Harvard Law Review Harvard Law Review Harvard Law Review

Constitutional Law

Two Models of the Right to Not Speak

Introduction: “One Damned Case After Another”1×1. Eugene Volokh, The First Amendment: Problems, Cases, and Policy Arguments 546 (2001) (quoting Joseph H. Sommer, Against Cyberlaw, 15 Berkeley Tech. L.J. 1145, 1161 (2000)).

Following the revelations of foreign interference in the 2016 presidential election, several states passed election laws obliging internet platforms to disclose information about who purchased political ads on their websites.2×2. See Md. Code Ann., Elec. Law § 13-405 (West 2020); Democracy Protection Act, 2018 N.Y. Sess. Laws 358 (McKinney) (codified as amended at N.Y. Elec. Law § 14-100, -106–07, -126 (McKinney 2020)); 2018 Wash. Sess. Laws 2453. Maryland’s version of these laws quickly bowed to a First Amendment challenge.3×3. See Wash. Post v. McManus, 355 F. Supp. 3d 272, 278 (D. Md. 2019). While “[s]unlight is said to be the best of disinfectants,”4×4. Louis D. Brandeis, Other People’s Money and How the Bankers Use It 92 (1914). even the sunlight of disclosure laws aimed at preserving a healthy democracy must suffer the right to free speech. The question is when such legal compulsions violate the right to not speak.

Compelled speech has long perplexed courts and commentators. At its core, the principle of the negative speech right is simple: freedom of speech “includes both the right to speak freely and the right to refrain from speaking at all.”5×5. Wooley v. Maynard, 430 U.S. 705, 714 (1977) (first citing W. Va. State Bd. of Educ. v. Barnette, 319 U.S. 624, 633–34 (1943); and then citing id. at 645 (Murphy, J., concurring)). The Supreme Court has reiterated this principle of First Amendment law in various contexts.6×6. See, e.g., Agency for Int’l Dev. v. All. for Open Soc’y Int’l, Inc., 570 U.S. 205, 213 (2013); Knox v. SEIU, Local 1000, 567 U.S. 298, 309 (2012) (“The government may not . . . compel the endorsement of ideas that it approves.”); Rumsfeld v. Forum for Acad. & Institutional Rights, Inc., 547 U.S. 47, 61 (2006) (“[F]reedom of speech prohibits the government from telling people what they must say.”); Hurley v. Irish-Am. Gay, Lesbian & Bisexual Grp. of Bos., 515 U.S. 557, 573 (1995); Riley v. Nat’l Fed’n of the Blind of N.C., Inc., 487 U.S. 781, 796–97 (1988); Pac. Gas & Elec. Co. v. Pub. Utils. Comm’n, 475 U.S. 1, 9 (1986); Harper & Row, Publishers, Inc. v. Nation Enters., 471 U.S. 539, 559 (1985); Miami Herald Publ’g Co. v. Tornillo, 418 U.S. 241, 256 (1974). Recent examples include compelled disclosures by state abortion services7×7. See Nat’l Inst. of Family & Life Advocates v. Becerra, 138 S. Ct. 2361, 2376 (2018). as well as agency-shop arrangements.8×8. See Janus v. AFSCME, Council 31, 138 S. Ct. 2448, 2463 (2018). Individual Justices have relied on the negative speech right as a basis for exempting objectors from public accommodation obligations in longstanding civil rights laws.9×9. See Masterpiece Cakeshop, Ltd. v. Colo. Civil Rights Comm’n, 138 S. Ct. 1719, 1744–45 (2018) (Thomas, J., concurring in part and concurring in the judgment); see also Brush & Nib Studio, LC v. City of Phx., 448 P.3d 890, 896 (Ariz. 2019); Leslie Kendrick, A Fixed Star in Shifting Skies: Barnette and Civil Rights Law, 13 Fla. Int’l U. L. Rev. 729, 739–40 (2019). More broadly, compelled speech is central to today’s deregulatory First Amendment. See Charlotte Garden, The Deregulatory First Amendment at Work, 51 Harv. C.R.-C.L. L. Rev. 323, 339 (2016); Amanda Shanor, The New Lochner, 2016 Wis. L. Rev. 133, 197; Morgan N. Weiland, Expanding the Periphery and Threatening the Core: The Ascendant Libertarian Speech Tradition, 69 Stan. L. Rev. 1389, 1466–67 (2017). Indeed, the right to not speak has arisen with enough variation to raise doubts about ever identifying a theory that explains compelled speech.10×10. See Larry Alexander, Compelled Speech, 23 Const. Comment. 147, 161 (2006); Steven H. Shiffrin, What Is Wrong with Compelled Speech?, 29 J.L. & Pol. 499, 515–16 (2014) (discussing elements of compelled speech that the Court has found constitutionally suspect); Nat Stern, The Subordinate Status of Negative Speech Rights, 59 Buff. L. Rev. 847, 849 (2011) (noting that “the underlying idea [of a negative speech right] has lost much of its coherence and explanatory power”). The result has seemingly been little more than “one damned case after another.”11×11. Volokh, supra note 1, at 546; see also David B. Gaebler, First Amendment Protection Against Government Compelled Expression and Association, 23 B.C. L. Rev. 995, 1002 (1982); Leslie Gielow Jacobs, Pledges, Parades, and Mandatory Payments, 52 Rutgers L. Rev. 123, 131 (1999); Stern, supra note 10, at 905 (“[N]egative speech rights have followed an unusually unsteady trajectory.”).

When lost, first principles can help reorient.12×12. William Baude, Adjudication Outside Article III, 133 Harv. L. Rev. 1511, 1513 (2020); see Robert Post, Recuperating First Amendment Doctrine, 47 Stan. L. Rev. 1249, 1255 (1995) (“First Amendment analysis is relevant only when the values served by the First Amendment are implicated.”). This Note argues that, as a matter of first principles, the negative speech right can be understood in two distinct ways: The autonomy-influenced “speech production model” involves compelling a shift from silence to speech. The antisuppression-influenced “speech restriction model” involves compelled speech that, due to limits on the amount of speech possible, constrains what the speaker can say. The two models are implicated in different categories of compelled speech cases and track different harms. They therefore call for different doctrines tailored to their respective First Amendment values. Though not every case maps onto them, the models nonetheless bring coherence to much of the confused compelled speech precedent. More importantly, by highlighting the harms to First Amendment principles, these two models suggest a way to evaluate whether a compelled speech law survives the First Amendment.

Although the conceptual problem appears minor, it carries substantial implications. Compelled speech over the internet technically alters speech, but, as a result of the internet’s limitless nature as a speech medium,13×13. See Richard L. Hasen, Cheap Speech and What It Has Done (to American Democracy), 16 First Amend. L. Rev. 200, 202 (2018); see also Eugene Volokh, Cheap Speech and What It Will Do, 104 Yale L.J. 1805, 1819 (1995). it does not restrict opportunities to produce speech.14×14. See Jack M. Balkin, Digital Speech and Democratic Culture: A Theory of Freedom of Expression for the Information Society, 79 N.Y.U. L. Rev. 1, 7 (2004). Instead, compelled internet speech principally involves the autonomy concern that lends itself to the lesser scrutiny of the speech production model. Yet because the current doctrine treats all compelled speech that alters speech as though it also restricts speech, it requires applying strict scrutiny to all instances of compelled internet speech despite the total absence of the speech-suppression concern that justifies strict scrutiny.15×15. See Sarah C. Haan, The Post-truth First Amendment, 94 Ind. L.J. 1351, 1393–95 (2019) (discussing the difficulties of squaring traditional First Amendment doctrine with modern speech). This Note concludes by attempting to realign compelled speech doctrine with the two models of the negative speech right to show that laws like Maryland’s — currently cropping up around the country — do not actually present the First Amendment issues the current doctrine implies. They are, in fact, innocuous. This Note seeks to show why.

I. The Two Models Defined

Courts and commentators have struggled to pinpoint what it is about compelled speech that raises constitutional concerns and how courts should adjudicate those concerns.16×16. See, e.g., Shiffrin, supra note 10, at 515–16. The best attempt to make sense of compelled speech jurisprudence has come from Professor Eugene Volokh, who has shown how the Court’s cases can be categorized and distinguished to produce consistency.17×17. See Eugene Volokh, Essay, The Law of Compelled Speech, 97 Tex. L. Rev. 355 (2018) (attempting to distill compelled speech doctrine into cohesive black letter law). Though instructive as a doctrinal summary, Volokh’s article largely avoids linking that doctrine to the First Amendment values needed to explain why compelled speech is a constitutional problem at all.18×18. See id. at 357–58. Still missing, then, is a normative account for determining the scope of the compelled speech right — especially important, given that delineating categories of cases doesn’t help courts identify compelled speech problems in the first instance.

So let’s start anew, not from the compelled speech cases Volokh considers, but from the negative speech axiom itself. Doing so suggests that the negative speech right involves more than one model of compelled speech; here, as in other parts of constitutional law, we ought not to let the use of a single term obfuscate important differences.19×19. See Frederick Schauer, Categories and the First Amendment: A Play in Three Acts, 34 Vand. L. Rev. 265, 276–77 (1981). In fact, though all compelled speech derives from the negative speech right, that right lends itself to two distinct models representing two distinct approaches to compelled speech: compelled speech production and compelled speech restriction. Distinguishing the two models will help identify the relevant constitutional values implicated across various instances of compelled speech.

A. Compelled Speech Production

Intuitively, the right to free speech necessarily implicates the right to choose what not to say. The characteristic element of this negative speech right model is a compelled movement from silence to speech. A prohibition occurs as a function of the government regulation, but it is a prohibition on silence. Where one otherwise would remain silent, the government regulation compels the speaker to engage in public discourse. The speech production model does not interfere with speech that the speaker is attempting to engage in. Rather, what suffers in the case of regulations touching on speech production is the capacity for the unwilling speaker to remain silent, without any correlated effect on whether she also wants to express speech of her own.20×20. See Caroline Mala Corbin, Compelled Disclosures, 65 Ala. L. Rev. 1277, 1351 (2014). For this reason, Volokh has called the cases of compelled speech that roughly map onto the speech production model “pure speech compulsions.” Volokh, supra note 17, at 358. It is a right to be left alone by the government, protected in the privacy of one’s preference for silence.21×21. See Pub. Utils. Comm’n v. Pollak, 343 U.S. 451, 467–68 (1952) (Douglas, J., dissenting).

The original compelled speech cases follow the speech production model of the negative speech right. West Virginia State Board of Education v. Barnette,22×22. 319 U.S. 624 (1943). the original compelled speech case, followed this model: school children had no capacity to opt out of reciting the Pledge of Allegiance and saluting the flag.23×23. Id. at 626. If they could, they would have remained silent at their desks. Instead, the West Virginia regulation required them to enter public discourse, to engage in speech where they otherwise would not have done so.24×24. Id.; see also Opinions of the Justices to the Governor, 363 N.E.2d 251, 255 (Mass. 1977) (noting that a proposed bill requiring teachers to lead students in the Pledge of Allegiance would violate the teachers’ negative speech rights). Wooley v. Maynard25×25. 430 U.S. 705 (1977). likewise followed the speech production model: but for the New Hampshire law announcing the state’s motto, drivers could have remained silent as to whether they accepted the notion of “Live Free or Die.”26×26. Id. at 717 (describing the First Amendment as protecting the “right to avoid becoming the courier for [the government’s] message”). Following this model too are cases of compelled subsidies,27×27. See, e.g., Janus v. AFSCME, Council 31, 138 S. Ct. 2448, 2464 (2018). in which the speaker unwillingly supports a particular cause she would not otherwise have contributed to, and compelled hosting of either government speech or private speech through the forced use of the unwilling speaker’s property.28×28. See, e.g., PruneYard Shopping Ctr. v. Robins, 447 U.S. 74, 85–88 (1980) (distinguishing compelled hosting of government and private speech). Requiring expressive public accommodations to serve people on a nondiscriminatory basis would also fall under this category.29×29. See, e.g., Elane Photography, LLC v. Willock, 309 P.3d 53, 64 (N.M. 2013) (upholding an antidiscrimination law against a compelled speech challenge).

Despite Barnette’s language about “fixed star[s] in our constitutional constellation,”30×30. W. Va. State Bd. of Educ. v. Barnette, 319 U.S. 624, 642 (1943). the Court has been willing to uphold compelled speech deriving from the speech production model. Implicated by the inability to remain silent, the speech production model is largely one concerned about speaker autonomy.31×31. See Corbin, supra note 20, at 1298 (“A person cannot be said to be autonomous in body if forced to speak when she would rather stay silent.”). It is the right to be able to say what one wishes to say and nothing else. But since every law implicates autonomy to some degree, the Court has been more lenient unless the infringement on speaker autonomy raises additional concerns under the circumstances. The government does have some capacity to compel production of speech expressing a particular viewpoint given its need to take positions on political issues.32×32. Helen Norton, The Government’s Speech and the Constitution 203 (2019). Where the government requires adoption of another private speaker’s speech, and does so on the basis of viewpoint, that too strikes close enough to the heart of autonomy as to raise maximal constitutional suspicion.33×33. See PruneYard, 447 U.S. at 97–98 (Powell, J., concurring in part and concurring in the judgment); Abood v. Detroit Bd. of Educ., 431 U.S. 209, 235 (1977); see also Elster v. City of Seattle, 444 P.3d 590, 594–95 (Wash. 2019) (distinguishing compelled subsidy cases on the ground that Seattle’s Democracy Voucher Program did not require taxpayers to individually associate with a particular political message). And the autonomy concern is at its peak with forced confessions of loyalty to the government.34×34. See Seana Valentine Shiffrin, What Is Really Wrong with Compelled Association?, 99 Nw. U. L. Rev. 839, 854, 860 (2005) (describing the autonomy harms that stem from compulsory recitation of the Pledge of Allegiance). In other circumstances, though, compelled speech production need not trigger maximal constitutional suspicion if the law does not meaningfully infringe on speaker autonomy.35×35. See, e.g., Rumsfeld v. Forum for Acad. & Institutional Rights, Inc., 547 U.S. 47, 62 (2006); Zauderer v. Office of Disciplinary Counsel, 471 U.S. 626, 651 (1985).

B. Compelled Speech Restriction

The second model of the negative speech right involves compelled speech that restricts speech. The amount of possible speech supported by any given speech medium is often limited. Forcing someone to speak thereby forces the speaker to occupy a portion of a limited speech medium with expression that she would not otherwise have engaged in. The result is that she no longer has the room to say what she otherwise would have used the limited speech medium to say. The cases that roughly map onto this model are those that Volokh calls “[i]nterference[s] with a . . . coherent speech product.”36×36. Volokh, supra note 17, at 361. Typically, they involve speech compulsions that alter the content of speech already being made. See id. The speech compulsion functions as a speech restriction akin to the kind of content-based prohibition that the First Amendment has traditionally been understood to protect against.37×37. See Agency for Int’l Dev. v. All. for Open Soc’y Int’l, Inc., 570 U.S. 205, 218 (2013). Where the speech restriction model is implicated, the speech compulsion is a speech prohibition.38×38. In this respect, the Court spoke too broadly in Zauderer v. Office of Disciplinary Counsel, 471 U.S. 626 (1985), by identifying “material differences between disclosure requirements and outright prohibitions on speech.” Id. at 650. True, disclosures and prohibitions are often different. However, this model shows the manner in which disclosures can function as prohibitions.

The speech restriction model has been on full display in Miami Herald Publishing Co. v. Tornillo,39×39. 418 U.S. 241 (1974). Riley v. National Federation of the Blind of North Carolina, Inc.,40×40. 487 U.S. 781 (1988). and Hurley v. Irish-American Gay, Lesbian and Bisexual Group of Boston, Inc.41×41. 515 U.S. 557 (1995). In Tornillo, the Court applied strict scrutiny and invalidated a Florida right-of-reply statute that required newspapers to publish the response of a public figure about whom the newspapers had previously published criticism.42×42. Tornillo, 418 U.S. at 244, 258. In so doing, the Court relied on the notion that the limited nature of the newspaper medium meant that newspapers could publish only so much speech.43×43. Id. at 258. By compelling some speech, the law stopped the newspapers from fully expressing what they wanted to say.44×44. Id. at 256–57. Similarly, the law at issue in Riley forced professional fundraisers to use their limited contact with potential donors to share the average percentage of gross receipts that the fundraisers actually turned over to charities, which prevented them from saying what they wanted to say.45×45. See Riley, 487 U.S. at 795–97. Hurley found the same impositions: by compelling the hosts of a parade to include participants whom they would have otherwise excluded, the law both compelled expression and prohibited the hosts from saying something they otherwise would have said.46×46. See Hurley, 515 U.S. at 572–78. That is, the Court presumed a limited amount of space in the parade; admitting the challengers would require excluding other participants. See id. at 574–75.

As the foregoing suggests, the Court is highly skeptical of compelled speech restrictions and tends to treat them as a form of content-based regulation. All content-based laws are subject to strict scrutiny.47×47. See Reed v. Town of Gilbert, 135 S. Ct. 2218, 2226 (2015). That was one of the clarifications that Reed v. Town of Gilbert48×48. 135 S. Ct. 2218. brought.49×49. To be sure, content-based restrictions were often subject to strict scrutiny before Reed. See, e.g., Bd. of Airport Comm’rs v. Jews for Jesus, Inc., 482 U.S. 569, 573 (1987) (quoting Perry Educ. Ass’n v. Perry Local Educators’ Ass’n, 460 U.S. 37, 45 (1983)); cf. Eugene Volokh, Freedom of Speech, Permissible Tailoring and Transcending Strict Scrutiny, 144 U. Pa. L. Rev. 2417, 2417 (1996). Before Reed, however, the Court had suggested that a content-based law with an innocuous legislative intent might not trigger strict scrutiny — a possibility Reed removed. See Genevieve Lakier, Reed v. Town of Gilbert, Arizona, and the Rise of the Anticlassificatory First Amendment, 2016 Sup. Ct. Rev. 233, 235. The second concerned what constitutes a content-based law. The Reed Court’s answer was that laws are content-based when they classify on the basis of content.50×50. See Reed, 135 S. Ct. at 2228. Apparently neutral laws can still be susceptible to strict scrutiny as content-based regulations if they conceal a purpose to discriminate against speech based on content. Id. at 2228–29. Because compelled speech restrictions require the speaker to alter her speech content and thereby prohibit her from saying everything she wants to say, the Court generally treats laws altering coherent speech products the same as it treats content-based regulations.51×51. See Riley v. Nat’l Fed’n of the Blind of N.C., Inc., 487 U.S. 781, 795 (1988) (“Mandating speech that a speaker would not otherwise make necessarily alters the content of the speech. We therefore consider the Act as a content-based regulation of speech.”); Corbin, supra note 20, at 1283.

C. Somewhere in Between?

We have, then, two distinct models of the negative speech right upon which to base compelled speech doctrine, each with distinct justifications and harms. We also have two lines of corresponding compelled speech cases, one involving forcing silent persons to speak, the other involving interference with people already speaking. Facially, the two models seem to map neatly onto these two compelled speech categories. The reality is more complicated, mostly because distinguishing between forced speech and an interference with speech is often arbitrary.

National Institute of Family & Life Advocates v. Becerra52×52. 138 S. Ct. 2361 (2018). (NIFLA) illustrates the problem. There, the Court evaluated a First Amendment challenge to the FACT Act,53×53. Cal. Health & Safety Code Ann. §§ 123470–72 (West 2020). a California law requiring state-licensed crisis pregnancy clinics to provide patrons with information about state-sponsored reproductive health services and facilities.54×54. Id. The employees themselves were not required to inform patrons.55×55. See NIFLA, 138 S. Ct. at 2369. Rather, the clinics could simply provide state-printed brochures in their waiting rooms before proceeding to engage in the speech that reflected their ideologies and values.56×56. See id. (explaining that the government-drafted notice “must be posted in the waiting room, printed and distributed to all clients, or provided digitally at check-in”). The Court treated this compelled disclosure law as a content-based regulation and held it unconstitutional.57×57. Id. at 2378. Technically, NIFLA did not apply strict scrutiny because it found the law invalid under intermediate scrutiny. However, the Court made its ruling after finding the law to be content-based and noting that all content-based regulations trigger strict scrutiny. Id. at 2371, 2375.

Why, though? The Court could have interpreted the licensed notice requirement as a movement from nonspeech to speech, or even as a hosting of government speech. In that way, NIFLA could have gone the route of Rumsfeld v. Forum for Academic and Institutional Rights, Inc.58×58. 547 U.S. 47 (2006). (FAIR): the government could have compelled someone not speaking to serve as a platform for the government’s purely factual information.59×59. See Brief for the State Respondents at 42–47, NIFLA, 138 S. Ct. 2361 (No. 16-1140) (arguing that if the Court declined to dispose of the case under Zauderer, then FAIR should resolve it). It isn’t literally true that nobody was speaking in NIFLA, but it wasn’t literally true in FAIR either: a federal law obliged law schools to send emails, post flyers, and otherwise engage in speech that facilitated on-campus military recruiters to the same extent as the law schools were providing for other employers on campus.60×60. See FAIR, 547 U.S. at 54. In both cases, coherent speech products could be understood as individual instances of communication, with the compelled speech taking place outside the parameters of the precisely defined speech act. But the Court took opposite routes. In FAIR, the Court identified the relevant speech narrowly, as communications between law schools and prospective employers, making the compelled speech supplemental to, not an alteration of, the law schools’ speech.61×61. Id. at 61–62. Yet NIFLA took the opposite route, defining the relevant speech act as the totality of the clinics’ expressions.62×62. See NIFLA, 138 S. Ct. at 2371. The problem isn’t simply that FAIR took one approach while NIFLA took the other: it is that no principled basis seems to account for why the Court should have gone with the category of compelled speech that it chose in either case. Since the category determined the outcome in both cases, the theory of compelled speech is unintelligible if it cannot identify the category to apply.

Understanding the doctrinal distinctions between cases like NIFLA and FAIR helps sort through the Court’s decisions, but it doesn’t explain why a particular instance of compelled speech should follow one line of cases rather than the other. For that, we need something beyond mere categories. These categories would be sufficient to resolve close cases if they mapped onto the two models of the negative speech right, but many don’t. Restoring compelled speech doctrine requires recognizing that divergence.63×63. Cf. Post, supra note 12, at 1255.

II. Collapsed Categories and Divergent Harms

In the aftermath of recent First Amendment cases confirming that strict scrutiny applies to all content-based laws, commentators have warned about the capacity for the Court’s holding to sweep in similar laws that the Court might not have intended to invalidate.64×64. See, e.g., Lakier, supra note 49, at 258–59; Note, Free Speech Doctrine After Reed v. Town of Gilbert, 129 Harv. L. Rev. 1981, 1998 (2016); see also Adam Liptak, Court’s Free-Speech Expansion Has Far-Reaching Consequences, N.Y. Times (Aug. 17, 2015), https://nyti.ms/1NAPcQd [https://perma.cc/X7SZ-NKMR]. NIFLA has received a similar response to Reed. See, e.g., Erwin Chemerinsky & Michele Goodwin, Constitutional Gerrymandering Against Abortion Rights: NIFLA v. Becerra, 94 N.Y.U. L. Rev. 61, 111–12 (2019); see also Marc O. DeGirolami, The Sickness unto Death of the First Amendment, 42 Harv. J.L. & Pub. Pol’y 751, 796–97 (2019). After all, at least for the Free Speech Clause, the Court’s strict scrutiny rules are essentially per se invalidations.65×65. Only five cases have triggered strict scrutiny and survived the analysis. See Williams-Yulee v. Fla. Bar, 575 U.S. 433 (2015); Holder v. Humanitarian Law Project, 561 U.S. 1 (2010); Burson v. Freeman, 504 U.S. 191 (1992); Austin v. Mich. Chamber of Commerce, 494 U.S. 652 (1990); Buckley v. Valeo, 424 U.S. 1 (1976). For a discussion of the three earliest cases, see Volokh, supra note 49. When the Court upholds a law against a First Amendment challenge, it tends to do so by finding that a lower level of scrutiny is appropriate.66×66. See, e.g., Morse v. Frederick, 551 U.S. 393, 408–10 (2007); Garcetti v. Ceballos, 547 U.S. 410, 423, 425–26 (2006). The widespread talk of a “weaponiz[ed] First Amendment,” Janus v. AFSCME, Council 31, 138 S. Ct. 2448, 2501 (2018) (Kagan, J., dissenting), often gives the impression that First Amendment protection has uniformly ballooned. See, e.g., Genevieve Lakier, The First Amendment’s Real Lochner Problem, 87 U. Chi. L. Rev. (forthcoming 2020) (manuscript at 65) (on file with the Harvard Law School Library); Robert Post & Amanda Shanor, Adam Smith’s First Amendment, 128 Harv. L. Rev. F. 165, 181–82 (2015); Shanor, supra note 9, at 137; Weiland, supra note 9, at 1393. Given that the Court does often uphold speech restrictions, the real problem seems to be the inconsistency with which it grants or withholds strong free speech rights. The level of scrutiny that attaches is therefore the principal determinant of how a First Amendment claim will be decided.67×67. See Ashutosh Bhagwat, Purpose Scrutiny in Constitutional Analysis, 85 Calif. L. Rev. 297, 305 (1997). That standard of scrutiny is justified only to the extent that it actually tracks the First Amendment values implicated by the law.68×68. See generally David A. Strauss, The Ubiquity of Prophylactic Rules, 55 U. Chi. L. Rev. 190, 207 (1988) (noting that constitutional doctrine consists of “prophylactic rule[s]” that serve as proxies for constitutional values and principles). Current compelled speech doctrine tends to apply the presumptive unconstitutionality rule reserved for content-based regulations whenever laws alter the content of speech.

However, the Court is wrong to treat all content alteration as per se invalid. As this Part will show, the two models of the negative speech right serve as metrics for the relevant First Amendment concerns, helping to identify when compelled speech warrants heightened scrutiny and when it does not.

A. Why Strict Scrutiny for Content-Based Laws?

Although the Court has provided several justifications for the First Amendment’s suspicion of content-based regulations, at their core the justifications derive from a concern about speech suppression. When acting as sovereign, the government may seek “to suppress unpopular ideas or information.”69×69. Turner Broad. Sys., Inc. v. FCC, 512 U.S. 622, 641 (1994). Of course, ideas might be unpopular because they carry no value or otherwise cause more harm than benefit. But the government should not be making that determination through “invidious discrimination of disfavored subjects.”70×70. City of Cincinnati v. Discovery Network, Inc., 507 U.S. 410, 424 n.19 (1993); see Ward v. Rock Against Racism, 491 U.S. 781, 791 (1989); N.Y. Times Co. v. Sullivan, 376 U.S. 254, 275 (1964) (noting that, for James Madison, “the censorial power is in the people over the Government, and not in the Government over the people” (quoting 3 Annals of Cong. 934 (1794)). The ideas that a government is likely to disfavor, no doubt, are those that pose a threat to the government. The concern about censorship is thus a concern about the government insulating itself: if the government attempts to prescribe orthodoxy then it will prescribe itself as that orthodoxy.71×71. See Citizens United v. FEC, 558 U.S. 310, 340 (2010); Leathers v. Medlock, 499 U.S. 439, 448–49 (1991). The idea is democratic. For the people to select the government, rather than the government selecting itself, the people must be able to express criticism of the government along political, social, and economic dimensions.72×72. See Sorrell v. IMS Health Inc., 564 U.S. 552, 583 (2011) (Breyer, J., dissenting); Simon & Schuster, Inc. v. Members of the N.Y. State Crime Victims Bd., 502 U.S. 105, 116 (1991). Such criticism is required in a democracy.73×73. See McCutcheon v. FEC, 134 S. Ct. 1434, 1441–42 (2014) (“[T]hose who govern should be the last people to help decide who should govern.”); Cincinnati, 507 U.S. at 420.

Not every content-based law will result in such orthodoxy. That isn’t the concern. The concern is the risk of insulation in itself, which is inherent to content regulation.74×74. See Turner, 512 U.S. at 641. This risk formed the theory on which the Reed Court rejected the Ninth Circuit’s holding that the content-based concern was the government’s intent to discriminate, the absence of which could justify lower scrutiny for content-based laws.75×75. Reed v. Town of Gilbert, 135 S. Ct. 2218, 2230 (2015); cf. Renton v. Playtime Theatres, Inc., 475 U.S. 41, 48 (1986). Intent to discriminate or not, the risk is palpable that sanctioning a nondiscriminatory content-based law could legitimate future attempts at suppressing disfavored ideas.76×76. See Hill v. Colorado, 530 U.S. 703, 743 (2000) (Scalia, J., dissenting). Strict scrutiny is therefore warranted to render content-based regulations per se invalid.77×77. See Strauss, supra note 68, at 197 (noting that the per se invalidation of content-based regulations is purposely overprotective); see also Matthew Tokson, Blank Slates, 59 B.C. L. Rev. 591, 613 (2018) (“The more that a proxy fails to encompass significant normative considerations, the less effective it will be . . . .”).

The Court has offered other justifications for its suspicion of content-based regulations, but they either collapse into the democratic concern or are otherwise incomplete. In the marketplace of ideas, content-based prohibitions circumscribe the ability of society to “preserve an uninhibited marketplace of ideas in which truth will ultimately prevail.”78×78. McCullen v. Coakley, 573 U.S. 464, 476 (2014) (quoting FCC v. League of Women Voters, 468 U.S. 364, 377 (1984)). This value seems inseparable from the democratic theory concern. After all, the model doesn’t describe much: most topics of public speech lack “truth” in the objective sense, and those that have it rarely benefit from a widespread attempt to chip away at the falsity. Thus, the model is accurate as a descriptive matter only if one understands the relevant “truth” to be that which people decide to be true.79×79. See R.H. Coase, Advertising and Free Speech, 6 J. Legal Stud. 1, 27 (1977). Still others defend the marketplace as a search for an objective theory of truth. See, e.g., Kent Greenawalt, Free Speech Justifications, 89 Colum. L. Rev. 119, 130 (1989). This Note doesn’t mean to evaluate that assertion here; it aims not to argue about organizing principles justifying freedom of speech, but rather to identify the principles that explain the Court’s content-regulation rules. Thus, the objective-truth conception of the marketplace theory has little application to political or social ideas that lack objective “truth,” and for those ideas that could have an objective truth, the model seems empirically inaccurate. See C. Edwin Baker, Scope of the First Amendment Freedom of Speech, 25 UCLA L. Rev. 964, 974–75 (1978); Daniel E. Ho & Frederick Schauer, Testing the Marketplace of Ideas, 90 N.Y.U. L. Rev. 1160, 1160 (2015). In fact, if the aim was to reach an objective truth, the prohibition on content-based restrictions, especially those that function as speaker-based distinctions, see, e.g., Sorrell v. IMS Health Inc., 564 U.S. 552, 557 (2011), seems diametrically opposed to that end. See Kathleen M. Sullivan, Discrimination, Distribution and Free Speech, 37 Ariz. L. Rev. 439, 445 (1995); Ari Ezra Waldman, The Marketplace of Fake News, 20 U. Pa. J. Const. L. 845, 866 (2018). Seen in this light, the marketplace of ideas, though perhaps epistemologically, even psychologically, valuable,80×80. See Learned Hand, Sources of Tolerance, 79 U. Pa. L. Rev. 1, 13–14 (1930). is little more than a tautology unless aimed at the value of self-government. That value lies in allowing the people to come to conclusions about political, social, and economic ideas, free from government orthodoxy.81×81. See Nat’l Inst. of Family & Life Advocates v. Becerra, 138 S. Ct. 2361, 2375 (2018) (“[T]he people lose when the government is the one deciding which ideas should prevail.”). Any other approach to this interest seems unrelated to the compelled speech context.82×82. See Gaebler, supra note 11, at 1003 (“[W]hen government compels rather than prohibits expression there is not the same conflict with the policy of preserving an opportunity for full and free discussion.”). In other words, the concern boils down once more into democratic distrust of government insulation: if truth is nothing but what the people widely accept as being true and act on in public life, then the people, and not the government, should decide the content of those political, social, and economic ideas.

Perhaps, though, the real concern underlying the per se invalidation is a speaker-based autonomy value. That is, maybe the search for truth is itself a First Amendment value. The Court has intimated as much at times, suggesting that content-based prohibitions are suspect because they inhibit “democratic self-government.”83×83. Reed v. Town of Gilbert, 135 S. Ct. 2218, 2233 (2015) (Alito, J., concurring) (citing Consol. Edison Co. v. Pub. Serv. Comm’n, 447 U.S. 530, 537 (1980)). However, the actual meaning of this phrase is opaque. If it refers to the notion that the government should not be the one dictating social values because it risks insulating the government from the people, then it is the same democracy-enhancing concern described above. If it is a more general concern about autonomy and the ability to participate freely in public life, then it fails to justify the invalidation rule applied to content-based regulations.

In short, the presumptive invalidity for content-based regulations derives from the democracy-enhancing concern about the government suppressing ideas. Whether as a means for government to insulate itself from criticism or otherwise control the dominating ideas, content-based speech prohibitions always heighten the risk of government censorship and thereby merit a rule of strict scrutiny.

B. Compelled Speech Alteration and the Speech Medium

Reorienting compelled speech around the two models of the negative speech right reveals that the Court’s rule for content-based regulations does not automatically apply to those circumstances in which the compelled speech alters the speech content.

At the heart of the Court’s content-based rule is a concern about the government suppressing speech. That concern aligns with the speech restriction model of the right to not speak. In those cases, compelling speech functions to suppress speech, and it will often do so on a content-specific basis.84×84. See, e.g., Agency for Int’l Dev. v. All. for Open Soc’y Int’l, Inc., 570 U.S. 205, 218 (2013). To the extent that compelled speech stops speakers from expressing what they wish to say, it functions as a restriction on speech precisely as though the government had directly prohibited speech.85×85. Even without an outright prohibition, compelled speech can operate as a direct burden on speech to discourage its expression, which is the same. See, e.g., Pac. Gas & Elec. Co. v. Pub. Utils. Comm’n, 475 U.S. 1, 10 (1986). In those circumstances, it makes sense to treat interferences with coherent speech products — that is, compelled alterations of speech — as speech restrictions. The concern about the government suppressing speech bears out in reality. By altering what can be said, the government is, in fact, suppressing speech.86×86. See, e.g., Miami Herald Publ’g Co. v. Tornillo, 418 U.S. 241, 256–57 (1974) (“The first phase of the penalty resulting from the compelled printing of a reply is exacted in . . . taking up space that could be devoted to other material the newspaper may have preferred to print.” Id. at 256.).

But collapsing the category of speech alterations into content-based regulations writ large ignores the substantial degree to which speech content can be altered without implicating the concerns of the speech restriction model. After all, compelled speech principally functions to prohibit speech where the speech medium is meaningfully limited. Outside of those circumstances, when the speech medium is not meaningfully limited, speech compulsions do not restrict speech.87×87. The First Amendment is often sensitive to the material nature of the speech medium. Compare Red Lion Broad. Co. v. FCC, 395 U.S. 367, 390–92 (1969) (suggesting compelled carry is less concerning for cable TV), and Turner Broad. Sys., Inc. v. FCC, 512 U.S. 622, 656–57 (1994) (permitting compelled carry for cable TV), with Tornillo, 418 U.S. at 258 (rejecting compelled carry for newspapers). Different communicative media involve different constraints so as to change the effect of speech regulation across those media. See Volokh, supra note 13, at 1844. There is no reason for compelled speech to be an exception to the ordinary tailoring of First Amendment doctrine to the regulated medium. The speaker can still say everything that she wants to say. Moreover, she can distinguish herself from the compelled speech, or present so much oppositional speech of her own that little of the compelled speech remains.88×88. See Gaebler, supra note 11, at 1003 (“Compulsion to express a particular view does not by itself preclude the opportunity to disavow whatever one has been compelled to express.”). The government simply cannot suppress disfavored ideas when the speaker is free to counter, drown out, distinguish, or undermine the compelled speech. Moreover, not every compelled speech alteration that restricts speech does so by reaching the limits of speech permitted by the medium.89×89. See, e.g., McIntyre v. Ohio Elections Comm’n, 514 U.S. 334, 348–49 (1995) (describing the harm of compulsory self-identification on political pamphlets in terms that track the production model of the negative speech right). In each instance, the question of content regulation to ask is whether the speech compulsion is restricting speech.90×90. One could argue that answering this question requires making precisely the same determination about what the relevant medium is that led to ambiguity in section I.C above. To some extent, this criticism is true, but the theory advanced here removes the significance of the latitude that remains in selecting the speech medium. Focusing on the harm in NIFLA and FAIR, for example, doesn’t permit a definition of the speech medium that warrants heightened scrutiny, even if it fails to require a precise definition of the speech medium. While the specifics of the speech medium are not the only determinant factors in answering this question,91×91. See, e.g., Agency for Int’l Dev. v. All. for Open Soc’y Int’l, Inc., 570 U.S. 205, 218–19 (2013) (finding that norms prevented speakers from correcting speech they were compelled to utter). they will often be central.92×92. Of course, not all interference with coherent speech products on limited speech mediums gets at the speech-suppression concern. For example, nothing in this Note unsettles the longstanding permissibility of health and safety warnings or “purely factual and uncontroversial” commercial disclosures. See Nat’l Inst. of Family & Life Advocates v. Becerra, 138 S. Ct. 2361, 2376 (2018). Initially, it isn’t obvious that, even for limited speech mediums, such compulsions actually restrict, rather than merely alter, speech. See Zauderer v. Office of Disciplinary Counsel, 471 U.S. 626, 651 (1985). Even if they do, the only speech they conceivably “suppress,” by rendering it unavailable, is that which is confusing or deceptive — speech entitled to less constitutional protection. See In re R.M.J., 455 U.S. 191, 201 (1982); Ohralik v. Ohio State Bar Ass’n, 436 U.S. 447, 455–56 (1978); Va. State Bd. of Pharmacy v. Va. Citizens Consumer Council, 425 U.S. 748, 772 n.24 (1976). This Note affects only compelled speech that falls outside those established exceptions to the negative speech axiom.

As such, the Court is in error when it collapses the category of interference with coherent speech products into content-based regulations, subject to strict scrutiny like other laws in that category. The First Amendment justification for the strict scrutiny rule of content-based regulations is gone from these circumstances. The problem is not simply that which is endemic to prophylactic rules93×93. See Strauss, supra note 68, at 197. — namely, that the rules will result in some false positives by invalidating laws that do not actually create the harm about which the Court is concerned.94×94. Cf. Lakier, supra note 49, at 258. As discussed above, the per se invalidation for content-based regulations has no false positives because the harm done by such laws results from the laws’ very existence.95×95. See supra p. 2368. But applying the prophylactic rule for content-based regulations to compelled speech without a meaningful limitation on the speech medium does not implicate the same First Amendment values whatsoever.96×96. See Schauer, supra note 19, at 286–87 (explaining the importance of distinguishing categories of speech that are qualitatively different in the type of harm they involve).

This is not to say that every application of the content-based presumptive unconstitutionality rule to nonrestrictive speech alteration laws will improperly invalidate those laws. The rule may invalidate nonrestrictive speech alteration laws that would have been invalidated if reviewed under a doctrine properly sensitive to the correct model of the negative speech right. The point is simply that such overlap is entirely random.

Accounting for the material conditions of the speech medium in question is particularly important for compelled internet speech.97×97. See Volokh, supra note 13, at 1844 (“The law of speech is premised on certain (often unspoken) assumptions about the way the speech market operates. If these assumptions aren’t valid for new technologies, the law may have to evolve to reflect the changes.”). The Court has recognized the importance of the internet to contemporary discourse.98×98. See Packingham v. North Carolina, 137 S. Ct. 1730, 1736 (2017). The next step will be to understand the balance of interests as manifested in internet speech, as distinct from analog speech arenas around which contemporary doctrines have been fashioned. Internet speech differs from prior speech situations in important respects.99×99. See, e.g., Danielle Keats Citron, Hate Crimes in Cyberspace 26–27 (2014) (noting the pervasiveness of internet speech). Most importantly, the internet creates a problem of audience and attention scarcity, but remedies any conceivable problem of speech scarcity.100×100. See Balkin, supra note 14, at 7; see also Mary Anne Franks, When Bad Speech Does Good, 43 Loy. U. Chi. L.J. 395, 396–97 (2012) (noting that the “sheer volume and mindlessness” of much bad internet speech has a “dilution effect” in relation to the totality of online speech, id. at 397). Today’s internet speech problems are fundamentally problems not of speech suppression but of speech surplus.101×101. See Tim Wu, Is the First Amendment Obsolete?, 117 Mich. L. Rev. 547, 554–58 (2018). The repeated claim of the remedy for bad speech being more speech102×102. See, e.g., United States v. Alvarez, 567 U.S. 709, 727–28 (2012). is actually possible on the internet as perhaps the only speech medium not restricted in speech availability.103×103. This is not to say that internet speech is either unmediated or undifferentiated in the audience it reaches. See Rebecca MacKinnon, Consent of the Networked, at xxii (2012) (noting the many ways in which governments regulate internet speech by acting on online platforms with the aim of affecting downstream user speech); Jack M. Balkin, Old-School/New-School Speech Regulation, 127 Harv. L. Rev. 2296, 2309 (2014) (same). The Court’s First Amendment analysis should not blind itself to that fact.

C. Nonrestrictive Speech Alteration

Compelled alterations of speech that fail to meaningfully restrict speech are not without their First Amendment harms. Yet the harm is not the same as when such laws do, in fact, restrict speech. In the latter circumstance, the concern about government suppressing disfavored ideas is a live one, and adopting a rule of per se invalidation for such laws is appropriate. But where no such concern applies, adopting that invalidation rule is inappropriate; it becomes a solution untethered from the applicable problem, a cure unaligned with the relevant illness.

What is the harm of nonrestrictive speech alteration, then? It is the same harm as the one befalling the speech production model of the negative speech right. One could interpret Justice Jackson’s appeal to the constitutional constellation in Barnette, which echoes the speech production model, as deriving from a suspicion that the government may compel conformity to a particular ideology.104×104. See W. Va. State Bd. of Educ. v. Barnette, 319 U.S. 624, 634 (1943). That prescription raises the same suspicion of government insulation at work as do content-based regulations. However, this explanation faces a problem: pure compelled speech raises constitutional concerns regardless of whether the speaker compelled into speech agrees or disagrees with the ideology expressed in the compelled speech. If the concern is that the government is attempting to prescribe orthodoxy, then that concern would have little relevance where the speaker already adheres to that orthodoxy. The Court has insisted that the problem with pure compelled speech does not depend on the speaker’s opposition to the compelled ideology.105×105. See, e.g., id.; cf. Wooley v. Maynard, 430 U.S. 705, 715 (1977). Rather, the point is that the speaker should be free to come to that set of ideas on her own, free from government compulsion.106×106. See Shiffrin, supra note 34, at 854 (noting the harm compelled speech does to freedom of thought and the autonomous agent’s control over her own mind).

Instead, underlying the category of pure compelled speech is the principle of speaker autonomy. Freedom of speech protects a correlative right to remain silent insofar as a compulsion to speak would mean that such speech is not done freely. That itself is the harm — the inability of the speaker to decide what to say.107×107. See Genevieve Lakier, Not Such a Fixed Star After All: West Virginia State Board of Education v. Barnette, and the Changing Meaning of the First Amendment Right Not to Speak, 13 Fla. Int’l U. L. Rev. 741, 744 (2019). Nothing is directed toward the health of a democratic polity or the search for truth, but simply the exercise of a speaker’s autonomy protecting the capacity to speak freely in the literal sense.108×108. See Richard H. Fallon, Jr., Two Senses of Autonomy, 46 Stan. L. Rev. 875, 876 (1994); Martin H. Redish, The Value of Free Speech, 130 U. Pa. L. Rev. 591, 593 (1982). The first state-based cases of compelled speech, involving employers’ service letters, derive from this principle,109×109. See, e.g., Cleveland, C., C. & St. Louis Ry. Co. v. Jenkins, 51 N.E. 811, 812–13 (Ill. 1898); Atchison, T. & S.F. Ry. Co. v. Brown, 102 P. 459, 460 (Kan. 1909). and it is the best explanation for the rest of the category as it has arisen in the Supreme Court.110×110. See, e.g., Abood v. Detroit Bd. of Educ., 431 U.S. 209, 234–35 (1977); see also 2 The Writings of James Madison 186 (Gaillard Hunt ed., 1901) (“Who does not see . . . [t]hat the same authority which can force a citizen to contribute three pence only of his property for the support of any one establishment, may force him to conform to any other establishment in all cases whatsoever?”).

Similarly, when compelled speech alters the content of speech without restricting what can be said, the applicable concern is that of harm done to the speaker’s autonomy. She no longer has the freedom to say what she wishes to say and nothing more. But the first step is to realize that this autonomy concern is indeed the implicated interest. Any overlap in outcome with the concern about government suppressing speech is purely happenstance. The task of fixing compelled speech doctrine is to recognize the variable circumstances in which the compulsion arises and that the relevant constitutional values differ across those dimensions.111×111. See Robert C. Post, Subsidized Speech, 106 Yale L.J. 151, 195 (1996).

NIFLA again demonstrates the point. The Supreme Court may not have been wrong to invalidate the FACT Act as a compelled disclosure burdening the First Amendment interests of the crisis pregnancy clinics; the clinics’ expression differed from what they would have expressed in the absence of the disclosure requirement. But the Court was wrong to imagine that the law was content-restrictive. Employees at the clinics were free to tell patrons anything they liked. The speech medium likely suffered some meaningful limitation by displaying the government’s brochures; patrons spent a limited amount of time at the clinics, and the clinics had only so much interior space to communicate their ideologies. But this limitation on what the clinics affirmatively wished to say was negligible. If the targeted speech content that California sought to suppress was the position adhered to by the crisis clinics,112×112. At oral argument, Justice Alito suggested the FACT Act had such a target, introducing a hypothetical about a law that was “neutral on its face, but . . . gee, it turns out that just about the only clinics that are covered by this are pro-life clinics.” Transcript of Oral Argument at 38, Nat’l Inst. of Family & Life Advocates v. Becerra, 138 S. Ct. 2361 (2018) (No. 16-1140), https://www.supremecourt.gov/oral_arguments/argument_transcripts/2017/16-1140_lkgn.pdf [https://perma.cc/NP3A-M98M“>https://perma.cc/NP3A-M98M”>https://perma.cc/NP3A-M98M]. one can only conclude that the state failed, as the clinics could still say whatever they wished. The FACT Act did amount to a state preference for one position with respect to abortion information, but the state can take sides on policy issues.113×113. See Norton, supra note 32, at 203. The question is whether that nonneutrality amounted to either an attempted or actual suppression of speech. The inevitable conclusion is that it did not. After abandoning the notion that content alteration is content regulation, nothing remains of the concern that justifies applying the per se invalidation of content-based regulations to California’s FACT Act.

Still, the FACT Act implicated some First Amendment harm: by forcing the clinics to provide information that they would not otherwise have provided, the law impacted the clinics’ ability to promote their ideology on their own terms. Under the Court’s precedents, disagreement with the message contained in the compelled speech is not relevant to the autonomy concern, but perhaps the autonomy infringement was more substantial than normal. Or perhaps it wasn’t and this was an ordinary, run-of-the-mill disclosure law whose observance comes with operating an institution open to the public. Yet the question of whether the FACT Act impermissibly burdened the clinics’ autonomy is distinct from whether it restricted speech. By glossing over the fact that the speech medium wasn’t meaningfully limited, the NIFLA Court applied a prophylactic rule for weeding out speech suppression where no First Amendment concern of speech suppression was viable. In other words, the case represents total incongruence between First Amendment harm and First Amendment protection.

The difference is more than academic. As explained above, the Court has granted the government greater latitude when it has sought to compel pure speech than when it has compelled content-based alterations to coherent speech products.114×114. See, e.g., Rumsfeld v. Forum for Acad. & Institutional Rights, Inc., 547 U.S. 47, 62 (2006). Of course, the Court remains suspicious of pure speech compulsions that involve requirements to affirm a particular ideology115×115. See Torcaso v. Watkins, 367 U.S. 488, 495 (1961). or that involve compelling the hosting of third-party speech in a viewpoint-discriminatory way. But where the pure speech compulsion is factual, or involves hosting the speech of another private party selected in a viewpoint-neutral manner, the Court is often willing to uphold such regulations.116×116. See Volokh, supra note 17, at 368–70. Certainly it has been more willing to inquire into the actual harm done to autonomy.117×117. See id. at 371–72.

It follows that, as with ordinary pure speech compulsions, compulsions that alter the content of a coherent speech product but that do not seriously implicate speaker autonomy should be granted greater constitutional leeway. In such circumstances, the issue is not whether the law alters the content of speech; every law involving a coherent speech product will alter the content of speech, by definition. If the changed content does not restrict or prohibit any speech, the law of content-based regulation is inapplicable. What should apply instead in such situations is the Court’s more lenient, autonomy-based approach.

III. Tomorrow’s Talkers and Compelled Speech

One reason for doctrine’s inevitable incongruence with reality is that it bears the unenviable responsibility of predicting the future.118×118. See Vincent Blasi, The Pathological Perspective and the First Amendment, 85 Colum. L. Rev. 449, 467 (1985). To the extent that the creation of constitutional doctrine resembles common law adjudication, constitutional doctrine resolves present disputes in the shadow of that which doesn’t exist yet.119×119. Cf. David A. Strauss, Common Law Constitutional Interpretation, 63 U. Chi. L. Rev. 877, 879 (1996).

Nowhere is this shifting landscape more apparent than in constitutional cases involving new technologies.120×120. See generally Ronald K.L. Collins & David M. Skover, The Death of Discourse, at xxii–xxiii (2019); M. Margaret McKeown, The Internet and the Constitution: A Selective Retrospective, 9 Wash. J.L. Tech. & Arts 133, 138 (2014). The Court has described itself as proceeding cautiously, which it takes to mean erring on the side of strict rules favoring speech.121×121. See, e.g., Packingham v. North Carolina, 137 S. Ct. 1730, 1736 (2017). But maintaining the role of speech as a defender of democracy requires not just blanket protection of speech but also recognition of the limits of protected speech. Unable to prohibit internet speech, governments have instead sought to use the limitless speech medium of the internet to combat its own vices.122×122. For a discussion of net neutrality as a case study in this phenomenon, see Geoffrey A. Manne et al., A Conflict of Visions: How the “21st Century First Amendment” Violates the Constitution’s First Amendment, 13 First Amend. L. Rev. 319, 333–58 (2014). Fake news or attempts to capitalize on invisible algorithms that create the architecture for information access cannot safely be prohibited,123×123. A growing literature on whether prohibitions on fake news are constitutional has debated this proposition at length. See generally Waldman, supra note 79, at 862–66. but perhaps they can be mitigated through transparency obligations and disclosure laws.124×124. See generally Jack M. Balkin, How to Regulate (and Not Regulate) Social Media 18–21 (Jan. 11, 2020) (unpublished manuscript), https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3484114“>https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3484114”>https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3484114 [https://perma.cc/MKA8-Y2WE] (noting the value of digital consumer protection laws). This is not to say that such compelled speech must be permissible. Yet the importance of these laws to the health of a functional modern democracy125×125. For a summary of how misinformation threatens democracy, see Hunt Allcott & Matthew Gentzkow, Social Media and Fake News in the 2016 Election, 31 J. Econ. Persp. 211, 219 (2017). at least requires a compelled speech doctrine that responsibly considers their permissibility. With proper sensitivity to implicated values, such a doctrine invalidates what is impermissible and leaves standing that which does not threaten First Amendment values but instead functions only to combat toxic speech.126×126. See generally Nathaniel Persily, Can Democracy Survive the Internet?, 28 J. Democracy 63, 67–70 (2017). Whether disclosure laws promote healthier speech is an empirical question about how people process information. See D.J. Flynn et al., The Nature and Origins of Misperceptions: Understanding False and Unsupported Beliefs About Politics, 38 Advances Pol. Psychol. 127, 142 (2017); Brendan Nyhan & Jason Reifler, Do People Actually Learn from Fact-Checking? Evidence from a Longitudinal Study During the 2014 Campaign 1, 32 (Nov. 30, 2016) (unpublished manuscript), https://www.dartmouth.edu/~nyhan/fact-checking-effects.pdf [https://perma.cc/2RVC-9WJN]. Whether disclosure laws can attempt to promote healthier speech is a constitutional question.

NIFLA did not involve the internet, but its refusal to separate the speech act from the speech medium means that its strict rule readily bears upon the analysis of compelled internet speech. And, as lower courts have begun to demonstrate, NIFLA’s identification of the relevant coherent speech product with the totality of the speech has led to use of the rule where the relevant concern will not be present. This Part explores this phenomenon as it intersects with new frontiers of internet speech. It concludes by attempting to reorient the analysis to the two models of the negative speech right.

A. The Error Demonstrated

This Note began by recounting the recent efforts states have undertaken to combat foreign interference in American elections. We now are in a position to see the Fourth Circuit’s decision in Washington Post v. McManus127×127. 944 F.3d 506 (4th Cir. 2019). as doctrinally correct, which demonstrates the doctrine’s misalignment.

The Maryland law at issue created disclosure obligations for online platforms with at least 100,000 unique monthly visitors and that published paid political advertisements.128×128. Md. Code Ann., Elec. Law § 1-101(dd-1) (West 2020). In signing the bill, the governor expressed constitutional concerns about the disclosure obligations,129×129. See 944 F.3d at 512. as did several newspapers, which challenged two provisions of the disclosure obligations under the First Amendment immediately after the law’s passage.130×130. Wash. Post v. McManus, 355 F. Supp. 3d 272, 278 (D. Md. 2019). One provision was a “publication requirement” that obliged online platforms to post information about political ads they displayed on their websites — including the identity of the purchaser, the people controlling the purchaser, and the amount paid for the ad.131×131. Id. at 282–83; Md. Code Ann., Elec. Law § 13-405(b)(6). Under the publication requirement, that information had to be kept available on the platforms’ websites for at least a year following the election.132×132. McManus, 355 F. Supp. 3d at 283 (citing Md. Code Ann., Elec. Law § 13-405(b)(3)(ii)). The other provision was an “inspection requirement” that obliged platforms to collect records on their political ad purchasers, maintain them for at least a year after the election, and make them available to the State Board of Elections upon request.133×133. Id. (citing Md. Code Ann., Elec. Law § 13-405(c)).

The Fourth Circuit invalidated both provisions as violations of the “most basic First Amendment principles” — including those pertaining to compelled speech.134×134. McManus, 944 F.3d at 523; see id. at 514. Both provisions “force[d] elements of civil society to speak when they otherwise would have refrained,” and that compulsion compromised the First Amendment.135×135. Id. at 514. In particular, by obliging the platforms to make readily available on their websites information about the purchasers of political ads, Maryland’s law “force[d] news outlets to publish certain information on their websites and,” as a result, “interfere[d] with the content of a newspaper or the message of a news outlet.”136×136. Id. at 517. Citing NIFLA, the panel applied heightened scrutiny and invalidated both provisions as unjustified alterations of speech.137×137. Id. at 518, 520–23.

NIFLA’s error runs rampant throughout the Fourth Circuit’s panel opinion. The critical misstep was to classify the entire website as the coherent speech product. After that, the case decided itself. If the relevant speech product is the entire website, then any compelled speech automatically changes the speech content and therefore constitutes a content-based regulation. The error is even more apparent than in NIFLA, as the internet lends itself to near limitless amounts of speech. The provisions therefore permitted the platform, compelled into speech, to make the disclosed information available on the website without stopping it from rendering available any other information that it wanted to share. As a result, the law raised zero concern of speech suppression, only of speaker autonomy.138×138. The panel did explain that one effect of the disclosure obligations falling on the online platforms would be to make hosting political ads more expensive as compared to ads of other content, possibly leading platforms to decline political ads altogether. Id. at 516–17. This concern could touch on the speech restriction model of the negative speech right given that burdening or chilling speech can be the same as suppressing it outright. See Watchtower Bible & Tract Soc’y of N.Y., Inc. v. Village of Stratton, 536 U.S. 150, 167–68 (2002); Hannegan v. Esquire, Inc., 327 U.S. 146, 151–52 (1946). But nothing about the panel’s eventual analysis of the First Amendment infirmity involved this concern. Indeed, the issue became apparent only after the district court had issued its opinion invalidating the law. See McManus, 944 F.3d at 516–17. But this is not the harm that justifies the content-based regulation rule of per se invalidation: it is the harm that admits of a more lenient First Amendment analysis outside of a few circumstances that render the threat to autonomy substantial.

The panel ignored this incongruence. Instead, after identifying the relevant coherent speech product as the entire website, any compelled speech necessarily altered the content of the speech and thereby triggered strict scrutiny as a content-based regulation.139×139. McManus, 944 F.3d at 517. The fact that the disclosure took place on the internet, rather than in a newspaper or a parade, made no difference to the panel.140×140. To its credit, the district court did find this fact worth a (singular) mention. Wash. Post v. McManus, 355 F. Supp. 3d 272, 300 (D. Md. 2019) (“The veritable infiniteness of cyberspace does not cure this constitutional infirmity.”). That is, it used a strict rule of per se invalidation to avoid the risk of government speech suppression because of a tangential threat to the platforms’ autonomy.141×141. McManus, 944 F.3d at 520–23.

B. Realigning Doctrine and Values

By now, the proper mode of analysis for laws like the one in McManus should be clear. The question is not whether the compelled disclosure changes the content of the speech. Unless that compulsion functions to compel silence through limitations on the speech medium, the constitutional infirmity of speech compulsions that change the content of the speech is directly proportional to the degree of harm done to speaker autonomy. Where that harm is minimal, even negligible, then the applicable First Amendment rule should account for that.

The Maryland law required online platforms to make available information pertaining to their political ad purchasers and to keep that information available for the State Board of Elections. Those compulsions affected the content of the platforms’ speech, but they did not stop the platforms from engaging in the speech that they wished to engage in. The infiniteness of the internet medium supporting the platforms’ speech meant that the law at issue did not stop the platforms from saying anything. Understanding the law as a content-based regulation subject to strict scrutiny was therefore inappropriate; any risk of the government suppressing speech or using the law as precedent for future speech suppression was foreign to the provisions.

To what extent did the publication and inspection requirements strike at the autonomy interest protected by the First Amendment? No more than minimally. They did not involve affirming a government-issued ideology142×142. Cf. W. Va. State Bd. of Educ. v. Barnette, 319 U.S. 624, 634 (1943). or supporting the speech of another private speaker on a viewpoint-discriminatory basis selected by the government.143×143. Cf. Janus v. AFSCME, Council 31, 138 S. Ct. 2448, 2463 (2018). Rather, the provisions were as innocuous in their effect on speaker autonomy as one can imagine. Once platforms posted the information to their websites, they did not need to revisit those pages or that information unless asked to do so by the State Board of Elections. Moreover, the revealed information pertained principally to the ad purchasers, with minimal expressive relevance to the platforms. Given the amount of speech that online platforms can support, the platforms could drown out their compelled disclosures or otherwise distinguish themselves from that forced speech.144×144. See generally Wu, supra note 101, at 565. A still better response would be for the platforms to just ignore the speech once made. Far from Barnette or Wooley, and even less intrusive to autonomy than FAIR, the function of the law was to make information available with a negligible effect on speaker autonomy, and given the digital medium a compulsion with less effect on speech is difficult to imagine.

Without a significant autonomy interest implicated, a court should have had no difficulty upholding the publication and inspection provisions of Maryland’s compelled disclosure law. Reaching that point, however, required recognizing just how minimal the First Amendment interest implicated was. To some extent, this orientation stands at odds with the notion of a prophylactic constitutional rule. Yet the sense in which it does not lies in the willingness to ensure not that the rule perfectly calibrates costs and benefits, but that the doctrine as a whole aims at the relevant First Amendment interest.


Not everything about the free speech landscape has changed with the internet; the internet has merely made salient what was already present.145×145. See Balkin, supra note 14, at 2 (“Instead of focusing on novelty, we should focus on salience.”); cf. Volokh, supra note 13, at 1843–44. The idea that NIFLA ought to have recognized how minimally the FACT Act implicated the First Amendment becomes self-evident when applied to the internet in cases like McManus. With that application comes the associated conclusion that the negative speech right simply carries a different significance on the internet than in a newspaper or a parade. In those contexts, compelled speech is not without constitutional infirmity, but it is a different infirmity. If compelled speech doctrine is to carry coherence, the relevant differences in the underlying speech mediums are a requisite component of the analysis. They are more than differentiating facts: they variously implicate harms according to the two models of the right to not speak.146×146. See generally Volokh, supra note 13, at 1844. Courts would do well to recognize the distinction as internet speech increasingly involves speech compulsions rather than direct speech prohibitions.147×147. See Dawn C. Nunziato, First Amendment Values for the Internet, 13 First Amend. L. Rev. 282, 313–14 (2014).