Navigating Trademark Law’s Empirical Turn

I. Introduction

Intellectual property law and policy rely on best guesses about human behavior. Copyright and patent law estimate which kind of property rights will promote the creation of works of authorship and invention, an appraisal that requires an understanding of psychology as much as economics.^[1] Trademark law also depends on psychological estimates, but of consumers rather than authors or inventors.^[2]

The dominant theoretical justification for trademark law is the reduction of consumer search costs.^[3] This normative theory can only guide trademark law to the extent that there is trustworthy data on the mental processes of the buying public. The search costs model needs reliable evidence of consumer perception to fully determine where to set trademark law’s boundaries.

As a result, an accurate reading of consumer behavior has long been the holy grail of trademark law. At the same time, it has been assumed that such a reading would never be fully realizable and the courts would just have to do their best in the face of empirical uncertainties.^[4] This Article asks what if, thanks to new techniques for studying and predicting consumer behavior, this assumption is no longer justified? From neuroscience to linguistic databases to artificial intelligence, new technologies point to a potential sea change in our understanding of the consumer. More and better information about how we perceive commercial symbols may alter the kinds of evidence used in trademark disputes as well as the doctrine crafted to resolve such disputes.

I argue below that trademark law is already beginning to recalibrate itself in response to this behavioral data in two ways. New forms of evidence about consumer perception are being advanced by scholars and practitioners as critical curatives for the guesswork long characterizing much of trademark law adjudication. At the same time, the Supreme Court has been reworking the structure of trademark law, removing or deemphasizing doctrinal constructs that are not focused on consumer perception and privileging those that are. The combination is what I refer to as the empirical turn.

There are reasons to be cautious about the empirical turn and to resist efforts to base trademark law exclusively on estimates of consumer thought. Despite the search cost theory’s dominance, trademark law operates according to several additional normative considerations not strictly dependent on consumer reaction. Trademark law is not only about promoting informational efficiencies. It also implicates free expression,^[5] competition policy,^[6] public safety,^[7] and other concerns inevitably bound up in a legal regime regulating speech in the marketplace.^[8] Sometimes these alternative considerations have been implicit, allowed to influence trademark adjudications thanks to the inherent difficulty in accurately measuring consumer thought. A turn towards more and more reliance on behavioral readings could prompt an unwitting rejection of some long-held trademark law principles.

This is not to say that empirical evidence of consumer thought is not valuable. In other work, I have championed the role new technologies can play in improving trademark law outcomes, not just in individual cases but in revising legal rules for trademark protection to better match behavioral realities. The issue lies in balancing the prospective theories behind those rules with descriptive accounts of consumer thought. Ignoring non-empirical inputs can have negative consequences. Yoking trademark law solely to empirical understandings of buyer behavior risks making consumers less self-reliant and spreading marketplace logics to noncommercial spaces. Too much emphasis on the behavioral data surfaced by new technologies could sacrifice competing considerations of competitive need and free expression.

In this Article, I describe the empirical turn in trademark law and prescribe how it should be addressed. Part II documents the turn.^[9] This is not the first time that trademark law has had to respond to new methods for understanding shopper sentiment. Still, I argue that there is something different going on today, both in the probativity of these new understandings of the consumer and in the Supreme Court’s recent closing off of trademark doctrines untethered to assessments of consumer behavior.

Part III of the Article charts a course for reconciling the empirical with the non-empirical. A completely empirical approach to trademark law would be undesirable, but so would blinding our eyes to better evidence of consumer behavior. The key is a considered balance of trademark law’s descriptive aspects with its prescriptive ones. The Article provides some suggestions—embedding doctrinal means for encouraging discourse about normative goals, maintaining epistemic humility about predicting human behavior, and borrowing non-empirical constructs from other bodies of law—for successfully maintaining this balance in the face of trademark’s empirical turn.

II. Tracing the Turn

A growing array of technologies—neuroscience, corpus linguistics, artificial intelligence—make possible new understandings of human thought. Meanwhile, as these technologies are embraced by trademark practitioners and scholars, the Supreme Court has issued a series of decisions that signal either disfavor or outright rejection of some of trademark doctrine’s traditional avenues for non-empirical inputs. This part ends by discussing potential consequences if trademark’s empirical turn continues.

A. New Tools for Studying Consumer Perception

Judges have long expressed frustration at existing means for predicting buyer behavior.^[10] Likewise, market researchers have long been confounded by focus group and survey data that ends up diverging from real-world consumer decision-making. Research subjects, when directly queried, often fail to articulate their actual reactions, either because they have difficulty verbalizing their own thought processes, they are embarrassed by those thought processes, or because they decide to tell their interlocutors what they think they want to hear. In addition, subtle changes in question wording can produce different results both in market research and trademark surveys.^[11]

Neuroscientific methods of studying consumer thought avoid some of these problems by directly recording responses to commercial stimuli as they happen without the intermediating effects of questioning and the respondent’s own decisions as to how to respond. These methods have been used to study a variety of questions at the heart of trademark doctrine. By calculating the speed at which consumers identify a brand with a particular product, researchers devised an “objective measure of trademark fame” for dilution analysis.^[12] By using fMRI data to quantify the biology of “repetition suppression” in the brain—i.e., the phenomenon whereby the brain’s response to a repeated stimulus declines over time—scientists offered a neural index of visual mark similarity that potentially could be used to help inform assessments of consumer confusion.^[13] Experiments identifying unique neural signatures for different brand reputations in the brain have been flagged as potential tools for evaluating mark distinctiveness.^[14]

Moving from the biological to the lexical, advances in computing power and data collection afford new insights into the way people use language. Generally unknown until just a few years ago, corpus linguistic analysis has been utilized by scholars on a number of legal issues, including determining the “ordinary meaning” of legal terms.^[15] In the trademark context, linguistic databases have been suggested as a means for assessing when a descriptive trademark has become distinctive,^[16] combating allegations that a once distinctive mark has become generic,^[17] and evaluating when a mark is famous enough to warrant anti-dilution protection.^[18] Corpus-based analyses are touted as superior to surveys for analyzing consumer sentiment given their comprehensive sample size and relative cost-effectiveness.^[19]

In a related but much less sophisticated manner, judges and trademark examiners are increasingly looking to tallies of search engine hits, page views, and likes on social media to determine a mark’s distinctiveness.^[20] Scholars have offered a variety of proposals for anchoring trademark doctrine in new and easily accessible technological means for recording the utterances of crowds. Google search results, it is theorized, can be swapped for judicially crafted factors for determining a mark’s relative distinctiveness as well as the likelihood of a defendant’s use causing consumer confusion.^[21]

Finally, artificial intelligence is being celebrated as a way to more accurately predict consumer behavior. Daryl Lim proposes “[a]lgorithmically determining confusion” via AI to detect “deviant” lower court decisions and to police against the disguised use of considerations—like the defendant’s intent—that he argues are extraneous to a true understanding of consumer confusion.^[22] AI is already being used by trademark attorneys to enhance the efficacy of the trademark search process.^[23] This process involves not only looking for exact matches, but identifying existing marks with confusingly similar visual, textual, or semantic content.^[24] Greater search efficacy may in turn impact trademark defenses like descriptive fair use, which already consider the probability that a search will reveal the plaintiff’s trademark.^[25]

To be clear, in trademark litigation, none of these new tools have yet displaced older methods, like surveys, for gathering data on consumer sentiment. Nor, to this point, have courts jettisoned the less empirical proxies for consumer perception they have developed over the years to evaluate mark distinctiveness and confusion. Nevertheless, it is worth taking seriously the potential consequences of integrating neuroscience, corpus linguistics, and AI into trademark law. Championing a more empirical approach to trademark law comes alongside a broader conversation about adopting a more data-driven approach for all legal decision-making. Some forecast a “legal singularity” whereby massive troves of data, predictive analytics, and digital judges will be deployed to remove all uncertainty from the legal system.^[26] In the midst of these intellectual and technological trends, on multiple fronts, the U.S. Supreme Court has closed off longstanding avenues for deciding trademark cases independent of consumer perception.

B. Closing off Non-Empirical Avenues

Exhibit A is a 2020 case involving a travel website’s attempt to federally register the mark BOOKING.COM.^[27] Generic terms—terms referring to an entire class of products or services rather than a particular producer—are not eligible for trademark registration.^[28] The U.S. Patent and Trademark Office (USPTO) determined that the term “booking” was generic for hotel reservation services, and that the addition of the “.com” suffix should not alter that determination.^[29]

Overturning the decision of the USPTO and writing for an 8–1 majority, Justice Ruth Bader Ginsburg made consumer perception the alpha and omega in determining genericness. If consumers interpreted BOOKING.COM, not as a source identifier, but as a label for an entire class of business—in this case, businesses providing online hotel reservation services—then the mark should be refused registry. But if consumers perceived the term as representing a particular business, then the USPTO could not block the registration for being generic.^[30] A survey submitted by the trademark applicant showed that three out of four respondents identified BOOKING.COM as “a brand name,” rather than a generic term—evidence deemed compelling by the courts below.^[31] Justice Ginsburg bluntly endorsed a completely empirical view of genericness: “That should resolve this case: Because ‘Booking.com’ is not a generic name to consumers, it is not generic.”^[32]

Justice Stephen Breyer penned a dissent making the case for a genericness evaluation involving more than an empirical reading of consumer thought. Justice Breyer second-guessed the evidence at hand, noting the difficulty in uncovering genericness via a survey, particularly for a well-advertised term.^[33] His main focus, however, was rejecting a definition of genericness that hinged exclusively on consumer perception. Regardless of consumer perception evidence, he said, allowing the now ubiquitous “.com” suffix to transform a generic term into a distinctive one was “[un]sound as a matter of law and logic.”^[34] Prior cases evaluating how mark distinctiveness should be determined expressed caution at merely relying on consumer sentiment, particularly given the ability of well-funded businesses to shape that sentiment via advertising.^[35] Justice Breyer worried that allowing an otherwise generic term to escape a finding of genericness with the simple addition of “.com” would endanger the “linguistic commons.”^[36] But his was a lonely voice—the other eight Justices agreed that the consumer’s view of a particular word should govern regardless of the competitive dangers of giving one business exclusive rights in a term referring to an entire product category.^[37]

One might be tempted to shrug off Booking.com as a single case on the isolated issue of genericness and top-level domains. But the Court’s other trademark decisions of the past few years have also acted to close off doctrinal channels not reliant on evidence of consumer mental state. A 2023 case took aim at the Rogers test, a mechanism developed by courts to sidestep trademark law’s likelihood of confusion analysis and its heavy (and costly for litigants) dependence on assessments of consumer perception.^[38] Under Rogers, when an expressive work is involved, so long as the unauthorized trademark usage has some “artistic relevance to the underlying work” and does not “explicitly mislead[],” the case must be dismissed without investigating the likelihood of consumer confusion.^[39]

In the case, the producer of a “Bad Spaniel[]” dog chew toy contended that Rogers should apply to plaintiff Jack Daniel’s claim of trademark infringement.^[40] The Court sided with Jack Daniel’s, however, holding that Rogers did not apply because the chew toy maker used “Bad Spaniels” as a source identifier, i.e., as a trademark. Even if a party is using another’s mark to make an “expressive comment,” the Court explained, if the use is also in part for “source identification,” then "Rogers has no proper role" and the normal likelihood of confusion analysis governs the outcome.^[41]

Relative to the Rogers test, the likelihood of confusion test is more dependent on observations about consumer behavior and more solicitous of empirical evidence. In every federal circuit, likelihood of confusion requires consideration of several different factors. Some, like mark strength and actual confusion, lend themselves to surveys of consumer thought. Others, like relatedness of goods, rely on mere proxies for that thought.^[42] But, in total, the multifactor confusion analysis is designed to obtain a read of shopper behavior.^[43] The analysis can even be framed somewhat mathematically as a determination as to whether consumer confusion is “probable.”^[44] By contrast, when substituted for the likelihood of confusion test, the Rogers defense is less an empirical investigation of consumer thought and more a proactive carve out engineered to protect artistic expression.^[45] Survey evidence demonstrating confusion is insufficient to show that a defendant’s use is “explicitly misleading” under Rogers because the test is about “the nature of the behavior of the defendant, not the impact of the defendant’s use.”^[46]

Hence, as in Booking.com, the Jack Daniel’s decision makes consumer perceptions paramount, portending future trouble for trademark doctrines designed to downplay those perceptions in favor of other normative goals beyond reducing search costs.^[47] Even though the Court stressed that its decision was “narrow” in that it did not proscribe all applications of the Rogers defense,^[48] it did not champion Rogers either, explaining “we take no position” on Rogers’s application in other circumstances not involving use as a source identifier.^[49] Responding to the argument that it had neutered a key doctrinal innovation meant to safeguard free speech, the Court explained that “[w]hen a mark is used as a mark (except, potentially, in rare situations), the likelihood-of-confusion inquiry does enough work to account for the interest in free expression.”^[50]

Even as the Court downplayed free speech concerns to restrict availability of the Rogers defense, in other recent cases, it mobilized those same concerns to prune additional legal pathways also reflecting considerations beyond consumer confusion.^[51] The Lanham Act bars various kinds of symbols from federal trademark registration.^[52] Some provisions of the Act are calibrated to prevent consumer confusion and depend on assessments of consumer perception. For example, § 2(d) of the Act disallows registration of marks that are likely “to cause confusion, or to cause mistake, or deceive” with regard to a previously used or registered mark.^[53] Other registration provisions do not necessarily hinge on obtaining an accurate read of consumer perception. Section 2(a) contains provisions prohibiting the registration of marks that are “immoral,” “disparag[ing],” or “scandalous.”^[54] Section 2(c) bars marks consisting of “a name, portrait, or signature identifying a particular living individual” without their consent as well as marks including “the name . . . of a deceased President of the United States during the life of his widow.”^[55]

In 2017, the Court took aim at the “disparaging” bar, which had been construed for decades to apply to marks denigrating racial and religious minorities.^[56] It deemed the bar invalid under the First Amendment.^[57] Two years later, it invalidated the “scandalous” bar on the same grounds.^[58] The Court viewed these bars as too remote from the “central purpose” of trademark law and, hence, lacking in sufficient justification for their abridgement of speech.^[59]

These decisions may be legally correct under established First Amendment precedent. I flag them here to illustrate how their logic adds fuel to trademark law’s empirical turn. It may have been clear to the Court that Congress acted ultra vires in making considerations of morality and protection of minorities part of the trademark registration process, but, to others, such considerations are an important and longstanding part of trademark law.^[60] Notably, these particular doctrinal pathways were not created by the USPTO or judges, but explicitly enshrined by Congress in federal law.^[61] Nevertheless, the Court did not feel the need to defer to legislative authority, instead electing to further narrow trademark law’s focus to the question of consumer confusion.^[62] Other registration bars untethered to consumer perception will likely suffer a similar fate.^[63]

C. Consequences for Trademark Law’s Prescriptive Goals

Trademark doctrine has long reflected a balance between descriptive and prescriptive components. Trademark law concerns itself not only with reducing consumer search costs, but also with prospectively removing competitive barriers, developing consumer capabilities, and quarantining some territories from the full force of market logic. These are aspirations for a better future rather than an acceptance of current conditions. The Court’s recent decisions make trademark law more reactive to consumer behavior and less proactive in shaping that behavior.

Technologies that make evidence of consumer sentiment more reliable, combined with the Court’s closing off of avenues for alternate considerations, tilt the balance away from prescriptive goals towards a strictly descriptive analysis of consumer behavior. First, the empirical turn could reshape the playing field for commercial actors, solidifying the market power of established interests and raising the barriers to entry for newcomers. Justice Breyer flagged the “anticompetitive consequences” of rubber-stamping evidence of consumer association with a mark without attention to other inputs.^[64] Some scholars have made the same point, warning that tethering eligibility for trademark protections solely to consumer perception will exacerbate pre-existing competitive advantages.^[65] Whether it is BOOKING.COM or something else, businesses can spend their advertising dollars to build an association in people’s heads between a particular trade symbol and a commercial enterprise.^[66] To some degree, these self-interested efforts should positively influence decisions to award trademark protection. If marketing has made a brand have a particular resonance for consumers, this also means that that particular resonance can be deployed in harmful ways by unscrupulous commercial rivals.^[67] If trademark law did not exist at all, commercial actors would be hamstrung in their ability to communicate the availability of new and better products and markets would become less competitive.^[68]

But the engineered reality of advertising should not be the sole input in awarding trademark rights.^[69] Some parts of trademark law need to take into account the competitive effects of trademark protections, regardless of current mental associations between consumers and a particular producer. This need has become all the more apparent as recent research shows that, rather than there being an endless supply of equally effective potential trademarks, psychologically effective trademarks are a finite and declining resource.^[70] If trademark rights are awarded only on the basis of consumer perception, there is the danger of a first-mover advantage for the firms that claim rights in a handful of descriptive terms that best convey a product class’s desired qualities.^[71] There is little doubt that trademarks can both help and hinder competitive markets and that trademark law must balance policing against consumer deception with avoiding anticompetitive blockades.^[72] The concern is that too much focus on empirical proof of secondary meaning and confusion will upset this balance.^[73]

Second, rather than merely describing consumer perception, trademark law actively shapes that perception.^[74] Indeed, the very term “consumer” is legally constructed.^[75] As noted, the reduction of search costs serves as the preeminent justification for modern trademark law, but, until now, the difficulty in measuring every instance of consumer confusion meant that the law tolerated some moments of cognitive consternation in the buying public.^[76] New technologies hold the potential for detecting even fleeting experiences of confusion.^[77] But just because some varieties of consumer thought can be measured that does not mean those measurements should dictate legal outcomes. Graeme Dinwoodie advocates for an explicitly proactive view of the role of trademark law in shaping consumer capabilities.^[78] Along similar lines, Fred Yen notes the benefits of ensuring that a certain amount of consumer confusion is non-actionable: “At the very least, modest confusion concerning trademarks actually helps consumers avoid confusion by helping them develop valuable cognitive skills that make distinguishing and understanding trademarks possible.”^[79] Altering the current balance of empirical and non-empirical inputs—whether from accepting additional evidence of consumer perception beyond surveys or by removing doctrines instantiating concerns beyond consumer perception—could change trademark law’s role in actively shaping the mentality of consumers, perhaps not for the better.

Finally, trademark law, despite the current dominance of the search costs rationale, has always resisted application of market logic to certain precincts, characterizing particular representations as inappropriate for the marketplace, regardless of their appeal to consumers. In 1909, the USPTO refused to register a mark for GROVER CLEVELAND cigars, despite the former President’s willing endorsement, because “[t]o use the names of ex-Presidents of the United States as trademarks tends to detract from the dignity of the high office which they have held.”^[80] The primary drafter of the Lanham Act, Edward Rogers, offered a similar rationale for the current § 1052(c) prohibition on marks containing the name of a President during the life of their widow: “The idea of prostituting great names by sticking them on all kinds of goods is very distasteful to me.”^[81] Hence, the historical record shows a willingness to use trademark law not to simply ratify consumer sentiments but to construct a vision of the ideal marketplace by preventing commercialism from infiltrating certain territories. Perhaps concerns with the “dignity of the high office” seem antiquated in a time where “selling out,” even by Presidents, no longer has any meaning.^[82] My point in highlighting the bar on registering presidential trademarks is that, regardless of whether there is a harm to be addressed here or not, this is another example of an aspect of trademark law agnostic to evidence of consumer perception that seems headed for the Court’s chopping block.^[83]

III. Rules for Navigating the Turn

Despite the foregoing, it would be a mistake to simply ignore new means for measuring consumer behavior. A clearer window into human perception can make trademark law stronger, replacing ill-founded hunches and folk wisdom with verifiable data on how people actually interface with commercial symbols. Empirical research observing consumer behavior can not only make individual adjudications more accurate, but also prompt doctrinal revisions better aligned with how consumers actually perceive commercial symbols. At the same time, too much reliance on this data to decide trademark disputes risks overlooking other key considerations. The ultimate concern is that embracing empirical results to determine the content of legal terms like “distinctiveness,” “confusion,” and “fame” could lead to path dependence as other things we once interrogated when attempting to define those terms lose their relevance.

This is not to say that accepting some empirical results will automatically lead to a cascade of future acceptances swamping trademark law’s non-empirical supports. But it does mean that we should be on the lookout for such a cascade and should preemptively embed non-empirical off-ramps into the structure of trademark law. Technological inputs, though often framed in neutral terms, can generate effects beyond their initial planned use.^[84] Like other areas of the law, trademark doctrine needs to balance descriptive accounts with prescriptive guidelines. Exactly how to strike this balance is open to question, but, at the least, the empirical turn must not be allowed to close off trademark law’s proactive role in shaping economic conditions and preserving space for free expression. Here I offer four general principles for engineering those off-ramps.

A. Preserving Avenues for Non-Empirical Discourse

What makes people obey the law? This has been a subject of debate among philosophers for centuries. One answer, particularly in modern times, highlights the way law is constructed and communicated to its subjects. Jürgen Habermas locates law’s authority in the tension between law’s “facticity” and its “validity.” Facticity refers to law’s status as social fact, as “coercive state power” that “Holmes’s ‘bad man’” has to take into account in considering his behavior.^[85] Validity refers to law’s underlying claim for legitimacy, that it is worthy of being followed. According to Habermas, in modern society, without the same obedience to tradition and religion as in the past, this legitimacy can only be conferred through “discourse.” By offering reasons—and criticisms that are ultimately accepted or trumped by those reasons—for legal rules, legal actors transform the facticity of their edicts into something more.^[86]

Though empirical evidence can be marshalled in this discursive process, empirical claims are not the same as the legitimating reasons advanced for legal rules. Habermas posits that reasons can be redeemed through discourse whereas empirical claims are simply taken as they are.^[87] “[A]ll [legal] norms are inherently indeterminate,” says Habermas.^[88] In his view, the process of making and applying the law demands this indeterminacy, as it is in this process where “competing value orientations and interest positions” are rationally balanced.^[89]

A danger then of the empirical turn is that it short-circuits this discourse, potentially weakening public acceptance of the law. It is true one can have a discourse about empiricism, perhaps arguing in favor of the legitimacy of binding trademark law to empirical claims. But often empirical evidence is accepted as is without investigating its proper role.^[90] Dinwoodie notes that even though empirical evidence can militate in favor of a particular agenda—for example, the preservation of current consumer understandings—it is frequently cast in neutral terms with its normative motivations hidden.^[91] In its Booking.com decision, the Supreme Court favored an empirical showing of genericness while declining to acknowledge any other rationale for how genericness should be defined. Tersely insisting “[b]ecause ‘Booking.com’ is not a generic name to consumers, it is not generic” is an attempt to evade discourse rather than engage in it.^[92] If one accepts Habermas’s view, there need to be outlets for discourse within trademark doctrine to give that doctrine legal legitimacy and not just facticity.

One mechanism for facilitating the discourse essential to law’s persuasive authority is to construct legal definitions that explicitly require both empirical and non-empirical evaluations. Take, for example, the Eighth Amendment’s prohibition on “cruel and unusual punishments.”^[93] In determining whether a criminal punishment is “unusual,” courts have sought out empirical answers—for example, counting up acts of state legislatures to determine whether a particular punishment fits within a national consensus or is an outlier.^[94] But the word “cruel” suggests a non-empirical analysis as well. After all, a punishment could be unusual (e.g., being tickled with a feather) but not fit anyone’s definition of cruel. To figure out what is cruel, concerns not susceptible to empirical evaluation like “the dignity of man” or the proportionality of the punishment to the severity of the underlying crime, must be interrogated.^[95] The text of the Eighth Amendment defines prohibited punishments in a way that implicates both empirical and non-empirical inputs.

We might similarly insist on definitions for key trademark legal terms that blend the empirical with the non-empirical.^[96] In Booking.com, the majority’s operative definition of genericness was a term that “names a ‘class’ of goods or services, rather than any particular feature or exemplification of the class,” adding that “the relevant meaning of a term is its meaning to consumers.”^[97] Such a cramped definition allowed the majority to evaluate genericness based solely on evidence of consumer understanding and evade Justice Breyer’s call for discourse around the normative function of genericness. ^[98]

An alternative definition designed to preserve space for the non-empirical considerations at stake in genericness could take a page from the Eighth Amendment. This definition might instruct legal actors investigating genericness to consider consumer perception and competitive consequences. For example, a generic term could be defined as one that is competitively necessary to describe a product characteristic or one whose primary significance to the relevant public is the genus of which the particular product is a species.^[99] Admittedly, evaluating when a term is “competitively necessary” will invite debate and may only be relevant in particular circumstances. But this debate may produce the kind of discourse that legitimates trademark law’s role in balancing consumer protection with competitive need.

B. Valuing Vagueness

Trademark scholars, myself included, tend to criticize parts of trademark law for their vagueness. A common and valid concern is that vagueness in trademark doctrine can prompt extortionate attempts to pressure defendants to abandon use of particular terms, with potentially damaging consequences to free speech and competition.^[100] The empirical turn promises to replace vagueness with precision, as quantifiable and reliable metrics for consumer thought can now inform determinations of distinctiveness and confusion and judicial moves are made to remove trademark rules not dependent on such information. This may seem like an unmitigated good: empirical evidence not only improves our understanding of consumer thought but makes the law more precise and, therefore, more successful in guiding the desired behavior of commercial actors.^[101]

This analysis overlooks the benefits of vagueness, however. Though too much vagueness can allow arbitrary enforcement of legal rules, so can too much precision. Legal philosopher Timothy Endicott gives the example of a legal rule setting out the requirements for commencing a criminal trial.^[102] A concrete time limit (e.g., two months) would be plenty of time for preparing a shoplifting case but not enough time for a complex securities fraud matter. Insisting on a precise legal rule in this instance would generate arbitrary results; trial dates would be set regardless of the complexity of the case and, hence, regardless of the normative purpose behind the rule.

In other circumstances, legal precision should be highly valued. For something like a determination of eligibility to vote, lawmakers need a clear-cut rule to appear officially indifferent to different capacities for voting. A vague standard for voting eligibility (e.g., sufficient “maturity”) could lend itself to governmental corruption. “The challenge for law-makers,” Endicott says, “is to determine whether, in a given scheme of regulation, the arbitrariness resulting from precision is worse than the arbitrariness resulting from the application of a vague standard.”^[103]

To the extent trademark law is concerned with judgments of consumer psychological capacity, lawmakers need to weigh the value of precision in such judgments against the value of vagueness. In contrast to questions of voter capacity, we might conclude that vagueness has value for evaluating genericness if that doctrine is meant to implicate concerns beyond consumer perception.^[104] Only defining genericness according to whether a term is a generic name to consumers^[105] offers greater precision but handcuffs judges from considering the larger anticompetitive or expressive effects of granting trademark protection.^[106]

We may also find value in vagueness when examining the current legal assessment of likelihood of confusion. To determine likelihood of confusion, courts consider a collection of different factors like mark strength and proximity of goods, but these factors are not “rigidly weighed.”^[107] Moreover, even individual factors can contain multiple components and potentially serve multiple normative purposes.^[108] Judges applying the intentionally vague factors for assessing likelihood of confusion remain free to tailor their decisions to issues of free expression and competitive need, which are surely at least part of the normative basis for trademark regulation.^[109] In this way, a vague standard for infringement can be more faithful to the norms behind trademark law than a precise standard.^[110]

Of course, vagueness delegates power to judges and judicial discretion can generate its own arbitrariness. Again, however, the comparison to voting eligibility rules is instructive. We would not want to delegate broad discretion in determinations of voting capacity to the judiciary, but Congress has chosen to use common law decision-making to construct our system of trademark regulation.^[111] A strictly empirical approach to determining the presence of actionable confusion would ultimately take power away from judges engaging in common law adjudication. Though there are definitely things in the judge-made list of factors for evaluating trademark infringement to quibble about, the vagueness in these features is a feature and not a bug. Hence, one response to the empirical turn is to maintain the relative vagueness of the confusion factors and avoid the temptation to only highlight particular factors—like actual confusion, mark similarity, and mark strength—that lend themselves to empirical evidence of consumer capacity.^[112]

C. Borrowing from Other Subject Areas

Even as trademark doctrine becomes more devoted to quantifiable measures of consumer behavior, it can take steps to mimic allied bodies of law that are not so in debt to such measures. Take the doctrine of copyright misuse. Under the doctrine, a copyright holder’s abusive conduct can serve as an equitable bar to the enforcement of its copyright.^[113] For example, one copyright holder’s abusive licensing practices—in effect, forbidding a licensee of its copyrighted software from having any of its workers work on alternative (and potentially superior) software—ran so counter to the goals of copyright law as to forfeit the holder’s opportunity to enjoin the licensee’s infringing conduct.^[114] Copyright law borrowed its doctrine of misuse from patent law, but unlike in patent law, copyright misuse has been expanded beyond anticompetitive, antitrust-related violations to a broader constellation of disfavored conduct, including copyright overclaiming that restricts socially valuable speech.^[115]

Though courts have declined the invitation to create a corresponding doctrine of trademark misuse,^[116] the empirical turn makes further consideration of such borrowing warranted. Inter-regime doctrinal borrowing needs to be done thoughtfully and with attention to relevant historical, doctrinal, and public policy differences, but it can also offer valuable opportunities for doctrinal development.^[117] A trademark misuse doctrine could provide an off-ramp from the empirical turn in select situations where other values besides the reduction of search costs should be paramount. Perhaps more than some other legal subject areas, vigorous enforcement of trademark rights can be self-reinforcing as successful trademark suits reshape consumer perceptions of the frequency and likelihood of trademark ownership, which in turn makes future findings of confusion more likely.^[118] Hence, a defense against the assertion of trademark rights with an improper purpose might serve as an important counter against the inertia generated by motivated litigants and judicial receptivity to their empirical evidence of confusion.

Fleshing out the details of a trademark misuse defense would be a task for common law adjudication.^[119] One suggestion, in keeping with some efforts to enforce copyright misuse, has been to invoke trademark misuse when the trademark holder improperly invokes their property interest to stifle free speech.^[120] A trademark misuse defense would arguably be harder for the Court to dismantle than the Rogers test given the longstanding acceptance of such a defense in patent and copyright law. A misuse defense could also be deployed to facilitate competition interests that might otherwise be shortchanged in the infringement analysis.^[121] For example, a potential licensee engaging in unauthorized use of the licensor’s trademarked term might defend an infringement claim by contending that the terms of the proposed license violated public policy by requiring the licensee to never compete with the licensor, even after expiration of the license.^[122] Trademark misuse arguments could become more salient in situations where confusion does not occur at the point of purchase and there may be greater procompetitive benefits from the defendant’s behavior, such as when a generic drug manufacturer advertises its product as comparable yet cheaper than a name brand.^[123]

Another potential source for doctrinal borrowing is right of publicity law. Right of publicity doctrine is resolutely anti-empirical.^[124] Instead, the scope of publicity rights is largely determined by statutory carve outs and common law doctrines meant to safeguard competing interests beyond a rightsholder’s interest in controlling commercial exploitation of their identity.^[125] While few would argue that right of publicity jurisprudence is without its faults, trademark law can still take note of its doctrinal scaffolding, constructed outside the totalizing effects of the search costs justification.^[126] For example, robust defenses against publicity rights infringement exist for so-called “transformative” and “newsworthy” uses of someone’s persona. Both of these defenses are meant to provide space for expressive content that is considered socially valuable enough to outweigh a celebrity’s property interest.^[127] Given the Supreme Court’s own hostile signals and the lower courts’ broad interpretation of the Jack Daniel’s decision,^[128] a new safe harbor for expressive uses of a trademark that has already been judicially validated yet avoids the traditional, factually-intensive likelihood of confusion analysis might be warranted.

Both of these borrowing suggestions involve defenses, which may be the most appropriate location for non-empirical doctrinal safeguards. Conflating different analytical tasks in a single validity or infringement determination can be cognitively difficult, particularly for jurors unfamiliar with the different aims and concepts animating trademark law. Stand-alone defenses designed to separately address competition or free speech interests may be the best way to ensure that these interests do not receive short-shrift in the face of empirical evidence of consumer perception.

D. Practicing Epistemic Humility

Although this Article largely assumes that new technologies for studying consumer behavior are more accurate than what came before, this does not mean that they should be treated as providing a perfect window into such behavior. Past attempts to reshape trademark law through new means for elucidating consumer thought offer a cautionary tale.

In the early twentieth century, psychologists began to critique the manner in which courts were deciding trademark cases. Judges, in their view, were too enraptured by their own idiosyncratic perceptions of commercial stimuli. Their solution—ultimately accepted by the courts—was introducing expert-run studies of consumer behavior using “laboratory methods” into evidence.^[129]

The psychologists’ approach might sound like a neutral proposition: better evidence will lead to better outputs in trademark cases. Their argument aligned with the views of Progressive-era legal reformers, who supported replacing the moralizing about property rights sometimes evident in trademark opinions of the era with scientific evidence of human behavior. But, rather than being value neutral, this move altered trademark law in favor of mark owners as psychologist-led surveys tended to support stronger trademark rights.^[130]

One need not go back an entire century to realize that research into consumer psychology can sometimes rely on flawed understandings. Past demonstrations of consumer reactions once considered sufficient to justify new legal regulations on commercial behavior subsequently have been revealed as frauds.^[131] Recent experiments now cast doubt on prior research supposedly rationalizing trademark law’s anti-dilution provisions.^[132]

In light of this checkered history, the final modest suggestion I would make is that we should be rigorous in testing new evidence supporting particular characterizations of human behavior for its accuracy and that, even when the evidence survives such testing, we should not be hesitant to revise our judgments in the face of new data. Those of us responsible for theorizing and designing trademark law should adopt a fallibilistic theory of the truth of consumer behavior. Empirical knowledge claims will always be socially and politically contingent. This does not mean that one has to be a complete relativist, refusing to ever place one’s faith in empirical evidence of consumer perception. Rather, the fallibilistic perspective insists that a truth today may be revealed to no longer be a truth in some subsequent period.^[133] By being amenable to future review, courts can not only avoid too much reliance on inaccurate data, but also open themselves to renewed consideration of non-empirical inputs once an empirical foundation is found to be shakier than first realized.

IV. Conclusion

New technologies for measuring consumer perception promise to make our system for regulating the use of commercial symbols more effective. Enhanced understanding of consumer behavior facilitates assessments of which symbols designate source to consumers and which unauthorized uses of those symbols will confuse the public. But this Article also warns of the need for caution and proactive moves to insulate some aspects of trademark law from the empirical turn. Strictly calibrating the law to behavioral measures risks neglecting normative considerations beyond the reduction of search costs, like competitive need and free expression.

In an older, less empirical time for trademark jurisprudence, the marketplace was viewed as a spiritual proving ground where it was incumbent upon consumers to build character by navigating tempting and sometimes untrustworthy commercial representations.^[134] This nineteenth-century approach to trademark law reflected a very different balance than today, one that perhaps focused too intently on a particular vision of a chastening commercial sphere and not enough on measurable consequences to its participants. If carried too far, today’s empirical turn threatens the opposite, a balance that relies too heavily on what is measurable and not enough on a proactive vision of what the marketplace should be. Trademark law should always have room for both.

---

1. See, e.g., Stephanie Plamondon Bair, The Psychology of Patent Protection, 48 Conn. L. Rev. 297, 314–29 (2015) (describing past psychological interpretations of patent enforcement); Mark A. Lemley, Faith-Based Intellectual Property, 62 UCLA L. Rev. 1328, 1333 (2015) (noting psychological studies exploring why people create and the particular effects of financial incentives); Erez Reuveni, Copyright, Neuroscience, and Creativity, 64 Ala. L. Rev. 735, 738 (2013) (discussing debate over how to best design copyright law to foster creativity and the role of psychology in that debate).

2. Mishawaka Rubber & Woolen Mfg. Co. v. S.S. Kresge Co., 316 U.S. 203, 205 (1942) (“The protection of trade-marks is the law’s recognition of the psychological function of symbols.”).

3. See Daniel M. McClure, Trademarks and Competition: The Recent History, Law & Contemp. Probs., Spring 1996, at 13, 32–33; Mark P. McKenna, A Consumer Decision-Making Theory of Trademark Law, 98 Va. L. Rev. 67, 75 (2012).

4. See Societe Anonyme de La Grande Distillerie v. Julius Wile Sons & Co., 161 F. Supp. 545, 547 (S.D.N.Y. 1958) (“[Likelihood of confusion] does not readily lend itself to resolution by scientific appraisement or comparison. Rather, a finding on infringement is by necessity a subjective determination by the trial judge based on his visceral reactions . . . .”); Jeremy N. Sheff, Marks, Morals, and Markets, 65 Stan. L. Rev. 761, 810 (2013) (referring to the “inescapable empirical uncertainty” surrounding questions of consumer reaction in trademark law).

5. RiseandShine Corp. v. PepsiCo, Inc., 41 F. 4th 112, 120 (2d Cir. 2022) (discussing impact on free speech of extending trademark rights to different kinds of marks).

6. Platinum Home Mortg. Corp. v. Platinum Fin. Grp., Inc., 149 F.3d 722, 726 (7th Cir. 1998) (“[T]rademark protection should not interfere with the traditional policies of a competitive market, and courts have generally recognized that the public substantially benefits from competition.”).

7. David A. Simon, Trademark Law & Consumer Safety, 72 Fla. L. Rev. 673, 711, 739 (2020) (emphasizing need to reform trademark law when a product, particularly pharmaceuticals, presents a heightened risk of physical harm).

8. Michael S. Mireles, Jr., Towards Recognizing and Reconciling the Multiplicity of Values and Interests in Trademark Law, 44 Ind. L. Rev. 427, 450–54 (2011).

9. I am not the first person to suggest there has been a move to more empiricism in trademark doctrine and scholarship. See Graeme B. Dinwoodie, Trade Mark Law as a Normative Project, 2023 Sing. J. Legal Stud. 305, 316–19, 322–23 (2023). What I hope to contribute here is a particular description of that trend as well as a blueprint for how to preserve trademark law’s non-empirical considerations while still benefiting from better insights into consumer behavior.

10. See, e.g., Kraft Foods Grp. Brands LLC v. Cracker Barrel Old Country Store, Inc., 735 F.3d 735, 741 (7th Cir. 2013); Triangle Publ’ns, Inc. v. Rohrlich, 167 F.2d 969, 976 (2d Cir. 1948).

11. See Min Kyoung Kim, Note, Trademark Law Relies on Consumers, Consumers Rely on What?: Trademark Law and its Heavy Reliance on Consumer Perception, 50 AIPLA Q.J. 749, 762–68 (2022).

12. Suneal Bedi & Mike Schuster, Towards an Objective Measure of Trademark Fame, 54 U.C. Davis L. Rev. 431, 434, 459, 461–62, 488 (2020).

13. Mark Bartholomew et al., A New Addition to the Trademark Litigator’s Tool Kit: A Neuroscientific Index of Mark Similarity, 113 Trademark Rep. 789, 800–01, 805 (2023).

14. Sandra M. Virtue & Darren S. Cahr, Trademarks and the Brain: Neuroscience and the Processing of Non-Literal Language, 112 Trademark Rep. 695, 699–700 (2022).

15. Thomas R. Lee & Stephen C. Mouritsen, Judging Ordinary Meaning, 127 Yale L.J. 788, 828–30, 829 n.176 (2018).

16. Alexandra J. Roberts, Mark Talk, 39 Cardozo Arts & Ent. L.J. 1001, 1023–25 (2021).

17. Neal A. Hoopes, Reclaiming the Primary Significance Test: Dictionaries, Corpus Linguistics, and Trademark Genericide, 54 Tulsa L. Rev. 407, 415–20 (2019); see also James A. Heilpern et al., Generic Ab Initio, 70 Buff. L. Rev. 613, 660–61 (2022) (contending that corpus linguistics “can provide concrete answers to the plaguing question of how to determine whether a term is generic at the time a party adopts it”).

18. Jake Linford & Kyra Nelson, Trademark Fame & Corpus Linguistics, 45 Colum. J.L. & Arts 171, 191–94 (2022).

19. Quentin J. Ullrich, Corpora in the Courts: Using Textual Data to Gauge Genericness and Trademark Validity, 108 Trademark Rep. 989, 1054–55 (2018).

20. E.g., Intercollegiate Women’s Lacrosse Coaches Ass’n v. Corrigan Sports Enters., Inc., No. 1:20cv425, 2020 WL 4227546, at *4 n.2 (M.D.N.C. July 23, 2020) (“Internet searches may be probative of secondary meaning.”); In re Eagle Crest, Inc., 96 U.S.P.Q.2d 1227, 1229–30 (T.T.A.B. 2010) (citing large number of Google search hits as evidence a term “is commonly used”). See also Jamie Sternberg et al., Leveraging Social Media Analytics to Demonstrate a Trademark’s Acquired Distinctiveness or Fame, Int’l Trademark Ass’n (July 13, 2022), https://www.inta.org/perspectives/features/leveraging-social-media-analytics-to-demonstrate-a-trademarks-acquired-distinctiveness-or-fame/ [https://perma.cc/P2CF-J5WR] (“In the United States, the U.S. Patent and Trademark Office (USPTO), the Trademark Trial and Appeal Board (TTAB), and the federal courts recognize social media statistics as evidence supporting the strength or fame of a trademark.”).

21. Lisa Larrimore Ouellette, The Google Shortcut to Trademark Law, 102 Calif. L. Rev. 351, 356, 362, 366–68, 371–73 (2014) (“Google often does a better job than the courts of evaluating factual issues of consumer perceptions.”); see also Alexandra J. Roberts, How to Do Things with Word Marks: A Speech-Act Theory of Distinctiveness, 65 Ala. L. Rev. 1035, 1082–83 (2014) (approving use by trademark examining attorneys of such resources as online message boards, auction sites, and Twitter “to gauge whether a term is descriptive for a given product category”).

22. Daryl Lim, Trademark Confusion Simplified: A New Framework for Multifactor Tests, 37 Berkeley Tech. L.J. 865, 923 (2022).

23. Sonia K. Katyal & Aniket Kesari, Trademark Search, Artificial Intelligence, and the Role of the Private Sector, 35 Berkeley Tech. L.J. 501, 539–540, 545, 547–48. (2020).

24. Idan Mosseri et al., TradeMarker – Artificial Intelligence Based Trademarks Similarity Search Engine, in 1034 Communications in Computer and Information Science 97, 99–101 (2019).

25. See Kelly-Brown v. Winfrey, 717 F.3d 295, 313 (2d Cir. 2013); Fortune Dynamic, Inc. v. Victoria’s Secret Stores Brand Mgmt., Inc., 618 F.3d 1025, 1043 (9th Cir. 2010); see also Marketquest Grp., Inc. v. BIC Corp., 862 F.3d 927, 934–35 (9th Cir. 2017) (evaluating the adequacy of a trademark search in construing intent factor for reverse confusion).

26. See Abdi Aidid & Benjamin Alarie, The Legal Singularity: How Artificial Intelligence Can Make Law Radically Better 4, 8–10 (2023); Robert F. Weber, Will the “Legal Singularity” Hollow Out Law’s Normative Core?, 27 Mich. Tech. L. Rev. 97, 116 (2020).

27. USPTO v. Booking.com B.V., 140 S. Ct. 2298, 2303 (2020).

28. Park 'N Fly, Inc. v. Dollar Park & Fly, Inc., 469 U.S. 189, 194 (1985).

29. Booking.com, 140 S. Ct. at 2303–04.

30. Id. at 2304–05.

31. Booking.com B.V. v. USPTO, 915 F.3d 171, 183 (4th Cir. 2019).

32. Booking.com, 140 S. Ct. at 2304–05.

33. Id. at 2313 (Breyer, J., dissenting).

34. Id. at 2311.

35. E.g., Schwan’s IP, LLC v. Kraft Pizza Co., 460 F.3d 971, 975–76 (8th Cir. 2006) (concluding that survey evidence of secondary meaning should be omitted from genericness analysis if the term at issue was commonly used before its adoption by the mark proponent); Abercrombie & Fitch Co. v. Hunting World, Inc., 537 F.2d 4, 9 (2d Cir. 1976) (“[N]o matter how much money and effort the user of a generic term has poured into promoting the sale of its merchandise and what success it has achieved in securing public identification, it cannot deprive competing manufacturers of the product of the right to call an article by its name.”); Snyder’s Lance, Inc. v. Frito-Lay N. Am., Inc., 542 F. Supp. 3d 371, 381 (W.D.N.C. 2021) (“Therefore, even advertising, repeated use, and consumer association will not warrant affording trademark protection to a generic term.”).

36. Booking.com, 140 S. Ct. at 2309 (Breyer, J., dissenting).

37. Id. at 2304–05, 2308 (majority opinion).

38. Jack Daniel’s Props., Inc. v. VIP Prods. LLC, 143 S. Ct. 1578, 1587–88 (2023). See William McGeveran & Mark P. McKenna, Confusion Isn’t Everything, 89 Notre Dame L. Rev. 253, 289, 292–93 (2013) (describing high litigation costs in discerning likelihood of confusion analysis).

39. Rogers v. Grimaldi, 875 F.2d 994, 997, 999 (2d Cir. 1989).

40. Jack Daniel’s, 143 S. Ct. at 1585–86 (2023).

41. Id. at 1589.

42. See 5 J. Thomas McCarthy, McCarthy on Trademarks and Unfair Competition § 32:158 (5th ed. 2024) (listing factual issues in trademark law as to which survey evidence is relevant).

43. See, e.g., Elec. Design & Sales, Inc. v. Elec. Data Sys. Corp., 954 F.2d 713, 716 (Fed. Cir. 1992) (“[T]he inquiry generally will turn on whether actual or potential ‘purchasers’ are confused.”).

44. E.g., Estee Lauder Inc. v. The Gap, Inc., 108 F.3d 1503, 1510–11 (2d Cir. 1997) (“Likelihood of confusion means a probability of confusion . . . .”); see also 3 J. Thomas McCarthy, McCarthy on Trademarks and Unfair Competition § 23.3 & n.1 (5th ed. 2024) (collecting cases).

45. MGFB Props., Inc. v. Viacom Inc., 54 F.4th 670, 678–80 (11th Cir. 2022).

46. Id. at 681–82.

47. Jack Daniel’s Props., Inc. v. VIP Prods. LLC, 143 S. Ct. 1578, 1589 (2023).

48. Id. at 1592.

49. Id. at 1588. Perhaps most troubling for Rogers supporters, three Justices signed on to a brief concurrence written solely to cast doubt on Rogers’s general validity, warning “it is not entirely clear where the Rogers test comes from . . . it is not obvious that Rogers is correct in all its particulars,” and “lower courts should handle [the Rogers defense] with care.” Id. at 1594 (Gorsuch, J., concurring with Thomas, J., & Barrett, J.).

50. Id. at 1591 (majority opinion). The Court made a similar move with Jack Daniel’s accompanying dilution claim, finding that a statutory defense for “noncommercial use” did not include use as a source identifier, even if that use was meant as a parody. Id. at 1592.

51. See generally Matal v. Tam, 582 U.S. 218 (2017). See also Iancu v. Brunetti, 139 S. Ct. 2294, 2302 (2019).

52. 15 U.S.C. § 1052.

53. Id. § 1052(d).

54. Id. § 1052(a). Admittedly, empirical assessments of consumer reaction have been part of the analysis of whether a mark is “disparaging” or “scandalous.” Both courts and the USPTO have used dictionaries and survey evidence to help assess the perceived meaning of marks proposed for registration. E.g., In re Mavety Media Grp. Ltd., 33 F.3d 1367, 1372–73 (Fed. Cir. 1994); Pro-Football, Inc. v. Harjo, 284 F. Supp. 2d 96, 109–12 (D.D.C. 2003). Nevertheless, determinations of scandalousness and disparagement also involved non-empirical evaluations of whether a mark demeans a particular group. See, e.g., In re Geller, 751 F.3d 1355, 1361–62 (Fed. Cir. 2014) (concluding the proposed mark STOP THE ISLAMISATION OF AMERICA associated terrorism with a peaceful political movement and would therefore be disparaging to a substantial composite of American Muslims). At the least, these determinations involved a much more balanced mix of empirical and non-empirical analyses than the Court’s evaluation of genericness in Booking.com.

55. 15 U.S.C. § 1052(c). At first glance, § 2(c)'s bar on non-consensual invocations of a person’s identity might appear geared towards preventing consumer confusion. It has been construed, however, to protect personal “privacy and publicity” interests. See TMEP § 1206 (July 2022) (stating that § 2(c) is meant to protect “privacy and publicity” rights); Jennifer E. Rothman, Navigating the Identity Thicket: Trademark’s Lost Theory of Personality, the Right of Publicity, and Preemption, 135 Harv. L. Rev. 1271, 1308 (2022) (contending the provision “was adopted to preserve the dignity of those depicted without consent”).

56. Tam, 582 U.S. at 227, 232–33.

57. Id. at 247.

58. Iancu v. Brunetti, 139 S. Ct. 2294, 2297 (2019).

59. Tam, 582 U.S. at 253 (Kennedy, J., concurring with Ginsburg, J., Sotomayor, J. & Kagan, J.) (“The central purpose of trademark registration is to facilitate source identification. . . . Whether a mark is disparaging bears no plausible relation to that goal.”); see also Brunetti, 139 S. Ct. at 2297 (“We hold that this provision infringes the First Amendment for the same reason [the disparagement bar did]: It too disfavors certain ideas.”).

60. See Michael P. Goodyear, Queer Trademarks, 2024 Ill. L. Rev. 163, 170–71 (noting the noncommercial symbolic value in registering and maintaining a trademark); Ned Snow, Immoral Trademarks After Brunetti, 58 Hous. L. Rev. 401, 421 (2020) (“For many consumers, protection from offensive language is at least as beneficial as protection from confusion or deception.”).

61. See 15 U.S.C. § 1052(a).

62. See Brunetti, 139 S. Ct. at 2302.

63. See id. at 2317 n.12 (Sotomayor, J., concurring in part and dissenting in part with Breyer, J.) (remarking that, like the scandalous bar, statutory bans on registering flags, insignias, and unapproved symbols of deceased U.S. Presidents during the lives of their spouses “could be characterized as content discriminatory”); Lisa P. Ramsey, Free Speech Challenges to Trademark Law After Matal v. Tam, 56 Hous. L. Rev. 401, 456–61 (2018) (contending that, after the Supreme Court declared unconstitutional the law denying registration to disparaging marks, the prohibition on dilutive marks should be considered unconstitutional as well). While the Supreme Court did recently hold that § 1052(c)'s prohibition on trademarks consisting of someone’s name without their consent did not violate the First Amendment, the Court was badly fractured as to the rationale for its decision, leaving the constitutionality of other “viewpoint-neutral, content-based trademark restriction[s]” unclear. Vidal v. Elster, 144 S. Ct. 1507, 1513, 1524 (2024).

64. USPTO v. Booking.com, 140 S. Ct. 2298, 2314 (Breyer, J., dissenting).

65. Christopher Buccafusco et al., Competition and Congestion in Trademark Law, 102 Tex. L. Rev. 437, 441–42 (2024) (“[F]ocus on source significance has obscured greater questions of anticompetitive advantage.”); Jeanne C. Fromer, Against Secondary Meaning, 98 Notre Dame L. Rev. 211, 214–15 (2022) (“The move to enshrine secondary meaning seems to be about grounding trademark protection in what consumers actually perceive as terms indicating product source. Yet it is too easy—especially for well-resourced firms—to show that consumers perceive terms as source-indicating, including for marks whose protection undermines trademark law’s goals.”).

66. Buccafusco et al., supra note 65, at 451–52.

67. Id. at 447.

68. Mark A. Lemley & Mark McKenna, Irrelevant Confusion, 62 Stan. L. Rev. 413, 428–29 (2010).

69. Of course, not every effort to imbue a particular symbol with source-identifying significance will succeed. See Jake Linford, Valuing Residual Goodwill After Trademark Forfeiture, 93 Notre Dame L. Rev. 811, 840–42 (2017) (discussing role of preexisting meanings and uses by consumers and competitors in shaping trademark significance). But some research suggests it is not that difficult for an advertiser to turn a merely descriptive term into a source identifier for consumers. Thomas R. Lee et al., An Empirical and Consumer Psychology Analysis of Trademark Distinctiveness, 41 Ariz. St. L.J. 1033, 1078–79 (2009). And, regardless of a trademark’s actual meaning to consumers, some maintain that it is too easy to legally establish secondary meaning and secure trademark rights. Fromer, supra note 65, at 230–34.

70. See Barton Beebe & Jeanne C. Fromer, Are We Running out of Trademarks?: An Empirical Study of Trademark Depletion and Congestion, 131 Harv. L. Rev. 945, 972, 982, 986–88, 1042 (2018).

71. Buccafusco et al., supra note 65, at 459–64.

72. Glynn S. Lunney, Jr., Trademark Monopolies, 48 Emory L.J. 367, 374–88 (1999).

73. This is not to say that trademark doctrine has no mechanisms for addressing the effects of trademark rights on competition. Here are three examples along with their limitations. First, the functionality doctrine assesses a design feature’s potential to hinder competition regardless of consumer perception of that feature, but it tends to apply to product features, not symbols or words. SoClean, Inc. v. Sunset Healthcare Sols., Inc., 52 F.4th 1363, 1370 (Fed. Cir. 2022). Second, words that convey direct information about the characteristics or qualities of goods are “descriptive” whereas those that only indirectly relate to the product are “suggestive.” Because the ability to directly communicate product qualities is viewed as critical to competition, descriptive marks require the mark owner to spend resources proving secondary meaning in the mind of the public before trademark rights will be awarded. By contrast, the indirect allusions of suggestive marks are viewed as less essential to competition and may be immediately protected as trademarks. Nevertheless, descriptive marks can indeed be protected so long as the mark proponent offers sufficient evidence of sales and advertising expenditures, something that is not all that hard to do, particularly for well-resourced firms. Fromer, supra note 65, at 213–14, 230–31. Third, a descriptive fair use defense is available, but it is “notoriously fact intensive and difficult to apply” and only aids parties engaging in non-trademark uses, not those who want to use a descriptive term as part of their brand to compete with the trademark holder. Buccafusco et al., supra note 65, at 459.

74. James Gibson, Risk Aversion and Rights Accretion in Intellectual Property Law, 116 Yale L.J. 882, 907–08 (2007) (describing a feedback loop between trademark law and consumer thought as “what consumers view as the norm becomes the norm because consumer perception is trademark law’s touchstone”).

75. See Barton Beebe, Search and Persuasion in Trademark Law, 103 Mich. L. Rev. 2020, 2062–64 (2005) (discussing trademark law’s efforts to create an “ideal type” of consumer); Dustin Marlan, Is the Word “Consumer” Biasing Trademark Law?, 8 Tex. A&M L. Rev. 367, 389–98 (2021) (contending that the pejorative valence of the term “consumer” has led to unfortunate doctrinal outcomes in trademark law, including an increase in findings of confusion and a lack of standing to sue).

76. See Mark Bartholomew, Neuromarks, 103 Minn. L. Rev. 521, 569–71 (2018) (discussing ability of modern neuroscience to record rapid changes in mental state that might be cited as evidence of consumer confusion).

77. Nun Vanichkul et al., Visual-Based Confusion Detection Using a Cooperative Spatio-Temporal Deep Neural Networks, in International Conference on Digital Government Technology & Innovation 80, 81–83 (2022).

78. Dinwoodie, supra note 9, at 333.

79. Alfred C. Yen, The Constructive Role of Confusion in Trademark, 93 N.C. L. Rev. 77, 125–26 (2014).

80. Ex parte Banner Cigar Mfg. Co., 1909 Dec. Comm’r Pat. 9–10.

81. Trade-Marks: Hearing on H.R. 9041 Before the Subcomm. on Trade-Marks of the H. Comm. on Pats., 75th Cong. 79 (1983); see also Christine Haight Farley, The Lost Unfair Competition Law, 110 Trademark Rep. 739, 756 (2020). In more recent times, § 2(c) has been invoked to block “Obama” pajamas and “Eisenhower” greeting cards. In re Hoefflin, 97 U.S.P.Q.2d (BNA) 1174, 1176–78 (T.T.A.B. 2010); In re Masucci, 179 U.S.P.Q. (BNA) 829, 829–30 (T.T.A.B. 1973).

82. Ex parte Banner Cigar Mfg. Co., 1909 Dec. Comm’r Pat. 9–10; Rebecca Jennings, Everyone’s a Sellout Now, Vox (Feb. 1, 2024, 7:00 AM), https://www.vox.com/culture/2024/2/1/24056883/tiktok-self-promotion-artist-careerhow-to-build-following [https://perma.cc/6HR4-4U5J]; Claudia Lauer & Jill Colvin, Trump Hawks $399 Branded Shoes at ‘Sneaker Con,’ a Day After a $355 Million Ruling Against Him, AP News, https://apnews.com/article/trump-sneakers-sneaker-con-philadelphia-4de093eda6f8d1c68baf8fe8095f777b [https://perma.cc/JK68-XZC7] (last updated Feb. 17, 2024, 8:45 PM).

83. Generally speaking, decisions as to registrability require predictions about audience reaction to a new commercial symbol as opposed to a full gathering of evidence about the public’s historical understanding of a mark. Consequently, registrability decisions have tended to stand on more modest empirical terrain than infringement disputes, and thus afforded more opportunities for application of non-empirical reasoning. Dinwoodie, supra note 9, at 313–16. Nevertheless, in the decisions described above, the Court has acted to limit those opportunities, thereby prompting even more attention to behavioral measurement. Id. at 325–26.

84. See Anya Bernstein, Technologies of Language Meet Ideologies of Law, 2020 Mich. St. L. Rev. 1241, 1243, 1247–48. As one critic of using corpus linguistics to define legal terms has noted, that technology can turn some things into “objects of interest” while pushing other sources of meaning to the periphery. See id. at 1247–48.

85. Hugh Baxter, System and Lifeworld in Habermas’s Theory of Law, 23 Cardozo L. Rev. 473, 480 (2002).

86. Jürgen Habermas, Between Facts and Norms: Contributions to a Discourse Theory of Law and Democracy 447–48 (William Rehg trans., 1996); see also Lon L. Fuller, The Morality of Law 39, 91–92 (rev. ed. 1969) (discussing requirement of “publicity” for the rule of law); Tom R. Tyler, Why People Obey the Law 149 (1990) (highlighting importance of explanations and consideration of different views in legal systems to build acceptance of legal decisions).

87. Habermas, supra note 86, at 226 (“The validity of a judgment is certainly defined by the fact that its validity conditions are satisfied. Whether these are satisfied, however, cannot be clarified by direct access to empirical evidence or to facts given in an ideal intuition, but only discursively, precisely by way of a justification that is carried out with arguments.”); see also 1 Jürgen Habermas, The Theory of Communicative Action 249 (Thomas McCarthy trans., 1984) (“Validity claims differ from empirical claims through the presupposition that they can be made good by means of arguments.”); Felipe Jiménez, The Limits of Experimental Jurisprudence, in Cambridge Handbook of Experimental Jurisprudence (forthcoming 2025) (manuscript at 9) (contending that empirical evidence of “lay cognition” cannot “settle the question” of the proper account of a legal concept and “necessarily requires the assistance of some type of normative argument”).

88. Habermas, supra note 86, at 217.

89. Id. at 108.

90. See Diego Fernandez-Duque et al., Superfluous Neuroscience Information Makes Explanations of Psychological Phenomena More Appealing, 27 J. Cognitive Neuroscience 926, 926–27, 931 (2015) (describing how “superfluous neuroscience information increased the judged quality of the argument for both good and bad explanations”).

91. Dinwoodie, supra note 9, at 321.

92. USPTO v. Booking.com, 140 S. Ct. 2298, 2304–05 (2020).

93. . U.S. Const. amend. VIII.

94. Stanford v. Kentucky, 492 U.S. 361, 370–71 (1989).

95. John B. Wefing, Cruel and Unusual Punishment, 20 Seton Hall L. Rev. 478, 493 (1990).

96. The concept of the “reasonable person” in tort law tends to rely on such a blended definition. See Kevin P. Tobia, How People Judge What Is Reasonable, 70 Ala. L. Rev. 293, 296 (2018) (“Reasonableness is best understood as a hybrid notion that is partly statistical and partly prescriptive.”). It should be noted, however, that this is an unsettled question in both the courts and legal scholarship. See Alan D. Miller & Ronen Perry, The Reasonable Person, 87 N.Y.U. L. Rev. 323, 325 (2012) (discussing the two “contradictory paradigms of reasonableness”—normative and positive—and concluding “[o]nly normative definitions are logically acceptable”).

97. Booking.com, 140 S. Ct. at 2304.

98. Id. at 2304 n.3 (“Sufficient to resolve this case is the undisputed principle that consumer perception demarcates a term’s meaning.”).

99. Cf. Genesee Brewing Co. v. Stroh Brewing Co., 124 F.3d 137, 140 (2d Cir. 1997) (“[W]hen a producer creates a new product that differs from an established product class in a particular characteristic, the law of trademark will not grant the producer the exclusive right to label its product with words that are necessary to describe that new characteristic.”) (emphasis added).

100. See, e.g., Leah Chan Grinvald, Shaming Trademark Bullies, 2011 Wis. L. Rev. 625, 644, 650, 652.

101. See Rebecca Tushnet, What’s the Harm of Trademark Infringement?, 49 Akron L. Rev. 627, 646 (2016) (“As long as trademark purports to be guided by consumer reactions, it can only benefit from a better understanding of those reactions.”).

102. Timothy Endicott, The Value of Vagueness, in Philosophical Foundations of Language in Law 14, 23 (Andrei Marmor & Scott Soames eds., 2011).

103. Id. at 22–23.

104. See, e.g., Kellogg Co. v. Nat’l Biscuit Co., 305 U.S. 111, 116–17 (1938) (suggesting that SHREDDED WHEAT was generic, in part, because patent law gave competitors the right to copy the shredded wheat biscuit shape and, as a result, those competitors also had the right to use the term referencing those biscuits).

105. USPTO v. Booking.com, 140 S. Ct. 2298, 2305 (2020).

106. Rachel Paige Thompson, Booking.com and the Shrinking Hyperlinguistic Commons, 36 Berkeley Tech. L.J. 1473, 1494 (2021) (noting that in determining genericness, “[r]ather than focusing on consumers, another option for factfinders would be to inquire about competitors”).

107. Dreamwerks Prod. Grp., Inc. v. SKG Studio, 142 F.3d 1127, 1129 (9th Cir. 1998).

108. For example, Jake Linford’s article for this issue explains how the intent factor in the likelihood of confusion test fulfills trademark law’s goal of encouraging the transmission of accurate information to consumers. Jake Linford, An Information Theory of Intentional Trademark Infringement, 62 Hous. L. Rev. 275, 283 (2024). Robert Bone has described a different purpose for intent: as a means for penalizing morally objectionable behavior. Robert G. Bone, Taking the Confusion out of “Likelihood of Confusion”: Toward a More Sensible Approach to Trademark Infringement, 106 Nw. U. L. Rev. 1307, 1350–51 (2012).

109. See, e.g., William McGeveran, Rethinking Trademark Fair Use, 94 Iowa L. Rev. 49, 72−74 (2008) (discussing importance of accommodating the normative value of free speech within trademark doctrine). Regardless of how one weighs trademark law’s competing normative goals—reducing consumer search costs, facilitating competition, preserving free expression, etc.—few would argue that trademark law should maintain absolute fidelity to only one of these concerns. Id.

110. Cf. Scott Fetzer Co. v. House of Vacuums Inc., 381 F.3d 477, 485 (5th Cir. 2004) (advocating flexibility in applying likelihood of confusion factors: “otherwise, we risk inadvertently lowering the standard of confusion”).

111. Anthony E. Dowell, Note, Trade Dress Protection of Product Designs: Stifling the Progress of Science and the Useful Arts for an Unlimited Time, 70 Notre Dame L. Rev. 137, 169 (1994) (“The Lanham Act merely federalizes the common law of trademarks and unfair competition. . . . The doctrines of secondary meaning/inherent distinctiveness, functionality, and the multifactor confusion test are entirely judge-made and a product of the common law.”).

112. But see Lim, supra note 22, at 929, 938 (proposing a streamlined confusion analysis that only relies on the “three core factors” revealed in an empirical study of infringement decisions, and then training AI models based on that data set). I am not suggesting that courts currently rely exclusively on empirical evidence to assess these particular confusion factors. Rather, this suggestion springs from a prediction that such evidence will be more frequently available and offered for these particular factors, potentially causing them to loom larger in the likelihood of confusion analysis than others. See Bartholomew et al., supra note 13, at 801–04 (describing experiment using neuroscience to uncover empirical evidence of visual mark similarity); Moran Cerf et al., Using Single-Neuron Recording in Marketing: Opportunities, Challenges, and an Application to Fear Enhancement in Communications, 52 J. Mktg. Res. 530, 534 (2015) (discussing how neuroscience can be used to determine familiarity with a trademark, which aligns with acquired distinctiveness, one of the two ingredients in determining the mark strength factor); Linford & Nelson, supra note 18, at 175 (explaining how courts and litigants can use corpus evidence to help establish or call into question singularity and prominence of well-known trademarks, information relevant to the mark strength factor).

113. See Ass’n of Am. Med. Colls. v. Princeton Rev., Inc., 332 F. Supp. 2d 11, 17 (D.D.C. 2004) (“The doctrine forbids the use of the copyright to secure an exclusive right or limited monopoly not granted by the Copyright Office and which is contrary to public policy to grant.”).

114. Prac. Mgmt. Info. Corp. v. Am. Med. Ass’n, 121 F.3d 516, 521 (9th Cir. 1997).

115. Deepa Varadarajan, The Uses of IP Misuse, 68 Emory L.J. 739, 758–59 (2019).

116. See, e.g., KEMA, Inc. v. Koperwhats, 658 F. Supp. 2d 1022, 1036 (N.D. Cal. 2009) (questioning whether “trademark misuse ever exists as a separate affirmative defense”).

117. Mark Bartholomew & John Tehranian, Historical Kinship & Categorical Mischief: The Use and Misuse of Doctrinal Borrowing in Intellectual Property Law, 109 Iowa L. Rev. 51, 114 (2023).

118. See Gibson, supra note 74, at 907–08.

119. See Michael Grynberg, Things Are Worse Than We Think: Trademark Defenses in A “Formalist” Age, 24 Berkeley Tech. L.J. 897, 947–48, 948 n.232 (2009) (discussing common law development of trademark defenses).

120. William E. Ridgway, Note, Revitalizing the Doctrine of Trademark Misuse, 21 Berkeley Tech. L.J. 1547, 1572−74 (2006). Trademark law already has built-in First Amendment defenses, but those have been considered insufficient by many in the field, in part, because of their insistence on investigating and relying on the same empirical evidence of confusion at stake in the plaintiff’s case for infringement. See Stacey L. Dogan & Mark A. Lemley, Parody as Brand, 47 U.C. Davis L. Rev. 473, 506–07 (2013) (“[C]ritics have complained that nominative fair use simply re-entangles courts in the same likelihood of confusion inquiries that define infringement in the first place.”).

121. Ass’n of Am. Med. Colls. v. Princeton Rev., Inc., 332 F. Supp. 2d 11, 17 (D.D.C. 2004) (“[M]isuse often exists where the patent or copyright holder has engaged in some form of anti-competitive behavior.”).

122. Cf. Lasercomb Am., Inc. v. Reynolds, 911 F.2d 970, 978–79 (4th Cir. 1990) (anticompetitive language in software licensing agreement, including prohibitions on developing any of the same type of software for ninety-nine years, constituted copyright misuse).

123. See Daniel M. Lifton, Note, Towards a Trademark Rule of Reason, 9 N.Y.U. J. Intell. Prop. & Ent. L. 222, 236−38 (2020).

124. Arlen W. Langvardt, The Troubling Implications of a Right of Publicity “Wheel” Spun Out of Control, 45 U. Kan. L. Rev. 329, 352–53 (1997) (describing the analysis in right of publicity claims as “[c]onfusion or no confusion, the celebrity’s property has been used”).

125. Eric E. Johnson, Disentangling the Right of Publicity, 111 Nw. U. L. Rev. 891, 904–05 (2017) (discussing the right of publicity’s lack of definition and its unusual dependence on outside defenses to give it shape).

126. See Stacey L. Dogan & Mark A. Lemley, What the Right of Publicity Can Learn from Trademark Law, 58 Stan. L. Rev. 1161, 1190–93, 1220 (2006). Given its lack of interest in consumer behavior, some have suggested right of publicity law should adopt the more empirical stance of trademark law and limit publicity rights to situations where the consumers would be confused as to who authorized the commercial use of someone’s identity. Id.

127. Comedy III Prods., Inc. v. Gary Saderup, Inc., 21 P.3d 797, 808 (Cal. 2001); Dora v. Frontline Video, Inc., 15 Cal. App. 4th 536, 545 (Cal. Ct. App. 1993); Rosemont Enters., Inc. v. Random House, Inc., 58 Misc. 2d 1, 4 (N.Y. Sup. Ct. 1968).

128. See, e.g., Vans, Inc. v. MSCHF Prod. Studio, Inc., 88 F.4th 125, 138 (2d Cir. 2023).

129. Edward S. Rogers, The Unwary Purchaser: A Study in the Psychology of Trade Mark Infringement, 8 Mich. L. Rev. 613, 622 (1910).

130. Mark Bartholomew, Advertising and the Transformation of Trademark Law, 38 N.M. L. Rev. 1, 26 (2008).

131. See Sheri J. Broyles, Subliminal Advertising and the Perpetual Popularity of Playing to People’s Paranoia, 40 J. Consumer Affs. 392, 392, 396 (2006) (describing research debunking earlier reports on the effectiveness of subliminal advertising); David Mikkelson, Popcorn Subliminal Advertising, Snopes, https://www.snopes.com/fact-check/subliminal-advertising/ [https://perma.cc/8JFG-YWNP] (last updated May 3, 2011).

132. See Barton Beebe et al., Testing for Trademark Dilution in Court and the Lab, 86 U. Chi. L. Rev. 611, 661 (2019); Jake Linford et al., Trademark Tarnishmyths, 55 Ariz. St. L.J. 609, 674 (2023).

133. Victoria Nourse & Gregory Shaffer, Empiricism, Experimentalism, and Conditional Theory, 67 SMU L. Rev. 141, 183−84 (2014).

134. Donald E. Frey, America’s Economic Moralists: A History of Rival Ethics and Economics 91 (2009).