Software Speech

Andrew Tutt *

When is software speech for purposes of the First Amendment? This issue has taken on new life amid recent accusations that Google used its search rankings to harm its competitors. 1 Google Must Answer EU Antitrust Concerns Over Search, BBC News (May 21, 2012, 9:12 AM),; Matt Peckham, Is Google’s Social Search Anti-Competitive? FTC Complaint Looms, Time (Jan. 12, 2012), This spring, Eugene Volokh coauthored a white paper explaining why Google’s search results are fully protected speech that lies beyond the reach of the antitrust laws. 2 Eugene Volokh & Donald M. Falk, Google: First Amendment Protection for Search Engine Search Results 20 (2012), available at The white paper was commissioned by Google. Days later, Professor Volokh’s white paper was the subject of an article in the New York Times. See Noam Cohen, Professor Makes the Case that Google Is a Publisher, N.Y. Times (May 20, 2012), The paper sparked a firestorm of controversy, and in a matter of weeks, dozens of scholars, lawyers, and technologists had joined the debate. 3 See, e.g., Bruce E. Boyden, Do Video Games Dream of Electric Speech?, Marquette U.L. Sch. Fac. Blog (June 22, 2012),; Frank Pasquale, Search as Speech: Two Scenarios, Concurring Opinions (May 29, 2012, 9:53 AM),

The most interesting aspect of the positions on both sides—whether contending that Google search results are or are not speech—is how both get First Amendment doctrine only half right. On one side, those arguing that Google search results should be treated as speech robustly protected from government intervention have relied on an emerging, unbounded vision of the First Amendment that would seem to preclude the regulation of any software at all. 4 See, e.g., Mike Masnick, Speech-Via-Algorithm Is Still Speech, and Censoring It Is Still Censorship, TechDirt (June 22, 2012, 9:12 AM),; Eugene Volokh, Freedom of Speech and Information Produced Using Computer Algorithms, Volokh Conspiracy (June 21, 2012, 11:59 PM), On the other side, those who argue that Google search results are not speech rely on a definition of speech so cramped that mere conveyance of an idea via digital device might rob it of its First Amendment protections. 5 See, e.g., Tim Wu, Op-Ed., Free Speech for Computers?, N.Y. Times, (June 19, 2012), Both sides are writing in a vast and largely unexamined area of First Amendment doctrine the Supreme Court has assiduously avoided defining and that scholars have criticized for its “incoherence” and “insensitivity to obviously pertinent First Amendment considerations.” 6 Robert Post, Essay, Recuperating First Amendment Doctrine, 47 Stan. L. Rev. 1249, 1270 (1995). The criticisms of the Court’s approach are easy to explain. Almost any act that conveys information that could be meaningful to someone else could be “speech.” Even ordinary everyday objects can be speech if they are used for an expressive purpose. Sometimes a urinal is just a urinal. But a urinal in an art gallery is Speech because it’s Art. Scholars have therefore argued that since some line must be drawn, efforts to limit the scope of the word “speech” in the First Amendment should at least be rooted in some conception of the values the Amendment was designed to serve, rather than semantic arguments over whether hanging a banner, wearing an armband, or burning a flag are or are not properly considered communicative. See id. at 1253-55 & n.16.

The doctrines upon which the two sides rely emerge from the same October 2010 Supreme Court Term, an important one for defining what the First Amendment means when it says “speech.” Two cases, Brown v. Entertainment Merchants Ass’n 7 131 S. Ct. 2729 (2011). and Sorrell v. IMS Health Inc., 8 131 S. Ct. 2653 (2011). created two distinct, arguably inconsistent, regimes. Both cases mark a decisive departure from the Court’s old test, set out in Spence v. Washington, that speech is that which possesses “[a]n intent to convey a particularized message” likely to be “understood by those who viewed it.” 9 418 U.S. 405, 410-11 (1974) (per curiam); see also Texas v. Johnson, 491 U.S. 397, 404 (1989) (quoting Spence, 418 U.S. at 410-11).

But the rules these cases laid down are murky. Brown protected video games only because the Court considered them analogous to literature, a path to First Amendment protection that might imply that most software is not speech. On the other hand, Sorrell declared “information is speech,” a holding so broad and potentially far-reaching that the Court could not possibly have literally meant what it said.

In Brown, the Supreme Court struck down a California statute restricting the sale or rental of violent video games to minors. 10 131 S. Ct. at 2732. As a threshold matter, the Court had to decide whether video games were speech. Rather than reach beyond video games to software generally, the Court zeroed in on video games and held that they were speech because they communicated ideas through familiar literary devices. The Court reasoned that video games were speech because they expressed ideas in familiar ways: “Like the protected books, plays, and movies that preceded them, video games communicate ideas—and even social messages—through many familiar literary devices (such as characters, dialogue, plot, and music) and through features distinctive to the medium (such as the player’s interaction with the virtual world).” 11 Id. at 2733. This approach was also reflected in the case from the Seventh Circuit cited by the Court in its decision in Brown. See id. at 2738 (citing Am. Amusement Mach. Ass’n v. Kendrick, 244 F.3d 572, 577 (7th Cir. 2001)). In American Amusement Machine Ass’n, the Seventh Circuit Court of Appeals wrote: “We have emphasized the ‘literary’ character of the games in the record . . . . [I]f the games lacked any story line, . . . a more narrowly drawn ordinance might survive a constitutional challenge.” 244 F.3d at 579-80.

In Sorrell, the Court struck down a Vermont statute barring the sale, disclosure, and use of information about how often doctor’s chose to prescribe drugs. But the statute only barred sale and use for marketing purposes. The Court found this imposed both content and speaker based restrictions on protected speech. A threshold issue, then, was whether such information was “speech” at all or merely a commodity or product. On its way to holding that such information was protected speech the Court announced an extraordinarily expansive rule. The Court held that the “creation and dissemination of information are speech within the meaning of the First Amendment.” Wrote the Court:

Facts, after all, are the beginning point for much of the speech that is most essential to advance human knowledge and to conduct human affairs. There is thus a strong argument that prescriber-identifying information is speech for First Amendment purposes.

The State asks for an exception to the rule that information is speech, but there is no need to consider that request in this case. 12 Sorrell, 131 S. Ct. at 2667 (emphasis added).

Brown and Sorrell thus suggest conflicting answers to the question: when is software speech? Brown’s test is probably best read as defining “new speech” as that which is directly analogous in presentation and mode to “old speech.” On this view, it is because video games are like literature in how they convey their information that they cross into the “speech” category. 13 This understanding is bolstered when one considers that Justice Alito’s objection to the Court’s reasoning was that some of the games were so realistic in their depictions of gruesome violence that they crossed from speech to experience. Brown, 131 S. Ct. at 2749 (Alito, J., concurring in the judgment).  Sorrell’s test, on the other hand, that “information is speech” would seem to mean that anything that facilitates the gathering or communication of information at all, such as a computer, is fully constitutionally protected in its doing so.

Neither of these doctrines is desirable or workable as a means of deciding when a new medium of communication is speech for First Amendment purposes. In applying its new analogical test, the Supreme Court in Brown wrote that “whatever the challenges of applying the Constitution to ever-advancing technology, the basic principles of freedom of speech . . . do not vary when a new and different medium for communication appears.” 14 Id. at 2733 (quotation marks ommitted). But this statement merely highlights the test’s shortcomings. What are the “basic principles” of freedom of speech under the Brown test? The Court suggested in Brown that when software operates in a manner similar to literature it is speech. But what about when it does not? Brown’s test cannot resolve difficult questions about when a search engine or other software program entirely unique to computers should be considered speech. Nor is the Court well equipped to decide at what level of abstraction to classify categories of digital algorithms as amorphously defined as “search engines,” “word processors,” “operating systems,” and “video games.” To apply Brown, the Court would have to choose at what level of abstraction to draw the categorical lines to begin with.

As intractable as the principle set out in Brown is, Sorrell’s is even more problematic. Doesn’t a software program that dispenses information squarely obtain the constitutional protections of Sorrell, full stop? If someone writes a program that facilitates the creation and retrieval of information, how can the program not be protected under the rule that “information is speech”? It seems that if we take the Sorrell approach, no software will ever be regulable—all software will be protected speech. The famous antitrust judgment finding Microsoft liable for excluding Netscape from Windows would have come out the other way, dismissed as inconsistent with Microsoft’s First Amendment right to convey information as it wishes. Apple’s wish to exclude disfavored books from the iPad eBook reader, or banish Adobe Flash from its iPhone browser, would simply be Apple’s speech. But this approach does not make much sense. Operating Systems are general-purpose platforms for running other programs, Word Processors are general-purpose data entry and manipulation engines, and Search Engines are general-purpose data retrieval tools. Because these platforms are so far from having any specific purpose to convey any particular information, it seems absurd to imagine that we would consider them “speech” protected from all government interference by the First Amendment.

The Court should disregard both of these approaches and chart a new course with respect to software. To enact a sound information policy, the Court should neither embrace a seemingly absurd result (as Sorrell would counsel) nor look to narrow analogies (as Brown would counsel) and instead look to the broader and more difficult question of the degree to which a class or category of new media implicates the First Amendment’s core purposes. Rather than counseling greater protection from governmental interference, this may in fact suggest that the government have a freer hand in content-neutral software regulation. Software is sometimes primarily concerned with conveying ideas of the kind and in a manner that one would recognize as familiar and essential to a free society. At other times, software functions much more like a means by which data is gathered, manipulated, and relayed to and by a user and therefore difficult to think of as akin to “speech.”

Software, in other words, should be considered not for what it is or even what it says but for what it means to society to treat it like speech. Whether operating systems, search engines, and word processors are “speech” depends on the position these categories occupy within our democracy. 15 For thoughtful exegeses of somewhat similar approaches, see, for example, Post, supra note 6, at 1270-79, and 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, 50-58 (2004). Thus, videogames should be protected not because they convey information and not because they are like literature, but because they are a culturally recognized medium of expression. 16 For a thoughtful discussion of this recognition peculiarly appropriate to the online form in which this Essay appears, see Video Games as Art, Wikipedia, (last accessed Nov. 10, 2012). Whereas operating systems, word processors—even search engines—are not recognized as occupying a similar expressive position. At least not yet.

To see how this already comports with how we think about speech in the real world, one need only think of a urinal in an art gallery. What makes it “art” and therefore “speech” is a constellation of cultural phenomena that coalesce to render an otherwise intellectually inert and uninteresting object meaningful. The unit of expression—the gallery, the artist—is the source of meaning. To take the urinal from the gallery and the artist from the urinal is to take from it its claim to First Amendment protection.

Now consider videogames. An independent developer creates a game that simply flashes “this is not a game” repeatedly. While probably among the least fun “games” in the world, we nonetheless see that its very status as a videogame conveys something important about it, some indicia of expressiveness—even if the game expresses little or nothing at all. Newspapers and other media are similarly culturally contingent in their claim to First Amendment protection. When a newspaper, newsletter, book, or pamphlet is published or simply handed out on a street corner, we do not need to know what is in it—it could just as well be blank—to know that it is worthy of the First Amendment’s protection.

Operating systems, word processors, and search engines are not like that. There is no understanding of these categories—as categories—as expressively important. Bearing only nascent meaning, they lack the cultural positioning to obtain the First Amendment’s most extraordinary protections and for that reason should probably fall on the other side of the “pure speech” line. This is not to say that the Court cannot monitor software regulation more closely than other areas of general conduct regulation. The Court has a shopworn device for evaluating laws in circumstances implicating important First Amendment concerns that do not fit squarely within the First Amendment’s protections for canonical speech. The Court attaches an adjective and applies a lower standard of scrutiny. Categories like “associational speech,” “symbolic speech,” and “commercial speech,” et cetera. 17 See, e.g., Cass R. Sunstein, Democracy and the Problem of Free Speech 123 (1993); R. George Wright, Constitutional Cases and the Four Cardinal Virtues, 60 Clev. St. L. Rev. 195, 195 & n.1 (2012).

The Court can similarly create a category of “software speech” granting all software a modicum of heightened scrutiny given the important First Amendment concerns its regulation might raise. But by stopping short of calling software “speech,” entirely and unequivocally, the Court would acknowledge the many ways in which software is still an evolving cultural phenomenon unlike others that have come before it. In discarding tests for whether software is speech on the basis of its literal resemblance either to storytelling (Brown) or information dissemination (Sorrell), the Court would strike a careful balance between the legitimate need to regulate software, on the one hand, and the need to protect ideas and viewpoints from manipulation and suppression, on the other.