United States v. Michael Miselis

PUBLISHED

UNITED STATES COURT OF APPEALS
FOR THE FOURTH CIRCUIT

No. 19-4550

UNITED STATES OF AMERICA,

Plaintiff − Appellee,

v.

MICHAEL PAUL MISELIS,

Defendant – Appellant.

------------------------------

THE FREE EXPRESSION FOUNDATION, INC.,

Amicus Supporting Appellant.

No. 19-4551

UNITED STATES OF AMERICA,

Plaintiff − Appellee,

v.

BENJAMIN DRAKE DALEY,

Defendant – Appellant.

------------------------------

THE FREE EXPRESSION FOUNDATION, INC.,
 Amicus Supporting Appellant.

Appeals from the United States District Court for the Western District of Virginia, at
Charlottesville. Norman K. Moon, Senior District Judge. (3:18-cr-00025-NKM-JCH-2;
3:18-cr-00025-NKM-JCH-1)

Argued: January 31, 2020 Decided: August 24, 2020

Before KING, DIAZ, and RUSHING, Circuit Judges.

Affirmed by published opinion. Judge Diaz wrote the opinion, in which Judge King and
Judge Rushing joined.

ARGUED: Lisa M. Lorish, OFFICE OF THE FEDERAL PUBLIC DEFENDER,
Charlottesville, Virginia, for Appellants. Laura Day Rottenborn, OFFICE OF THE
UNITED STATES ATTORNEY, Roanoke, Virginia, for Appellee. ON BRIEF: Juval O.
Scott, Federal Public Defender, OFFICE OF THE FEDERAL PUBLIC DEFENDER,
Roanoke, Virginia, for Appellant Benjamin Daley. Raymond C. Tarlton, TARLTON |
POLK PLLC, Raleigh, North Carolina, for Appellant Michael Miselis. Thomas T. Cullen,
United States Attorney, OFFICE OF THE UNITED STATES ATTORNEY, Roanoke,
Virginia, for Appellee. Glen K. Allen, Baltimore, Maryland, for Amicus The Free
Expression Foundation, Inc.

2
DIAZ, Circuit Judge:

Michael Paul Miselis and Benjamin Drake Daley entered conditional guilty pleas to

one count each of conspiracy to commit an offense against the United States, in violation

of 18 U.S.C. § 371, with the substantive offense being a violation of the Anti-Riot Act, 18

U.S.C. §§ 2101–02. The charges arise from the defendants’ violent participation in three

white supremacist rallies during the year 2017: two in their home state of California, and

the third being the notorious “Unite the Right” rally in Charlottesville, Virginia.

On appeal, the defendants challenge their convictions on the grounds that the Anti-

Riot Act is facially overbroad under the Free Speech Clause of the First Amendment, as

well as void for vagueness under the Due Process Clause of the Fifth Amendment.

Reviewing these issues de novo, Giovani Carandola, Ltd. v. Fox, 470 F.3d 1074, 1079 (4th

Cir. 2006), we disagree that the statute is unconstitutionally vague. But we agree that it

treads too far upon constitutionally protected speech—in some of its applications.

While the category of speech that lies at the core of the Anti-Riot Act’s prohibition,

called “incitement,” has never enjoyed First Amendment protection, the statute sweeps up

a substantial amount of speech that remains protected advocacy under the modern

incitement test of Brandenburg v. Ohio, 395 U.S. 444 (1969) (per curiam), insofar as it

encompasses speech tending to “encourage” or “promote” a riot under 18 U.S.C.

§ 2101(a)(2), as well as speech “urging” others to riot or “involving” mere advocacy of

violence under 18 U.S.C. § 2102(b).

In all other respects, however, the statute comports with the First Amendment. And

because the discrete instances of overbreadth are severable from the remainder of the

3
statute, the appropriate remedy is to invalidate the statute only to the extent that it reaches

too far, while leaving the remainder intact.

Finally, because the factual bases for the defendants’ guilty pleas conclusively

establish that their own substantive offense conduct—which involves no First Amendment

activity—falls under the Anti-Riot Act’s surviving applications, their convictions stand.

I.

We begin with an overview of the defendants’ offense conduct, the procedural

history, and the Anti-Riot Act.

A.

The defendants (who are residents of Southern California) began in early 2017 to

associate with a local white supremacist group called the “Rise Above Movement,” or

“RAM” for short. Billing itself as a “combat-ready, militant group of a new nationalist

white identity movement,” the group’s chief purpose was to attend “purported ‘political’

rallies” (typically organized by other groups) at which its members engaged in violent

attacks on counter-protestors. J.A. 227, 232. And to prepare for such rallies, RAM

members spent their weekends training in martial arts and other combat techniques.

The charges in this case arise from three such rallies held in 2017. The first took

place on March 25, in Huntington Beach, California, where the defendants and their

colleagues first obtained front-page notoriety for RAM by carrying out numerous assaults

against counter-protesters. They celebrated this coverage among themselves and posted it

on various white supremacist platforms to recruit new members to their ranks.

4
 The second rally took place on April 15, in Berkeley, California. The defendants

and a handful of other RAM members drove up to Berkeley the day before, riding together

in an eleven-passenger rental van. Hundreds of white nationalists attended the rally, as did

dozens of counter-protestors, and violence again broke out amongst the camps. In one

clash, the defendants and their colleagues trampled a barrier separating the two camps and

assaulted a group of counter-protestors. In another, after the rally had been broken up and

the participants dispersed into the streets of downtown Berkeley, the defendants and their

colleagues chased after another group of counter-protestors, whom they proceeded to

punch, kick, and stomp; defendant Miselis even broke his hand in the effort.

After returning from Berkeley, RAM members became aware that the now-

infamous “Unite the Right” rally would be held at Emancipation Park in Charlottesville,

Virginia, on August 12, 2017. The rally had been organized by Jason Kessler, a self-styled

“white advocate,” to protest the City Council’s vote to remove a statue of the Confederate

general Robert E. Lee from the park. See Hawes Spencer & Sheryl Gay Stolberg, Virginia

Town Is on Edge Over Confederate Statue, N.Y. Times, Aug. 12, 2017, at A12. The

defendants and at least two of their RAM colleagues, Cole Evan White and Thomas Walter

Gillen (who were later charged alongside them), each purchased roundtrip airfare to attend.

The defendants and their colleagues arrived in Charlottesville on August 11, 2017.

That night, they joined hundreds of other white nationalists for a torch-lit march on the

campus of the University of Virginia, just west of downtown Charlottesville. There, the

torch-bearers chanted slogans such as “Blood and soil!” and “Jews will not replace us!” as

they made their way to the statue of Thomas Jefferson in front of The Rotunda (the

5
University’s signature building), where they confronted a smaller group of student counter-

protesters bearing a banner that read, “VA Students Act Against White Supremacy.” J.A.

230, 235. A brawl ensued between the two camps, in which defendant Daley and other

RAM members attacked multiple counter-protestors with their tiki torches.

The morning of August 12, the defendants arrived at Emancipation Park for the

long-planned “Unite the Right” rally. But by 11 a.m., violence erupted (yet again) between

groups of white nationalists and counter-protestors who had surrounded the park. See

Sheryl Gay Stolberg & Brian M. Rosenthal, White Nationalist Protest Leads to Deadly

Violence, N.Y. Times, Aug. 13, 2017, at A1. Police promptly declared the assembly

unlawful and began to clear the park, while officials from the city declared a state of

emergency, citing an “imminent threat of civil disturbance, unrest, potential injury to

persons, and destruction of public and personal property.” Id.

Much of the violence associated with the “Unite the Right” rally took place after it

had been made to disperse, in the streets of downtown Charlottesville. 1 For their part, the

defendants engaged in several skirmishes both during and after the rally, including a clash

near the 2nd Street NE entrance to the park in which they “collectively pushed, punched,

kicked, choked, head-butted, and otherwise assaulted” a group of counter-protestors, and

“not in self-defense.” J.A. 231, 236.

1
That violence culminated in the death of Heather D. Heyer, who was killed when
an avowed neo-Nazi deliberately plowed into her and over a dozen others with his car. See
Sheryl Gay Stolberg & Brian M. Rosenthal, White Nationalist Protest Leads to Deadly
Violence, N.Y. Times, Aug. 13, 2017, at A1.

6
 B.

Following a federal investigation, the defendants (along with Gillen and White)

were indicted on two counts each: (1) conspiracy to commit an offense against the United

States, in violation of 18 U.S.C. § 371, with the underlying offense being the substantive

violation set forth in Count 2; and (2) traveling in interstate commerce with intent to riot,

in violation of the Anti-Riot Act, 18 U.S.C. §§ 2101–02.

The defendants moved to dismiss the indictment, raising numerous challenges.

Following a hearing, the district court denied the motion. United States v. Daley, 378 F.

Supp. 3d 539, 545 (W.D. Va. 2019). The defendants each pled conditionally guilty to

Count 1 the next day, subject to their rights to appeal the constitutionality of the Anti-Riot

Act. The district court thereafter sentenced Daley to a 37-month prison term, while Miselis

received 27 months; each was also given two years of supervised release. They appealed. 2

C.

Congress passed the Anti-Riot Act as a rider to the Civil Rights Act of 1968, amidst

an era, not unlike our own, marked by a palpable degree of social unrest. See Anti-Riot

Act, Pub. L. No. 90-284 § 104(a), 82 Stat. 73, 75–77 (April 11, 1968). The statute’s

passage followed on the heels of what has been deemed the “long, hot summer of 1967,”

in which more than 150 cities across 34 states witnessed riots stirred by issues such as

2
Gillen also pled guilty and filed an appeal alongside the defendants, see United
States v. Gillen, No. 19-4553 (4th Cir. filed July 30, 2019), but moved to sever his appeal.
We granted the motion, and have since held Gillen’s appeal in abeyance pending our
decision here. As for White, he pled guilty to Count 1 as well, see United States v. Daley
et al., No. 3:18-mj-24 (W.D. Va. 2018), ECF Nos. 57–60, but hasn’t filed an appeal.

7
racial injustice and the war in Vietnam. See generally Malcolm McLaughlin, The Long,

Hot Summer of 1967: Urban Rebellion in America (2014). And the statute’s immediate

catalyst was the upheaval sparked anew, in over 100 American cities, by the assassination

of Martin Luther King, Jr. on April 4, 1968. See Marvin Zalman, The Federal Anti-Riot

Act and Political Crime: The Need for Criminal Law Theory, 20 Vill. L. Rev. 897, 912

(1975).

The turbulence that lingered throughout 1968 gave rise to most of the few cases in

which courts have addressed—and upheld—the constitutionality of the Anti-Riot Act on

overbreadth or vagueness grounds. See United States v. Dellinger, 472 F.2d 340, 355 (7th

Cir. 1972), cert. denied, 410 U.S. 970 (1973); United States v. Hoffman, 334 F. Supp. 504,

509 (D.D.C. 1971); In re Shead, 302 F. Supp. 560, 567 (N.D. Cal. 1969), aff’d sub nom.

on other grounds, Carter v. United States, 417 F.2d 384 (9th Cir. 1969). The statute wasn’t

challenged again until along came RAM, whose participation in the California rallies

described above also gave rise to the other recent facial challenge, and the first successful

one. See United States v. Rundo, No. 18-cr-759 (C.D. Cal. June 3, 2019), appeal docketed,

No. 19-50189 (9th Cir. June 12, 2019) (finding the Anti-Riot Act facially overbroad and

dismissing indictments against RAM members who didn’t travel to Charlottesville).

The Anti-Riot Act comprises three provisions that bear on the defendants’ facial

challenges: one that proscribes a range of speech and conduct, and two that contribute to

the definition of such speech and conduct. First and foremost, § 2101(a) provides that:

Whoever travels in interstate or foreign commerce or uses any facility of
interstate or foreign commerce, including, but not limited to, the mail,
telegraph, telephone, radio, or television, with intent—

8
 (1) to incite a riot; or

(2) to organize, promote, encourage, participate in, or carry on a riot;
or

(3) to commit any act of violence in furtherance of a riot; or

(4) to aid or abet any person in inciting or participating in or carrying
on a riot or committing any act of violence in furtherance of a riot;

and who either during the course of any such travel or use or thereafter
performs or attempts to perform any other overt act for any purpose specified
in subparagraph (A), (B), (C), or (D) of this paragraph[3]—

Shall be fined under this title, or imprisoned not more than five years, or
both.

18 U.S.C. § 2101(a).

Second, § 2102(a) defines the “riot” at the center of the statute, and which forms the

object of § 2101(a)’s laundry list of alternative purposes, to mean

a public disturbance involving (1) an act or acts of violence by one or more
persons part of an assemblage of three or more persons, which act or acts
shall constitute a clear and present danger of, or shall result in, damage or
injury to the property of any other person or to the person of any other
individual or (2) a threat or threats of the commission of an act or acts of
violence by one or more persons part of an assemblage of three or more
persons having, individually or collectively, the ability of immediate
execution of such threat or threats, where the performance of the threatened
act or acts of violence would constitute a clear and present danger of, or
would result in, damage or injury to the property of any other person or to
the person of any other individual.

Id. § 2102(a).

3
As codified, the statute contains a footnote in this location explaining that the
reference to “subparagraph (A), (B), (C), or (D)” is the result of a drafting mistake, and
should read “[sub]paragraph (1), (2), (3), or (4).” See 18 U.S.C. § 2101 n.1.

9
 And third, § 2102(b) glosses the ordinary meaning of each of the speech- and

conduct-related verbs found in § 2101(a)(1)–(2) as follows:

As used in this chapter, the term “to incite a riot”, or “to organize, promote,
encourage, participate in, or carry on a riot”, includes, but is not limited to,
urging or instigating other persons to riot, but shall not be deemed to mean
the mere oral or written (1) advocacy of ideas or (2) expression of belief, not
involving advocacy of any act or acts of violence or assertion of the rightness
of, or the right to commit, any such act or acts.

Id. § 2102(b). Because the statute’s constitutionality hinges on these three interlocking

provisions, we focus on them as we address the defendants’ appeal.

II.

Before turning to the defendants’ facial challenges to the Anti-Riot Act, we take up

an issue on which we sought supplemental briefing: whether the defendants have standing

to contest the constitutionality of a statute forming the object of their conspiracy

convictions under 18 U.S.C. § 371. We agree with the parties that they do.

It is well-established that a conspiracy consists of “an agreement among the

defendants to do something which the law prohibits.” United States v. Hedgepeth, 418

F.3d 411, 420 (4th Cir. 2005) (quoting United States v. Meredith, 824 F.2d 1418, 1428 (4th

Cir. 1987)); see also Salinas v. United States, 522 U.S. 52, 65 (1997) (“A conspirator must

intend to further an endeavor which, if completed, would satisfy all of the elements of a

substantive criminal offense . . . .”). Indeed, this axiomatic principle is embedded directly

in the text of § 371, which spells out a conspiracy “to commit any offense against the

United States.” See 18 U.S.C. § 371.

10
 Yet because the object of an agreement can’t be unlawful “if the statute defining [it]

is unconstitutional,” it follows that “no prosecution for conspiracy to commit that offense

will lie.” United States v. Rosen, 520 F. Supp. 2d 786, 792 (E.D. Va. 2007). We therefore

agree with our sister circuit that “the statutory requirement of conspiring to commit an

‘offense against the United States,’ 18 U.S.C. § 371, is not fulfilled by an offense which

fails to meet constitutional muster.” United States v. Baranski, 484 F.2d 556, 561 (7th Cir.

1973). Accordingly, because the defendants’ convictions under § 371 cannot stand if the

Anti-Riot Act is unconstitutional, we are satisfied that the defendants have standing to

pursue the facial challenges to which we now turn.

III.

The defendants contend that the Anti-Riot Act is facially overbroad, under the Free

Speech Clause of the First Amendment, in a variety of respects. We agree—in part.

In our view, the Anti-Riot Act sweeps up a substantial amount of speech that retains

the status of protected advocacy under Brandenburg insofar as it encompasses speech

tending to “encourage” or “promote” a riot under § 2101(a)(2), as well as speech “urging”

others to riot or “involving” mere advocacy of violence under § 2102(b). In all other

aspects, however, we find the statute consistent with the First Amendment. And because

we also find that the discrete areas of overbreadth are severable—meaning that the

remainder of the statute is constitutionally valid, capable of operating independently, and

consistent with Congress’s basic objectives—the appropriate remedy is to invalidate the

statute only to the extent that it reaches too far, while leaving the remainder intact.

11
 A.

We begin by setting out the principles that guide our overbreadth analysis. Here,

the defendants bring a facial challenge to the Anti-Riot Act, meaning they claim that the

statute is unconstitutional not as it applies to their own conduct, but rather “on its face,” as

it applies to the population generally. See Wash. State Grange v. Wash. State Republican

Party, 552 U.S. 442, 449 (2008). Such claims of facial invalidity “are disfavored for

several reasons.” Id. at 450 (cleaned up). For one thing, facial challenges “run contrary to

the fundamental principle of judicial restraint that courts should neither anticipate a

question of constitutional law in advance of the necessity of deciding it nor formulate a

rule of constitutional law that is broader than is required by the precise facts to which it is

to be applied.” Id. (cleaned up). Relatedly, facial challenges “threaten to short circuit the

democratic process by preventing laws embodying the will of the people from being

implemented in a manner consistent with the Constitution.” Id. at 451.

In light of these twin concerns, a facial challenge typically requires a showing that

“no set of circumstances exists under which the Act would be valid, i.e., that the law is

unconstitutional in all of its applications,” Wash. State Grange, 552 U.S. at 449 (cleaned

up); or “that the statute lacks any plainly legitimate sweep,” United States v. Stevens, 559

U.S. 460, 472 (2010) (cleaned up). And in assessing whether a statute meets one of these

high bars, courts must typically take care “not to . . . speculate about hypothetical or

imaginary cases.” Wash. State Grange, 552 U.S. at 50 (cleaned up).

12
 In the First Amendment context, however, the fear of chilling protected expression

“has led courts to entertain facial challenges based merely on hypothetical applications of

the law to nonparties.” Preston v. Leake, 660 F.3d 726, 738 (4th Cir. 2011). Under this

“second type” of facial challenge, a statute “may be invalidated as overbroad” as long as

“a substantial number of its applications are unconstitutional, judged in relation to the

statute’s plainly legitimate sweep.” Stevens, 559 U.S. at 473 (cleaned up).

This so-called overbreadth doctrine “allows a party to challenge a law facially under

the First Amendment by ‘describing a substantial number of instances of arguable

overbreadth of the contested law,’ even if the law is constitutional as applied to [himself].”

Preston, 660 F.3d at 738–39 (quoting Wash. State Grange, 552 U.S. at 449 n.6) (cleaned

up); see also Stevens, 559 U.S. at 483–84 (Alito, J., dissenting) (“[T]he over-breadth

doctrine allows a party to whom the law may be constitutionally applied to challenge the

statute on the ground that it violates the First Amendment rights of others.”). In fact, in the

overbreadth context, the “usual judicial practice” is to determine that the statute “would be

valid as applied” to the challenger’s own conduct before proceeding to a facial challenge

premised on the hypothetical conduct of others “unnecessarily.” Bd. of Trustees of State

Univ. of N.Y. v. Fox, 492 U.S. 469, 484–85 (1989); accord Preston, 660 F.3d at 738. 4

4
We adhere to the usual judicial practice here, being satisfied that the circumstances
under which the defendants raise their facial overbreadth challenge amount to a concession
that the Anti-Riot Act may be constitutionally applied to their own offense conduct. For
starters, because the defendants don’t appeal the district court’s rejection of their as-applied
challenge, see Daley, 378 F. Supp. 3d at 558–59, they’ve waived any argument to the
contrary, see, e.g., United States v. Hudson, 673 F.3d 263, 268 (4th Cir. 2012). Relatedly,
at oral argument, the defendants confirmed that they’ve abandoned their as-applied

13
 To maintain an “appropriate balance” between the “competing social costs” at issue

in the overbreadth context, the Supreme Court has “vigorously enforced the requirement

that a statute’s overbreadth be substantial, not only in an absolute sense, but also relative

to the statute’s plainly legitimate sweep.” United States v. Williams, 553 U.S. 285, 292

(2008). As the Court has explained,

On the one hand, the threat of enforcement of an overbroad law deters people
from engaging in constitutionally protected speech, inhibiting the free
exchange of ideas. On the other hand, invalidating a law that in some of its
applications is perfectly constitutional—particularly a law directed at
conduct so antisocial that it has been made criminal—has obvious harmful
effects.

Id. In consequence, it isn’t enough to render a statute susceptible to a facial attack that one

may simply “conceive of some impermissible applications.” Members of City Council of

Los Angeles v. Taxpayers for Vincent, 466 U.S. 789, 800 (1984).

Overbreadth analysis proceeds along several steps. Because “it is impossible to

determine whether a statute reaches too far without first knowing what the statute covers,”

we must first “construe the challenged statute.” Williams, 553 U.S. at 293. In so doing,

we must seek to avoid any “constitutional problems” by asking whether the statute is

“subject to [] a limiting construction.” New York v. Ferber, 458 U.S. 747, 769 n.24 (1982).

We must then determine whether, so construed, the statute “criminalizes a substantial

amount of protected expressive activity.” Williams, 553 U.S. at 297. Finally, if the statute

challenge. See Oral Arg. at 8:10–8:12 (“We’re never going to make an as-applied
challenge, Your Honor.”). Finally, for reasons we make clear in Part V, the record readily
substantiates the defendants’ tacit acknowledgement that the statute is “plainly legitimate
as applied” to their conduct. Cf. Stevens, 559 U.S. at 472–73 (proceeding under such an
assumption after finding that any as-applied challenge had been waived).
14
proves “impermissibly overbroad,” we must assess whether “the unconstitutional portion”

is “severable” from the remainder; if so, only that portion “is to be invalidated.” Ferber,

458 U.S. at 769 n.24. Altogether, these efforts to preserve a statute from facial invalidation

reflect the notion “that the overbreadth doctrine is strong medicine,” to be applied “only as

a last resort,” in cases where it is “truly warranted.” See id. at 769.

In conducting our analysis, we find it preferable, at least in the context of the Anti-

Riot Act, to begin (at step zero, as it were) by delineating the scope of unprotected speech

that the statute aims to regulate. Cf. Dellinger, 472 F.2d at 358 (“Ideally the analysis should

begin with a delineation of the scope of speech protected by the first amendment.”). With

that backdrop in mind, we’ll be better able to perceive where the statute overshoots its

target and purports to regulate a substantial amount of protected speech.

B.

A glance at the Anti-Riot Act reveals that the category of unprotected speech that

lies at the core of the statute’s prohibition is that which also lies at the origin of First

Amendment jurisprudence: “incitement.” In general legal parlance, “incitement” refers to

“[t]he act of persuading”—that is, of inducing—“another person to commit a crime.” See

Incitement, Black’s Law Dictionary (11th ed. 2019); cf. Persuade, Black’s Law Dictionary

(11th ed. 2019) (“To induce (another) to do something; to make someone decide to do

something[.]”). More important for our purposes is how the Supreme Court has defined

“incitement” in First Amendment jurisprudence. And notably, while the Court initially did

so much more broadly than the dictionary, the modern test does so almost as narrowly.

15
 The modern incitement test derives from the Court’s per curiam decision in

Brandenburg, see 395 U.S. 444, which came down in 1969, the year after the Anti-Riot

Act was enacted. That case concerned the conviction of a Ku Klux Klan leader under the

Ohio Criminal Syndicalism statute, id. at 444–45, which made it a crime to “advocate or

teach the duty, necessity, or propriety of violence as a means of accomplishing industrial

or political reform,” id. at 448 (cleaned up). 5

Though the Court had upheld an analogous statute in Whitney v. California, 274

U.S. 357 (1927), it asserted that Whitney “ha[d] been thoroughly discredited by later

decisions,” from which it distilled the principle that “the constitutional guarantees of free

speech” protected the “advocacy of the use of force or of law violation except where such

advocacy is directed to inciting or producing imminent lawless action and is likely to incite

or produce such action.” See id. at 447. And because the Ohio statute “purport[ed] to

punish mere advocacy” of lawless action as opposed to advocacy directed and likely to

produce imminent lawless action, the Court held that it fell “within the condemnation” of

the First Amendment. Id. at 449 (emphasis added).

While Brandenburg purported to draw its incitement test from midcentury cases,

it’s widely acknowledged that the Court had theretofore (including well after Whitney) used

a far more encompassing test, called the “clear and present danger” test, to determine when

5
The specific words giving rise to the Klansman’s prosecution in Brandenburg were
these: “We’re not a revengent organization, but if our President, our Congress, our
Supreme Court, continues to suppress the white, Caucasian race, it’s possible that there
might have to be some revengeance taken.” See id. at 446.

16
advocacy of lawlessness became unprotected incitement. See generally Wallace

Mendelson, Clear and Present Danger: From Schenk to Dennis, 52 Colum. L. Rev. 313

(1952). Under that test, “[t]he question in every case” is whether the speech was “of such

a nature” and “used in such circumstances . . . as to create a clear and present danger that

[it] w[ould] bring about the substantive evils that Congress has a right to prevent.” Schenck

v. United States, 249 U.S. 47, 52 (1919); see also Dennis v. United States, 341 U.S. 494,

509 (1951); Whitney, 274 U.S. at 374; Frohwerk v. United States, 249 U.S. 204, 206

(1919); Debs v. United States, 249 U.S. 211, 215 (1919). Devoid of any such limiting

criteria as directedness, likelihood, or imminence, the clear-and-present-danger test applied

to a wide range of advocacy that now finds refuge under Brandenburg. See Dennis, 341

U.S. at 516–17 (upholding conviction for mere advocacy of Communism); Whitney, 274

U.S. at 371–72 (same); Frohwerk, 249 U.S. at 206–07 (upholding conviction for mere

advocacy of disobedience to the draft); Schenck, 249 U.S. at 51–52 (same).

Brandenburg has thus been widely understood, starting with the two concurring

Justices, as having significantly (if tacitly) narrowed the category of incitement. See

Brandenburg, 395 U.S. at 449–50 (Black, J., concurring) (“[T]he ‘clear and present danger’

doctrine should have no place in the interpretation of the First Amendment.”); id. at 454

(Douglas, J., concurring) (“I see no place in the regime of the First Amendment for any

‘clear and present danger’ test . . . .”); see also, e.g., Denver Area Educ. Telecomms.

Consortium, Inc. v. FCC, 518 U.S. 727, 778 (1996) (Souter, J., concurring) (“[T]he clear

and present danger [test] of Schenk v. United States . . . evolved into the modern incitement

rule of Brandenburg v. Ohio . . . .”); see generally Comment, Staughton Lynd,

17
Brandenburg v. Ohio: A Speech Test for All Seasons?, 43 U. Chi. L. Rev. 151 (1975).

These days, then, advocacy of lawlessness retains the guarantees of free speech unless it’s

directed and likely to produce imminent lawlessness.

As a corollary, we’ve understood Brandenburg’s protection to be limited to mere or

“abstract” advocacy. Rice v. Paladin Enters., Inc., 128 F.3d 233, 243 (4th Cir. 1997); cf.

Brandenburg, 395 U.S. at 447–48 (“[T]he mere abstract teaching of the moral propriety

. . . [of] a resort to force and violence[] is not the same as preparing a group for violent

action and steeling it to such action.” (cleaned up)). Speech taking some form “other than

abstract advocacy,” by contrast, such as that which “constitutes . . . aiding and abetting of

criminal conduct,” doesn’t implicate the First Amendment under our Rice decision. See

128 F.3d at 239, 242–43 (holding that the publication of a Hit Man: A Technical Manual

for Independent Contractors, whose detailed and concrete instructions on “how to murder

and become a professional killer” assisted a man in taking three lives, wasn’t protected

abstract advocacy); see also Williams, 553 U.S. at 299–300 (suggesting that Brandenburg

only protects “abstract advocacy”). In other words, Rice effectively recognizes a second

category of unprotected speech inherent in that of incitement, which may be proscribed

without regard to whether it’s directed and likely to produce imminent lawlessness.

With this delineation in mind, we consider whether the Anti-Riot Act encompasses

the sort of advocacy that Brandenburg “jealously protects.” See Rice, 128 F.3d at 262. 6

6
Because we agree with the parties that our overbreadth analysis revolves around
the contours of protected advocacy under Brandenburg, we decline the Free Expression
Foundation’s invitation, as amicus supporting the defendants, to analyze the Anti-Riot Act

18
 C.

We find it useful to begin our analysis of the Anti-Riot Act by breaking § 2101(a)

down into the four essential elements of a violation, which are:

(1) “travel[ing] in . . . or us[ing] any facility of interstate commerce”;

(2) “with intent” either to a) “incite”; b) “organize, promote, encourage,
participate in, or carry on”; c) “commit any act of violence in furtherance
of”; or d) “aid or abet any person in inciting or participating in or carrying
on . . . or committing any act of violence in furtherance of”;

(3) “a riot”; and

(4) “perform[ing] or attempt[ing] to perform any other overt act,” for any
of the foregoing purposes, “either during the course of any such travel or
use or thereafter.”

See 18 U.S.C. § 2101(a). Stated otherwise, a violation requires two overt acts plus specific

intent to carry out one or more of numerous alternative purposes with respect to a riot.

The defendants argue that three of these elements tread on protected advocacy: (1)

the “any other” (or second, in addition to the antecedent “travel in . . . or use of any facility

of interstate commerce”) overt-act element; (2) the specific-intent element; and (3) the

definition of a “riot.” We construe the statute by focusing on each in turn.

under strict scrutiny. In that regard, we note the view of some commentators that
Brandenburg effectively operates as an even stricter stand-in for strict scrutiny when it
comes to regulating “advocacy of illegal conduct.” See, e.g., Eugene Volokh, Freedom of
Speech, Permissible Tailoring and Transcending Strict Scrutiny, 144 U. Penn. L. Rev.
2417, 2245 n.114 (1996).

19
 1.

We start with the defendants’ contention that the “any other” or second overt-act

element is overbroad because, by its plain meaning, it extends criminal consequences to

“speech and expression” (or even nonexpressive conduct) “far removed from violence,”

Defs.’ Br. at 10. In the defendants’ view, that means the statute fails to bear an adequate

relation between speech and violence under Brandenburg, which requires lawlessness to

be the likely and imminent result of speech and expression.

Appearing to agree that a straightforward reading of this element to require only “a

step toward” one of the purposes set forth under § 2101(a)(1)–(4) would pose overbreadth

problems, the government urges us to take after our sister circuit by construing it to require

the actual “fulfillment” of one or more of these purposes. Cf. Dellinger, 472 F.2d at 361–

62 (“assuming” such a view). So construed, the government contends that the statute

necessitates “an adequate relation between . . . speech and action.” See id.

We disagree with the parties. In our view, the presence of an overt-act element (or

two, in fact), together with specific intent to incite or engage in a riot, simply indicates that

the Anti-Riot Act was drafted as an attempt offense, of which it bears all the classic

hallmarks, rather than a commission offense. See Martin v. Taylor, 857 F.2d 958, 961 (4th

Cir. 1988) (“An attempt crime requires specific intent to commit a crime and some overt

act which tends toward but falls short of the consummation of the crime.”); United States

v. McFadden, 739 F.2d 149, 152 (4th Cir. 1984) (“The classical elements of an attempt are

intent to commit a crime, the execution of an overt act in furtherance of the intention, and

20
a failure to consummate the crime.”). Indeed, as the indictment in this very case illustrates,

the crime described by § 2101(a) is simply that of “Travel with Intent to Riot.” J.A. 73.

The inescapable conclusion that Congress drafted the Anti-Riot Act to encompass

unconsummated attempts to incite or engage in a riot explains why, as the defendants put

it, the statute “does not criminalize rioting” alone, but also “behavior far-removed” from

rioting. Defs.’ Br. at 27. It also explains why the statute’s overt-act elements don’t

implicate Brandenburg: because, as with inchoate offenses generally, the overt acts

themselves—“which may be entirely innocent when considered alone,” United States v.

Fleschner, 98 F.3d 155, 159–60 (4th Cir. 1996)—serve only to establish that a defendant

specifically intended to carry out (and went far enough toward carrying out) an unlawful

“purpose,” see Meredith, 824 F.2d at 1428. 7

Recall that an inchoate offense requires proof beyond a reasonable doubt that a

defendant “intend[ed] to further an endeavor which, if completed, would satisfy all of the

elements of a substantive criminal offense.” See Salinas, 522 U.S. at 65. Accordingly, to

obtain a conviction under the Anti-Riot Act, the government must at a minimum prove

that, notwithstanding any failure of consummation, the defendant acted with specific intent

to engage in unprotected speech or conduct under § 2101(a)(1)–(4). It’s therefore with

respect to the defendant’s intended speech, as opposed to actual speech (if any), that

7
Though we’re not aware of another instance in which Congress has sought to
proscribe the attempt to engage in unprotected speech, we see no bar to such legislation, in
the same way that Congress may proscribe any other attempt to engage in unlawful conduct
(provided, of course, that the conduct falls under Congress’s limited legislative authority).

21
Brandenburg mandates the adequate relation between words and lawless action for

purposes of the Anti-Riot Act.

So framed, the central overbreadth question becomes whether any of the purposes

included in the statute’s specific-intent element implicate protected advocacy. If so, those

purposes can’t form the basis of an attempt to engage in unlawful speech, rendering

overbroad the particular way of violating the statute described thereby.

We proceed to take up this question.

2.

The defendants contend that the specific-intent element is overbroad in two ways:

(1) with respect to the plain meaning of the string of speech-related verbs under

§ 2101(a)(2); and (2) with respect to the additional meaning that many of the speech-related

verbs under § 2101(a)(1)–(4) obtain under § 2102(b). We take up each provision in turn.

i.

Because the First Amendment protects speech (the sine qua non of expression) as

opposed to mere conduct, 8 and because the purposes set forth under § 2101(a)(1)–(4)

encompass both speech- and mere-conduct-related varieties, it’s necessary to distinguish

between them. Here, we agree with the parties, as well as our sister circuit, that the

8
Of course, the First Amendment does protect expressive conduct through an
intermediate (i.e., “less demanding”) level of scrutiny under United States v. O’Brien, 391
U.S. 367, 376–77 (1968). See Texas v. Johnson, 491 U.S. 397, 406–07 (1989). But the
defendants don’t argue—and properly so, in our view—that any of the statute’s conduct-
related purposes implicate expressive conduct or, if so, fail to pass muster under O’Brien.

22
purposes implicating speech are those embodied by the verbs “incite,” “organize,”

“promote,” and “encourage” under § 2101(a)(1)–(2). See Dellinger, 472 F.2d at 361. 9

With respect to “incite” under § 2101(a)(1), we have little difficulty concluding that

this verb encompasses no more than unprotected speech under Brandenburg. Thus, in the

world of Brandenburg, “incite” most sensibly refers to speech that is directed and likely to

produce an imminent lawlessness. The other conceivable definition is the dictionary one,

which, as noted, is even narrower than Brandenburg’s because it requires lawlessness to

occur, not just be likely. So either way, § 2101(a)(1) readily comports with the First

Amendment.

Turning to § 2101(a)(2), however, we find that two verbs in the string “to organize,

promote, [or] encourage” a riot fail to bear the requisite relation between speech and

lawlessness. The loosest such relation in the bunch belongs to “encourage,” which means

simply “to attempt to persuade (someone) to do something.” See Encourage, Merriam-

Webster Unabridged, https://unabridged.merriam-webster.com/unabridged/encourage

(last accessed July 30, 2020). Speech tending to encourage a riot thus encompasses all

hypothetical efforts to advocate for a riot, including the vast majority that aren’t likely to

produce an imminent riot (even assuming they’re directed to producing a riot). Indeed,

because mere encouragement is quintessential protected advocacy, the Supreme Court has

9
While the compound verb “to aid or abet” under § 2101(a)(4) can also implicate
speech, we agree with the government that any such speech would constitute “aiding and
abetting of criminal conduct,” which doesn’t implicate the First Amendment under Rice,
see 128 F.3d at 242–43—especially since none of the statutory objects of such aiding-and-
abetting speech are themselves overbroad (including, as we explain, “incite”).
23
recognized that “[t]he mere tendency of speech to encourage unlawful acts is not a

sufficient reason for banning it” under Brandenburg. Ashcroft v. Free Speech Coalition,

535 U.S. 234, 253 (2002); see also Williams, 553 U.S. at 300 (offering the statement, “I

encourage you to obtain child pornography,” as protected advocacy). It follows that

Brandenburg protects speech having a mere tendency to encourage others to riot.

The verb “promote” occupies a similarly overinclusive position on the continuum

of relation between advocacy and action. While “promote” admits of a wide range of

meanings depending on context, we think that, in the context of an enterprise like a riot,

it’s best understood to mean “to support or encourage something,” or “to advance” or

“further something by helping to . . . introduce it.” See Promote, Encarta Webster’s

Dictionary of the English Language (2d ed. 2004); see also Promoter, Encarta Webster’s

Dictionary of the English Language (2d ed. 2004) (“a supporter or advocate of

something”); cf. Williams, 553 U.S. at 294 (defining “promote” to refer to “the act of

recommending”). These definitions indicate that “promote” refers to a comparable, and

perhaps even wider, range of riot-oriented advocacy as “encourage” in the context of

§ 2101(a)(2). It thus suffers from the same overbreadth, subsuming an abundance of

hypothetical efforts to persuade that aren’t likely to produce an imminent riot. As a result,

Brandenburg also protects speech having a mere tendency to promote others to riot.

We reject the government’s argument that “promote” is readily susceptible of a

limiting construction under Williams. In Williams, the Supreme Court found that

“promote” isn’t overbroad within the meaning of 18 U.S.C. § 2252A, which proscribes

“[c]ertain activities relating to . . . child pornography,” by relying on the distinctly

24
“transactional connotation” arising from the statutory context at issue. See 553 U.S. at

294–95. The Court reasoned that, in relation to an object (grammatically speaking) like

child pornography, promotion doesn’t refer to “abstract advocacy” protected under

Brandenburg, but rather “to the recommendation of a particular piece of purported child

pornography with the intent of initiating a transfer.” Id. at 299–300.

But that reasoning is inapposite in the context of the Anti-Riot Act, where the object

of the promotional speech—the “riot” defined under § 2102(a)—is wholly non-

transactional, and can’t materialize until a sufficient number of people are persuaded to

show up at a certain future time and place and engage in lawless conduct. In this statutory

context, we think that “promote” refers to abstract advocacy.

We likewise reject the government’s invitation to limit both “promote” and

“encourage” to advocacy that is directed and likely to produce an imminent riot. For

starters, we don’t think either verb is “readily susceptible” of such an artificial limitation.

See Stevens, 559 U.S. at 481 (cleaned up). Moreover, because advocacy that is direct and

likely to produce imminent lawlessness is already called “incitement,” the government’s

proposed course would effectively require us to read these verbs as if they each said

“incite”—the same term already found under § 2101(a)(1). That, however, “requires

rewriting, not just reinterpretation,” and we may not “rewrite a law to conform it to

constitutional requirements.” See Stevens, 559 U.S. at 481 (cleaned up).

With respect to the verb “organize,” however, we reach a different outcome. As it

pertains to an event like a riot, “organize” is readily understood to mean “to form or

establish something . . . by . . . bringing people together into a structured group,” “to

25
oversee the coordination of the various aspects of something” or “to arrange the

components of something in a way that creates a particular structure.” See Organize,

Encarta Webster’s Dictionary of the English Language (2d ed. 2004). We think speech

tending to organize a riot might thus include communicating with prospective participants

about logistics, arranging travel accommodations, or overseeing efforts to obtain weapons

needed to carry out the planned violence.

Yet as these definitions and examples indicate, speech tending to “organize” others

to riot consists not of mere abstract advocacy, but rather of concrete aid. For, by the time

speech reaches the point of organizing a riot, it has crossed the line dividing abstract idea

from material reality, even if its components must still be brought together, coordinated,

arranged, or otherwise structured into form.

In other words, speech tending to organize a riot serves not to persuade others to

engage in a hypothetical riot, but rather to facilitate the occurrence of a riot that has already

begun to take shape. Such speech comes much closer to “preparing a group for violent

action” than merely “teaching . . . the moral propriety” of violence in the abstract,

Brandenburg, 395 U.S. at 48, and may even be characterized as the sort of “aiding and

abetting of criminal conduct” that doesn’t qualify for First Amendment protection, see

Rice, 128 F.3d at 242–43. It follows that speech tending to organize a riot under

§ 2101(a)(2), unlike that of encouraging and promoting a riot, doesn’t implicate mere

advocacy of lawlessness, and may thus be proscribed without reference to Brandenburg.

26
 ii.

Turning to § 2102(b), the defendants argue that this provision, which provides an

admittedly curious gloss on the statute’s specific-intent element, is overbroad in two ways.

Since these arguments track the provision’s two clauses, we take each in turn.

The first clause of § 2102(b) provides that the terms “‘to incite a riot’, or ‘to

organize, promote, encourage, participate in, or carry on a riot’, includes, but is not limited

to, urging or instigating other persons to riot.” 18 U.S.C. § 2102(b). Like the parties, we

understand this clause to gloss two more purposes onto each subparagraph under

§ 2101(a)(1)–(4) (excepting § 2101(a)(3),“to commit any act of violence in furtherance of

a riot”). These additional purposes are “urging” and “instigating” other persons to riot.

With respect to speech “instigating” others to riot, we agree with the parties this

verb is best understood as a direct synonym for the dictionary definition of “incite”—

which, as noted, is even narrower than Brandenburg’s. See Instigate, Encarta Webster’s

Dictionary of the English Language (2d ed. 2004) (“to cause a process to start”); see also

Instigate, Merriam-Webster Unabridged, https://unabridged.merriam-

webster.com/unabridged/instigate (last accessed July 30, 2020) (“provoke, incite”). In

consequence, just as speech “instigating” others to riot seems to be already accounted for

under § 2101(a)(1), so too is it consistent with the First Amendment.

As to speech “urging” others to riot, however, we agree with the defendants that this

verb suffers from a similarly inadequate relation between speech and lawless action as

“encourage” and “promote” under § 2101(a)(2). After all, to “urge” means simply to

“encourage,” “advocate,” “recommend,” or “advise . . . earnestly and with persistence.”

27
Urge, Encarta Webster’s Dictionary of the English Language (2d ed. 2004); see also Urge,

Merriam-Webster Unabridged, https://unabridged.merriam-webster.com/unabridged/urge

(last accessed July 30, 2020) (“to present in an earnest and pressing manner” or “advocate

or demand with importunity”). And because earnestness and persistence don’t suffice to

transform such forms of protected advocacy into speech that is likely to produce imminent

lawless action, Brandenburg renders the purpose of “urging” others to riot overbroad.

The second clause of § 2102(b) provides that the terms “‘to incite a riot’, or ‘to

organize, promote, encourage, participate in, or carry on a riot’ . . . shall not be deemed to

mean the mere oral or written (1) advocacy of ideas or (2) expression of belief, not

involving advocacy of any act or acts of violence or assertion of the rightness of, or the

right to commit, any such act or acts.” 18 U.S.C. § 2102(b). Phrased in simpler terms, this

clause provides that each of these purposes under § 2101(a) shall not be deemed to

encompass the mere advocacy of ideas or beliefs not involving advocacy of violence.

The defendants argue that the last phrase of this clause, beginning with “not

involving,” is overbroad. They point out that “mere advocacy of the use of force or

violence does not remove speech from the protection of the First Amendment” in a

Brandenburg world. See NAACP v. Clairborne Hardware Co., 458 U.S. 886, 927 (1982)

(emphasis omitted). And they contend that, owing to the double-negative construction of

the second clause of § 2102(b), the final phrase must be construed as affirmatively

criminalizing mere advocacy of violence, running afoul of its protected status.

The government concedes that mere advocacy of violence is protected speech under

Brandenburg, but argues that the phrase beginning with “not involving” needn’t be read to
28
affirmatively criminalize such advocacy. In the government’s view, because the second

clause of § 2102(b) starts with “shall not be deemed,” the entire clause can be limited to

subtracting from, without adding onto, the purposes of inciting, organizing, promoting,

encouraging, participating in, and carrying on a riot. So confined, the government posits

that the second clause’s exclusion of mere advocacy of violence can be read neither to

affirmatively criminalize nor to affirmatively exempt such advocacy. And since the First

Amendment already exempts it, the Anti-Riot Act doesn’t have to.

We think the last phrase of the second clause of § 2102(b) isn’t “readily susceptible”

of the government’s proposed limiting construction. See Stevens, 559 U.S. at 481 (cleaned

up). Rather, under the familiar rule that a double negative cancels itself out, the natural

meaning of this phrase is that the purposes of inciting, organizing, promoting, encouraging,

participating in, and carrying on a riot “shall . . . be deemed to mean the mere . . . advocacy

of any act or acts of violence or assertion of the rightness of, or the right to commit, any

such act or acts.” See 18 U.S.C. § 2102(b) (emphasis added); cf., e.g., House v. Bell, 547

U.S. 518, 538 (2006) (demonstrating this familiar rule in action).

Indeed, as our sister circuit herself pointed out in considering this issue, just as “[a]

true negation of a negation is an affirmation,” so too “a careful exclusion from an

exclusion” results “in an inclusion.” Dellinger, 472 F.2d at 363. And while the Dellinger

majority opted to avoid overbreadth by assuming that such inclusion doesn’t follow with

equal force from the double negative here, we agree with the separate opinion in that case

that such a view strains common sense, and thus amounts to judicial rewriting. See id. at

412 (Pell, J., concurring in part and dissenting in part).

29
 Moreover, because Congress drafted the Anti-Riot Act against the backdrop of a

long line of cases, from Whitney to Dennis, in which mere advocacy of violence was

regularly held to be unprotected, we find it all the more likely that the exclusion found in

the final phrase of § 2102(b) means to attach criminal consequences to such advocacy, and

isn’t just indifferent to it. We therefore hold this language to be overbroad as well.

3.

The defendants’ final overbreadth argument concerns the Anti-Riot Act’s definition

of a “riot” under § 2102(a). They contend that this definition is overbroad because it

contains the clear-and-present-danger test that Brandenburg displaced from the prevailing

incitement test. The government responds that, while the clear-and-present-danger test is

no longer part of the prevailing incitement test, it’s nonetheless flexible enough that we

may construe it consistently with Brandenburg’s tightened standard.

In our view, however, § 2102(a)’s clear-and-present-danger test doesn’t relate to the

same things under the Anti-Riot Act as it did under the First Amendment. Recall that,

before being replaced by the Brandenburg test, the clear-and-present-danger test referred

to the relation between unprotected incitement and “the substantive evils that Congress has

a right to prevent”—i.e., the lawless action being incited. See Schenck, 249 U.S. at 52.

And while Brandenburg tightened the required relation between those things, it didn’t alter

the fact that the object of any unprotected incitement is simply “lawless action” in general.

See 395 U.S. at 448.

In the context of the Anti-Riot Act, the object corresponding to “lawless action”

under Brandenburg is (of course) the “riot” defined under § 2102(a). Yet the relation

30
between incitement and rioting under the statute isn’t governed by § 2102(a)’s clear-and-

present-danger test, but rather directly by the verb “incite” under § 2101(a)(1) (which, as

noted, provides the necessary relation between speech and lawless action all by itself).

To revisit § 2102(a), that provision defines two types of riot: the first based on one

or more “acts of violence,” 18 U.S.C. § 2102(a)(1), and the second based on one or more

“threats” to commit one or more acts of violence, id. § 2102(a)(2). With respect to each

type, the clear-and-present-danger test governs only the relation between the act or threat

of violence forming the core of the riotous conduct and the resulting risk of “damage or

injury” to the “property” or “person” of any other individual. See id. § 2102(a). So,

whatever the precise measure of risk required by that test, a “riot” entails at bottom an act

or a threat of violence presenting “grave danger” to others. Cf. United States v. Matthews,

419 F.2d 1177, 1180–82, 1184 (D.C. Cir. 1969) (discussing the District of Columbia’s anti-

riot statute, passed by Congress in late 1967, which defines a “riot” similarly to § 2102(a)

as a public disturbance “which by tumultuous and violent conduct or the threat thereof

creates grave danger of damage or injury to property or persons”).

We think it plain that both types of riot describe conduct that Congress had the right

to prevent in enacting the Anti-Riot Act. Indeed, regardless of any risk of bodily injury or

property damage, acts of violence against others in and of themselves constitute well-

recognized forms of unlawful conduct, finding no protection under the first or any other

amendment. As for “threats of violence,” they too “are outside the First Amendment”

under the doctrine of true threats, which “protects individuals” from even “the possibility

that the threatened violence will occur.” R.A.V. v. City of St. Paul, 505 U.S. 377, 388

31
(1992); see also Virginia v. Black, 538 U.S. 343, 359–60 (2003) (plurality opinion)

(discussing “true threats”). And we have little trouble reading “threat” under § 2102(a) to

contemplate only such true threats, which are frequently made unlawful as well.

Thus, like our sister circuit, we conclude that Congress in § 2102(a) has managed

to describe “a disorder of a type which is enough of an assault on the property and personal

safety interests of the community” that inciting, engaging in, or aiding and abetting one

“can be made a criminal offense.” See Dellinger, 472 F.2d at 360–61. Accordingly, we

discern no overbreadth in the statute’s definition of a riot.

D.

Having found that the Anti-Riot Act is overbroad vis-à-vis Brandenburg insofar as

it proscribes speech tending to “encourage” or “promote” a riot, as well as speech “urging”

others to riot or “involving” mere advocacy of violence, we turn now to consider whether

the amount of overbreadth is substantial, “not only in an absolute sense, but also relative

to the statute’s plainly legitimate sweep.” Williams, 553 U.S. at 292. We conclude that it

is.

To be sure, the Anti-Riot Act has a plainly legitimate sweep. The statute validly

proscribes not only efforts to engage in such unprotected speech as inciting, instigating,

and organizing a riot, but also such unprotected conduct as participating in, carrying on,

and committing acts of violence in furtherance of a riot, as well as aiding and abetting any

person engaged in such conduct. In other words, it encompasses just about every form of

unprotected activity in relation to a riot. And the statute’s conduct-related applications

appear to form the basis of every reported prosecution under it.

32
 Yet the Anti-Riot Act nonetheless sweeps up a substantial amount of protected

advocacy. Whereas Brandenburg removes advocacy relating to a riot from the protection

of the First Amendment only if it is directed and likely to produce an imminent riot, the

statute purports to regulate any speech tending merely to “encourage,” “promote,” or

“urge” others to riot, as well as mere advocacy of any act of violence. Altogether, these

areas of overbreadth cover the whole realm of advocacy that Brandenburg protects, and

dwarfs that which it left unprotected. Thus, while the statute may have been perfectly

consistent with the contemporary understanding of the First Amendment when it was

enacted, Brandenburg causes it to encroach substantially upon free speech.

E.

Having concluded that the Anti-Riot Act is substantially overbroad in part, we turn

at last to consider whether the overbroad portions of the statute are severable from the

constitutionally valid remainder; if so, only those portions are “to be invalidated.” See

Ferber, 458 U.S. at 769 n.24. We agree with the government that they are.

Because facial invalidation “is strong medicine” that serves “as a last resort,” id. at

769, the “normal rule” in the case of a partially unconstitutional statute is “that partial,

rather than facial, invalidation is the required course,” Free Enter. Fund. v. Pub. Co.

Accounting Oversight Bd., 561 U.S. 477, 508 (2010) (cleaned up). Indeed, the Supreme

Court has repeatedly cautioned that “whenever an act of Congress contains unobjectionable

provisions separable from those found to be unconstitutional,” it is our “duty” as a court to

“maintain the act in so far as it is valid.” Regan v. Time, Inc., 468 U.S. 641, 652 (1984)

(plurality opinion) (cleaned up); see also Seila Law LLC v. Consumer Fin. Prot. Bureau,

33
140 S. Ct. 2183, 2209 (2020) (plurality opinion) (“Generally speaking, when confronting

a constitutional flaw in a statute, we try to limit the solution to the problem, severing any

problematic portions while leaving the remainder intact.” (cleaned up)).

As the Court recently observed, the Judiciary’s “power and preference” for partial

invalidation “has been firmly established since Marbury v. Madison.” Barr v. Am. Ass’n

of Political Consultants, Inc., 140 S. Ct. 2335, 2350 (2020) (plurality opinion). From then

to the present, the Court’s cases have developed “a strong presumption of severability.”

Id.; see, e.g., Bank of Hamilton v. Dudley’s Lessee, 27 U.S. (2 Pet.) 492, 526 (1829) (“If

any part of the act be unconstitutional, the provisions of that part may be disregarded while

full effect will be given to such as are not repugnant to the constitution of the United States

. . . .”).

Thus, “[e]ven in the absence of a severability clause, the traditional rule is that the

unconstitutional provision must be severed unless the statute created in its absence is

legislation that Congress would not have enacted.” Seila Law, 140 S. Ct. at 2209 (cleaned

up). Put differently, “we must retain those portions of the [a]ct that are (1) constitutionally

valid, (2) capable of functioning independently, and (3) consistent with Congress’ basic

objectives in enacting the statute.” United States v. Booker, 543 U.S. 220, 259 (2005)

(cleaned up); see also Seila Law, 140 S. Ct. at 2209.

Several of the Court’s cases illustrate just how “surgical” we ought to be in severing

unconstitutional language from an otherwise inoffensive statute. See Barr, 140 S. Ct. at

2350–51. Consider Regan, in which the Court held that the “purpose requirement” of a

prior version of 18 U.S.C. § 504—which authorized the use of certain photographic

34
reproductions of currency (otherwise proscribed under 18 U.S.C. § 474) “for philatelic,

numismatic, educational, historical, or newsworthy purposes”—constituted an invalid

time, place, and manner regulation under the First Amendment. See 468 U.S. at 647–48,

659 (cleaned up). But finding the remainder of the statute constitutional, a five-member

majority of the Court (including Justice Stevens, who concurred in the judgment in relevant

part) found that the proper fix was to excise the “select[] words” making up the purpose

requirement, even though they formed part of “a single integrated statutory phrase” in

which they flowed directly into the words of another element. See id. at 666–67 (Brennan,

J., concurring in part and dissenting in part). 10

More recently, in Barr, a seven-member majority of the Court (including Justices

Breyer, Ginsburg, Kagan, and Sotomayor, who concurred in the judgment with respect to

severability) agreed that the government-debt exception to the Telephone Consumer

Protection Act’s robocall restriction—which the Court found also constituted an invalid

time, place, and manner regulation, see 140 S. Ct. at 2346—could be excised from the

remainder of the statute, even though it consisted of a sentence fragment appended to a

single subparagraph, see id. at 2344–45 & n.2, 2352–54; cf. 47 U.S.C. § 227(b)(1)(A)(iii).

And while the Court noted that the statute included a severability clause, see 47 U.S.C.

10
As for Justice Brennan, even his concern with such selective excision would have
been quelled if Congress had offset the purpose requirement from its surrounding provision
with the disjunctive “or.” See id. at 667–68.

35
§ 608, the Court made clear that it would have excised the government-debt exception all

the same under the general “presumption of severability,” 140 S. Ct. at 2252–53.

Applying these principles to the Anti-Riot Act, we hold that the appropriate remedy

is to invalidate no more than the language responsible for the statute’s overbreadth. That

language consists of the words “encourage,” “promote,” and “urging” under §§ 2101(a)(2)

and 2102(b), as well as the final phrase of § 2102(b), beginning with the words “not

involving” and continuing through the end of that provision. Severed accordingly, these

provisions of the statute look like this:

(a) Whoever travels in interstate or foreign commerce or uses any facility of
interstate or foreign commerce, including, but not limited to, the mail,
telegraph, telephone, radio, or television, with intent—

....

(2) to organize, promote, encourage, participate in, or carry on a riot;

....

and who either during the course of any such travel or use or thereafter
performs or attempts to perform any other overt act for any purpose specified
in subparagraph (A), (B), (C), or (D) of this paragraph[]—

Shall be fined under this title, or imprisoned not more than five years, or both.

18 U.S.C. § 2101(a)(2).

As used in this chapter, the term “to incite a riot”, or “to organize, promote,
encourage, participate in, or carry on a riot”, includes, but is not limited to,
urging or instigating other persons to riot, but shall not be deemed to mean
the mere oral or written (1) advocacy of ideas or (2) expression of belief, not
involving advocacy of any act or acts of violence or assertion of the rightness
of, or the right to commit, any such act or acts.

Id. § 2102(b).

36
 Besides these discrete instances of overbreadth, the remainder of the Anti-Riot Act

“is perfectly valid.” See Booker, 543 U.S. at 258. It’s also capable of functioning

independently and thus “fully operative without the offending” language. See Seila Law,

140 S. Ct. at 2209. After all, that language makes up only a fraction of the statute’s

specific-intent element, consisting of just two items from a menu of alternative purposes

under § 2101(a)(1)–(4), plus two additional purposes glossed onto these by way of

§ 2102(b).

Moreover, though the Anti-Riot Act’s overbroad language consists of select words

within two subsections left otherwise intact, it nonetheless lends itself to being cleanly

excised from these “surrounding” provisions. Cf. Barr, 140 S. Ct. at 2352 & n.9 (noting

that, while “it is fairly unusual for the remainder of a law not to be operative,” a statute

may occasionally be drafted such that a “surrounding or connected provision” must be

severed alongside the “offending provision”). Whereas “encourage,” “promote,” and

“urging” are each set off from their adjoining purposes by the disjunctive “or” (in addition

to commas where appropriate), the last phrase of § 2102(b) is easily dropped off from the

rest of the clause in which it appears, much like the government-debt exception severed in

Barr. The remaining statute thus makes for smooth reading.

Further, such minimal severance is consistent with Congress’s basic objective in

enacting the Anti-Riot Act. We think that objective is to proscribe, to the maximum

permissible extent, unprotected speech and conduct that both relates to a riot and involves

the use of interstate commerce. And while Congress drafted the statute to encompass the

full scope of such unprotected speech as of 1968, our partial invalidation serves only to

37
remove the discrete purposes that Brandenburg rendered overbroad, thereby trimming the

statute’s scope without altering its meaning. We thus have no doubt that, if Congress could

have foreseen the Court’s decision in Brandenburg, it would have readily preferred to enact

this appropriately narrowed version of the statute than none at all.

The defendants’ arguments against partial invalidation rely on their view that the

Anti-Riot Act is significantly more overbroad than we have found it to be, including with

respect to its second overt-act element and its definition of a riot. But while these elements

of the statute might prove difficult to sever if in fact they were overbroad, we are sure

Congress “would prefer that we use a scalpel rather than a bulldozer” to cure the much

more limited overbreadth we have identified. See Seila Law, 140 S. Ct. at 2210–11.

Accordingly, because the defendants’ overbreadth challenge leaves the bulk of the

Anti-Riot Act intact, we proceed to consider their remaining challenge to the statute.

IV.

As an alternative ground for facial invalidation, the defendants contend that the

Anti-Riot Act is void for vagueness under the Due Process Clause of the Fifth Amendment.

We disagree.

“It is a basic principle of due process that an enactment is void for vagueness if its

prohibitions are not clearly defined.” Grayned v. City of Rockford, 408 U.S. 104, 108

(1972). The vagueness doctrine therefore “requires that a penal statute define the criminal

offense with sufficient definiteness that ordinary people can understand what conduct is

38
prohibited and in a manner that does not encourage arbitrary and discriminatory

enforcement.” Kolender v. Lawson, 461 U.S. 352, 357 (1983).

These twin concerns of inadequate notice and arbitrary or discriminatory

enforcement are especially pronounced “where a vague statute abuts upon sensitive areas

of basic First Amendment freedoms” because ambiguity “inevitably lead[s] citizens to steer

far wider of the unlawful zone than if the boundaries . . . were clearly marked,” thereby

chilling protected speech. See Grayned, 408 U.S. at 109 (cleaned up). That said, “perfect

clarity and precise guidance have never been required even of regulations that restrict

expressive activity.” Williams, 553 U.S. at 304 (cleaned up).

The defendants argue that the Anti-Riot Act is unduly vague primarily with respect

to its definition of a riot under § 2102(a). Not so. In our view, the definition provides more

than the “minimal guidelines” necessary to provide a sufficient standard of conduct and

enforcement for purposes of due process. See Kolender, 461 U.S. at 358.

Recall that § 2102(a) describes two types of “riot”: one based on actual violence and

another based on a threat of violence. See 18 U.S.C. § 2102(a). Each type breaks down

into roughly four elements. An actual-violence riot consists of (1) a “public disturbance,”

(2) involving one or more “acts of violence,” (3) committed “by one or more persons” who

form part of a group “of three or more persons,” and (4) that either “result[s] in[] damage

or injury to the property . . . or . . . person of any other individual” or “constitute[s] a clear

and present danger” of such damage or injury. See id. § 2102(a)(1). Similarly, a threat-of-

violence riot consists of (1) a “public disturbance,” (2) involving one or more “threats” to

commit an act of violence, (3) committed “by one or more persons” who form part of a

39
group of “three or more persons” and have “the ability of immediate execution” of the

threat or threats, and (4) that, if executed, would either result in “damage or injury to the

property . . . or . . . person of any other individual” or constitute a clear and present danger

of such damage or injury. See id. § 2102(a)(2).

The defendants largely take issue with the term “public disturbance,” which they

contend invites “wholly subjective judgments” about the scope of proscribed conduct,

much as with statutes that the Court has voided for criminalizing “‘annoying’ or ‘indecent’”

conduct. See Williams, 553 U.S. at 306 (citing Coates v. Cincinnati, 402 U.S. 611, 614

(1971); Reno v. Am. Civil Liberties Union, 521 U.S. 844, 870–71 & n.35 (1997)). But even

assuming that a statute criminalizing mere public disturbances might be unduly vague, we

believe that the other components of § 2102(a) provide sufficient “narrowing context.” See

Williams, 553 U.S. at 306.

In particular, § 2102(a)’s requirement that the public disturbance involve either an

act or threat of violence renders the scope of proscribed conduct significantly more definite.

Indeed, because the word “violence” has a settled and objective meaning, the definition’s

violence element serves to exclude a wide range of conduct that might constitute a “public

disturbance” judged subjectively—such as “making an unnecessary or distracting noise,”

see Breach of the Peace, Black’s Law Dictionary (10th ed. 2014); or, as the defendants

hypothesize, causing a “public uproar” on Twitter, see Defs.’ Br. at 31 n.5.

In fact, because any act or threat of violence inherently constitutes a disturbance

or breach of the peace, the definition’s public-disturbance element appears in context to

mean simply that the act or threat of violence must occur in a public setting—as, for

40
instance, with each of the three rallies at which the defendants conducted their acts of

violence. So construed, the core elements of § 2102(a) leave little to the imagination.

The statute’s definition of a riot is further narrowed by § 2102(a)’s remaining

elements. Under the third, the act or threat of violence constituting the public disturbance

must be committed by someone who forms part of a group of at least three people, thereby

ensuring that more ordinary instances of violence, accomplished by less than a crowd of

three, don’t rise to the level of riotous conduct. Under the fourth, the act or threat of

violence must either cause bodily injury or property damage or create a clear and present

danger of the same, thereby excluding violence that entails an insignificant or remote risk

of harm to others.

Altogether, these elements adequately define the range of conduct that constitutes a

riot within the meaning of § 2102(a)—which, after all, differs little from definitions that

courts have upheld under similar statutes. See Matthews, 419 F.2d at 1180–82 (finding

“scant room . . . for mistaking the conduct contemplated by” the District of Columbia’s

anti-riot statute, which, as noted, defines a “riot” in similar terms); State v. Beasley, 317

So.2d 750, 752–53 (Fla. 1975) (rejecting vagueness challenge to Florida statute

incorporating common law definition of a riot to mean “a tumultuous disturbance of the

peace by three or more persons, assembled and acting with a common intent, either in

executing a lawful private enterprise in a violent and turbulent manner . . . or in executing

an unlawful enterprise in a violent and turbulent manner”).

The defendants fare no better in contending that § 2102(a) is rendered unduly vague

by its inclusion of the clear-and-present-danger test in relation to the threat of injury posed

41
by the core act or threat of violence. While the defendants point out that this test requires

an inquiry into the “imminence and magnitude,” as well as the “likelihood,” of the risk of

injury posed by the violence, see Landmark Commc’ns. Inc. v. Virginia, 435 U.S. 829, 843

(1978), they fail to show that this inquiry is any more “imprecise” than similar tests found

in many “perfectly constitutional statutes,” such as “serious potential risk” or “substantial

risk,” see Sessions v. Dimaya, 138 S. Ct. 1204, 1214 (2018) (cleaned up). Indeed, even the

Brandenburg test demands an analogous inquiry into these risk-oriented variables.

Nor is § 2102(a) unduly vague because “close cases can be envisioned” under the

clear-and-present-danger test, since “[c]lose cases can be imagined under virtually any

statute.” See Williams, 553 U.S. at 305–06. For that reason, the vagueness doctrine

demands only that we be able to discern what sort of “incriminating fact” must be

established, even if it may prove difficult to determine whether that fact “has been proved”

in some cases. Id. at 306. The clear-and-present-danger test satisfies this demand.

The defendants’ next attack on § 2102(a), which focuses on the requirement that

any threat of violence undergirding a riot be capable of immediate execution, is similarly

misguided. As with the clear-and-present-danger inquiry, determining whether a particular

threat of violence could have been carried out forthwith entails the same sort of “abstract

assessment[s] of chance,” Defs.’ Br. at 33, that the law asks judges to make all the time.

The defendants’ reliance on Johnson v. United States, 135 S. Ct. 2551 (2015) is

therefore misplaced. In Johnson, the Supreme Court found the residual clause of the

Armed Career Criminal Act void for vagueness not because it required a “judicial

assessment of risk,” but rather because it tethered such assessment “to a judicially imagined

42
‘ordinary case’ of a crime, not to real-world facts or statutory elements.” Id. at 2557;

accord Dimaya, 138 S. Ct. at 1213–14. But where, as here, the relevant qualitative standard

is assessed by reference to “real-world conduct,” the vagueness doctrine takes no offense.

Johnson, 135 S. Ct. at 2561; accord Dimaya, 128 S. Ct. at 1215–16.

The defendants next assert that the same three verbs under § 2101(a)(2) that we

discussed earlier—“to organize, promote, [or] encourage,” are unduly vague as well.

Having already excised the latter two of these verbs, we consider only “organize.” And

here, the defendants fail to demonstrate any ambiguity in this familiar term, which they

themselves ask us to read (as we do) “with the plain meaning that persons of ordinary

intelligence would assign” to it. Defs.’ Br. at 35. Instead, the defendants largely repackage

their overbreadth argument, which we have rejected on the ground that speech tending to

organize a riot doesn’t constitute protected advocacy.

Finally, the defendants posit that the Anti-Riot Act violates due process because it

doesn’t require the second overt act (the one beyond traveling in or using a facility of

interstate or foreign commerce) to concur in time with specific intent to carry out a purpose

set forth under § 2101(a)(1)–(4). But we agree with the government, as well as three other

courts, that the statute is best read to require both overt acts to coincide with the same

specific intent. See United States v. Markiewicz, 978 F.2d 786, 813 (2d Cir. 1992);

Dellinger, 472 F.2d at 393; Hoffman, 334 F. Supp. at 509. Indeed, as the element itself

provides, the act must be performed “for any purpose specified” under § 2101(a)(1)–(4).

18 U.S.C. § 2101(a) (emphasis added). We thus have little trouble concluding that this

element must be accomplished with specific intent to achieve one of those purposes.

43
 V.

So far, we have held that the Anti-Riot Act is substantially overbroad to the extent

that it proscribes the attempt to engage in speech tending to “encourage” or “promote” a

riot under § 2101(a)(2), as well as speech “urging” others to riot or “involving” mere

advocacy of violence under § 2102(b). But we have also held that the statute is severable

to the same partial extent, allowing the remainder to be left intact. Finally, we have held

that the statute isn’t void for vagueness. All that remains is to consider the appropriate

disposition of the defendants’ convictions. That disposition, we hold, is to affirm.

In arguing that their convictions must be vacated even though the Anti-Riot Act

remains largely operative, the defendants assert that the indictment and, by extension, their

guilty pleas (which invoke “Count 1 of the Indictment,” J.A. 238, 250) are premised on a

conspiracy to violate the statute as a whole, without specifying which of its alternative

purposes they conspired to (and in fact did) carry out. But it’s well-established that a

conviction under a statute that “specifies several alternative ways” to commit an offense

“will stand” as long as the record evidence suffices to prove “one or more of the means of

commission,” even if the indictment alleged “the several ways” in conjunction. United

States v. Brandon, 298 F.3d 307, 314 (4th Cir. 2002) (cleaned up); accord Turner v. United

States, 396 U.S. 398, 420 (1970); cf. Mathis v. United States, 136 S. Ct. 2243, 2249 (2016)

(describing such statutes).

That’s essentially the situation we face here, except that a few of the Anti-Riot Act’s

alternative purposes happen to be overbroad and, thus, invalid. And because the record, as

44
we explain, establishes conclusively that the defendants’ substantive offense conduct falls

under the statute’s surviving purposes, their convictions must stand.

Before accepting the defendants’ guilty pleas, the district court was required to

“determine that there [was] a factual basis” for them, Fed. R. Crim. P. 11(b)(3), which it

did by accepting the defendants’ respective Statements of Offense. In those Statements,

the defendants stipulated that the substantive offense conduct underlying their respective

conspiracy convictions consists (beyond such overt acts as traveling to rallies through

interstate commerce, conducting combat training, and buying supplies) of engaging “in

violent confrontations,” J.A. 227, which is to say “physical conflict,” J.A. 232, with

counter-protestors at each of the three rallies discussed above. Specifically, the defendants

admitted to having each (as part of an assemblage of three or more) “personally committed

multiple violent acts”—including but not limited to pushing, punching, kicking, choking,

head-butting, and otherwise assaulting numerous individuals, and none of which “were in

self-defense”—in Huntington Beach, Berkeley, and Charlottesville. J.A. 231, 236.

Such substantive offense conduct qualifies manifestly as “commit[ting] any act of

violence in furtherance of a riot” within the ordinary meaning of § 2101(a)(3), as well

“participat[ing] in” and “carry[ing] on a riot” within the ordinary meaning of

§ 2101(a)(2)—three wholly conduct-oriented purposes left unscathed by our partial

invalidation of the statute. By the same token, the defendants’ offenses have manifestly

nothing to do with speech tending to encourage, promote, or urge others to riot; mere

advocacy of violence; or any other First Amendment activity; as the district court properly

found. See Daley, 378 F. Supp. 3d at 559 (noting that the First Amendment doesn’t

45
“immunize[] violence,” even “within the broader context of a political demonstration”).

The defendants muster no argument to the contrary.

Moreover, as noted, the defendants have necessarily conceded—consistent with the

“usual judicial practice” in overbreadth cases, see Fox, 492 U.S. at 484–85; Preston, 660

F.3d at 737–38—that the Anti-Riot Act poses no constitutional concern as applied to their

own conduct. And indeed, none of the defendants’ overbreadth theories, including those

we have rejected, provide any basis for an as-applied challenge on the facts to which they

have stipulated. It follows that anything less than facial invalidation of the statute affords

the defendants no relief from their convictions. Cf. Regan, 468 U.S. at 659 (holding that

18 U.S.C. § 504 as partially invalidated wasn’t unconstitutional “as applied” to the

challenger, whose offense conduct qualified under “the remaining portions of the statute”).

* * *

For the foregoing reasons, the judgments of the district court are

AFFIRMED.

46