). Once again, the panel does something that is unique in First Amendment jurisprudence: It uses time, place, and manner reasoning in deciding the threshold question whether First Amendment interests are sufficiently implicated to even permit a facial overbreadth challenge.
The cases the panel cites as using the words “patently expressive or communicative,” or “integral to, and commonly associated with, expression” fail to support the panel’s own use of this language. The panel imports the phrase “commonly associated with expression” from City of Lakewood v. Plain Dealer Publishing Co., 486 U.S. 750, 760, 108 S.Ct. 2138, 2145, 100 L.Ed.2d 771 (1988), but that case did not deal with the distinction between expressive and non-expressive conduct. Lakewood involved the circulation of newspapers, and the Court simply described this activity as “expression or conduct commonly associated with expression.” Yet, the Roulette panel seizes upon that descriptive language and transforms it into a standard for determining whether conduct is sufficiently expressive to mount a facial challenge.
Moreover, the words “patently” and “integral to,” used by the panel as part of its new test, do not appear in any of the authorities cited by the panel.4 In sum, the panel cobbles its test together from its own imagination and language taken out of context.
Ill
In answering the threshold question in this facial overbreadth ease, the panel should have considered whether sitting or lying on the sidewalk has “a significant expressive element,” regardless of whether the plaintiff raises an as-applied or a facial challenge. This inquiry requires us to think about concrete instances where the conduct may be expressive. If the panel had considered the expressive elements of sitting or lying on a busy sidewalk in a commercial area, instead of merely indulging its subjective biases about the expressive value of “sitting” in the abstract under its new test, it would have considered that sitting or lying on the pavement may be inextricably intertwined with the messages of street people, and thus sufficiently expressive to permit a facial over-breadth challenge. As Judge Wilken asserted when she issued a preliminary injunction on First Amendment grounds against a Berkeley ordinance almost identical to the Seattle ordinance: “One message which may be communicated by the act of sitting on the sidewalk is the message that the solicitor is in serious need.... [and] too weak, ill, or *315defeated by circumstances to stand.” Berkeley Community Health Project v. City of Berkeley, 902 F.Supp. 1084, 1092-93 (N.D.Cal.1995).
If the panel had applied the correct analysis, it would have considered whether many people beg while seated because sitting is a non-threatening posture that signals passivity. Sitting can make the pedestrian feel safe, because that posture suggests that the solicitor is not aggressive and intends no harm. It can also communicate the beggar’s degree of desperation, by signalling surrender, weakness, and humiliation, thus altering the character of begging and making the solicitation more effective. There is an indigent woman who sits on a sidewalk along Rodeo Drive in Beverly Hills. Against the background of the retail mecca that literally defines American wealth, the image of this woman, plainly destitute and desperate, sitting against a lamppost, sends a powerful message about the plight of the downtrodden. Perhaps this particular woman chose this particular sidewalk because she believed the most effective means of begging was to confront the rich with her message of abject poverty in the face of their own affluence. And perhaps she chose to sit, placing herself at the feet of the wealthy, to amplify her message of degradation and dependence. In any case, it is presumptuous, if not arrogant, for the majority to disregard such real world possibilities and decide in a vacuum that “sitting” is not sufficiently expressive to trigger a facial overbreadth analysis.
Even beyond helping to impart the inescapable message of weakness and humility, sitting may be used to enhance a statement about poverty in general. When a dishe-velled man badly in need of a bath chooses to sit on the sidewalk where shoppers toss their cigarette butts and other trash, he conveys a message about the degradation that results from society’s failure to accommodate the essential needs of all its citizens. His message addresses what many consider to be the single most compelling problem facing our nation: the growing disparity between the haves and the have-nots.5
Sitting has, in fact, been closely tied to political messages. There is no question that sit-ins are a paradigm of political protest. See Brown v. Louisiana, 383 U.S. 131, 139, 86 S.Ct. 719, 722, 15 L.Ed.2d 637 (1966) (plaintiffs “sat and stood in the room, quietly, as monuments of protest against the segregation of the library”). No one can question that Gandhi used the posture of sitting to symbolize peaceful, nonviolent resistance.
Conclusion
The Seattle ordinance raises difficult and important First Amendment questions, which should not be dismissed out of hand, the way the panel has done. Without any basis, the panel falsely dichotomizes facial overbreadth and as-applied challenges from the start, and creates a First Amendment test that effectively immunizes the Seattle ordinance, and others like it, from facial challenge. It is especially troubling that the panel gives such short shrift to plaintiffs’ facial overbreadth claims, because the Seattle ordinance targets street people, who are among the most powerless in our society. If the real purpose of the ordinance were to combat sidewalk congestion, then the ordinance would have targeted all obstacles to pedestrian traffic. But the ordinance has an express exception allowing people to sit on the sidewalk, so long as they do so on a chair suppled by a merchant. Seattle Mun. Code *316§ 15.48.040(B)(4).6 Thus, those deemed “desirable” by shopkeepers may sit on the sidewalk and obstruct pedestrian traffic all day long, drinking their cappuccinos and reading their Wall Street Journals to their hearts’ content. But someone with an unpopular political message or an unsightly beggar symbolizing the failure of our society to achieve economic justice, may not sit, even to add power and content to his message. As Judge Pregerson said in dissent to the panel’s decision, “Seattle seeks economic preservation by ridding itself of social undesirables — homeless or otherwise — who sit or lie on the sidewalks, and this is done to protect the sensibilities of shoppers.” Roulette, 78 F.3d at 1432 (Pregerson, J., dissenting). Yet “a bare ... desire to harm a politically unpopular group cannot constitute a legitimate governmental interest.” Romer v. Evans, — U.S. -, 116 S.Ct. 1620, 134 L.Ed.2d 855 (1996) (quoting Department of Agriculture v. Moreno, 413 U.S. 528, 534, 93 S.Ct. 2821, 2825-26, 37 L.Ed.2d 782 (1973)).
Pedestrian safety and the free flow of pedestrian traffic in commercial centers are legitimate municipal concerns. No one argues that cities cannot address the problem of congestion through carefully tailored ordinances. However, municipalities must respect the First Amendment when they formulate their ordinances, and it is our job, in deciding eases such .as this one, to provide them guidance in doing so. Roulette offers no guidance to any municipality trying to fashion a remedial ordinance that is consistent with First Amendment values.
. See R.A.V. v. City of St. Paul, 505 U.S. 377, 112 S.Ct. 2538, 120 L.Ed.2d 305 (1992) (holding without discussion that display of symbols is clearly expressive speech); City of Houston v. Hill, 482 U.S. 451, 107 S.Ct. 2502, 96 L.Ed.2d 398 (1987) (holding without discussion that verbal interruption of police officers is speech); Schad v. Borough of Mount Ephraim, 452 U.S. 61, 101 S.Ct. 2176, 68 L.Ed.2d 671 (1981) (holding without discussion that live entertainment, including topless barroom dancing, has long been protected by the First Amendment); Broadrick v. Oklahoma, 413 U.S. 601, 93 S.Ct. 2908, 37 L.Ed.2d 830 (1973) (assuming without deciding that wearing political buttons or using bumper stickers is "arguably protected” conduct).
. See, Village of Schaumburg v. Citizens for a Better Environment, 444 U.S. 620, 100 S.Ct. 826, 63 L.Ed.2d 73 (1980) (holding solicitation has sufficient "speech interests” to merit First Amendment protection (relying on Bates v. State Bar of Arizona, 433 U.S. 350, 97 S.Ct. 2691, 53 L.Ed.2d 810 (1977); Thomas v. Collins, 323 U.S. 516, 65 S.Ct. 315, 89 L.Ed.430 (1945); Schneider v. State, 308 U.S. 147, 60 S.Ct. 146, 84 L.Ed. 155 (1939))), Buckley v. Valeo, 424 U.S. 1, 96 S.Ct. 612, 46 L.Ed.2d 659 (1976) (holding contributions and expenditures are protected speech (relying on Bigelow v. Virginia, 421 U.S. 809, 95 S.Ct. 2222, 44 L.Ed.2d 600 (1975); United States v. O’Brien, 391 U.S. 367, 88 S.Ct. 1673, 20 L.Ed.2d 672 (1968); Cox v. Louisiana, 379 U.S. 559, 85 S.Ct. 476, 13 L.Ed.2d 487 (1965))).
.By creating a more stringent threshold test for facial challenges than for as-applied claims, the Roulette panel also negates the very premise behind the facial overbreadth doctrine. That doctrine was created to enable plaintiffs to bring claims when “the statute's very existence may cause others not before the court to refrain from constitutionally protected speech or expression.” Broadrick v. Oklahoma, 413 U.S. 601, 612, 93 S.Ct. 2908, 2916, 37 L.Ed.2d 830 (1973). Thus, the doctrine rests on the notion that as-applied challenges may not be effective in combatting laws that have a speech-chilling effect because potential plaintiffs will remain silent rather than run the risk of facing civil or criminal penalties if an as-applied defense fails. See Brockett v. Spokane Arcades, Inc., 472 U.S. 491, 503, 105 S.Ct. 2794, 2801, 86 L.Ed.2d 394 (1985).
. See R.A.V. v. City of St. Paul, 505 U.S. 377, 112 S.Ct. 2538, 120 L.Ed.2d 305 (1992) (discussing pure speech, not conduct); Houston v. Hitt, 482 U.S. 451, 107 S.Ct. 2502, 96 L.Ed.2d 398 (1987) (same); Secretary of State of Maryland v. Joseph H. Munson Co., 467 U.S. 947, 104 S.Ct. 2839, 81 L.Ed.2d 786 (1984) (holding conduct "so intertwined with speech” deserves First Amendment protection); Schad v. Borough of Mt. Ephraim, 452 U.S. 61, 101 S.Ct. 2176, 68 L.Ed.2d 671 (1981) (holding conduct long held to be protected deserves First Amendment scrutiny); Buckley v. Valeo, 424 U.S. 1, 96 S.Ct. 612, 46 L.Ed.2d 659 (1976) (discussing pure speech, not conduct); Chase v. Davelaar, 645 F.2d 735 (9th Cir.1981) (discussing conduct "ordinarily regarded as expressive,” but not limiting overbreadth challenges to such conduct).
. A University of Michigan study recently found that the wealthiest 10% of American households held 66.8% of the nation's wealth in 1994, up over 5% since 1989. Meanwhile, the poorest 10% had debts averaging over $7,000, up almost $2,500 since 1989. Keith Bradsher, Rich Control More of U.S. Wealth, Study Says, as Debts Grow for Poor, N.Y. Times, June 22, 1996, at 31. The Census Bureau also reports that the gap between the most affluent Americans and eveiy-one else was wider in 1994 than it has been since the end of World War II. Steven A. Holmes, Income Disparity Between Poorest and Richest Rises, N.Y. Times, June 20, 1996, at Al; see also Robin Wright, U.S. Child Poverty Worst Among Richest Nations, L.A. Times, June 12, 1996, at A22 (reporting UNICEF found United States has highest child poverty rate among world's rich industrialized nations, but also houses world’s richest children).
. Seattle Mun. Code, § 15.48.040 provides:
B. Exceptions. The prohibition [against sitting and lying established by the ordinance] shall not apply to any person ... (4) sitting on a chair or bench located on the public sidewalk which is supplied by a public agency or by the abutting private property owner....