# The Future of Trademarks in a Global Multilingual Economy: Evidence and Lessons from the European Union

**Authors:** Barton Beebe, Jeanne C. Fromer
**Citation:** "The Future of Trademarks in a Global Multilingual Economy: Evidence and Lessons from the European Union," 112 *Trademark Reporter* 902 (2022) (with Barton Beebe)
**Source:** https://its.law.nyu.edu/faculty/profiles/representiveFiles/Beebe Fromer - Future of TMs in Global Multil Econ - TMR_24DDF1AB-E3BC-2F2C-D241EE5FBE0B00BB.pdf

## INTRODUCTION

*p. 6*
Companies around the world are increasingly pursuing global branding strategies in which they seek to use the same trademark in all of the national or regional markets in which they sell their goods. To that end, such companies typically attempt to register the same mark in each of the trademark offices associated with those markets. The result is that the various national and regional trademark systems of the world are integrating into a de facto global trademark system. Substantial proportions of trademark office registries intersect with other offices' registries. Though trademark law and individual trademark registrations typically remain delimited by national borders, trademarks themselves are increasingly transnational, even global, entities.

*p. 6*
The emergence of a global trademark system presents significant challenges to companies seeking to develop new brand names. They are facing mounting difficulties finding brand names that will be effective throughout the global marketplace, including in every one of its many languages, but that have not yet been claimed by another entity somewhere in that marketplace. The problem is that transnational market integration decreases the supply of marks at the same time that it increases the demand for competitively effective trademarks. On the supply side, as a market integrates, the number of trademarks that are effective in that market declines. To be effective throughout the market, a mark must be effective in each of the languages and cultures of which the market is composed; it must lie within the narrow intersection of the various sets of marks that are viable in each national submarket. Thus, for example, Microsoft would prefer that its Internet search engine not be known as BING in China, because BING may be understood to mean "sickness" in Mandarin. 1 Especially attractive in a global multilingual trademark system are what we call "multilanguage words," that is, cognates or loan words that are the same or closely similar in multiple languages (such as "fantastic," "idea," or "virus"). Because these words can be understood by people across jurisdictions and languages, it is easy to see how businesses can find them desirable as trademarks, but the supply of them is extraordinarily limited.

*p. 6*
Reducing the supply of available trademarks even further is what we designate the "reverse Babel problem" in global trademark Vol. 112 TMR law. 2 Most trademark systems conventionally hold that two orthographically and phonetically different marks from different languages (such as APPLE and MANZANA) may be confusingly similar if they convey the same meaning to any significant population of consumers capable of understanding the terms in both languages, something we call "translational similarity." 3 Thus, a registration for APPLE may also claim the translationally equivalent word in every other language in the market-as well as the phonetically and orthographically equivalent word in every other language. For this reason, when descendants of Baron von Richthofen sought in 2000 to register the English-language mark RED BARON at the European Union Intellectual Property Office ("EUIPO") for various goods and services, owners of the alreadyregistered Spanish-language mark BARON ROJO-meaning the same thing-succeeded in preventing the registration for the goods on which they were already using their mark. 4 On the demand side, at the same time that an integrating market and trademark law constrict the supply of competitively effective marks, an integrating market itself increases the demand for those marks. There are simply more and more entities claiming exclusive rights within the same commercial sign system. Forty years ago, there were only 75,000 registered marks in the automotive space, but now there are over 800,000. 5 This has led even top automakers to use the same, or overlapping, marks. In 2013, Infiniti rebranded its entire car line to begin with a Q followed by a number (such as Q50), even though Audi was simultaneously using the same letter Q followed by a number (such as Q5) as the brand name for many of its cars. 6 An industry expert worried that choosing distinct marks from "the shrinking pool of available words" will lead to car "names that sound like pharmaceuticals," notorious for their nonsensical brand names. 7 As the automobile industry example suggests, these forces of supply and demand create the conditions for severe levels of trademark depletion and trademark crowding in the global trademark system. In previous work, we defined trademark depletion as the process by which an increasing proportion of 2 "Come, let us go down, and there confuse their language, that they may not understand one another's speech." Genesis 11:7; infra section V.B.3.

*p. 7*
3 Infra section I.B.2. 4 OHIM Opposition Division Decision No. 3111/2000 (Dec. 21, 2000). 5 Lindsay Chappell, With 800,000 Auto Names Already Trademarked, It's Tougher than Ever to Find a Moniker that Works Worldwide, Auto. News, Sept. 30, 2013, https://www.autonews.com/article/20130930/OEM/309309978/with-800-000-auto-names-already-trademarked-it-s-tougher-than-ever-to-find-a-moniker-that-worksworldwide.

*p. 8*
Vol. 112 TMR 907 competitively effective trademarks are claimed by one or more registrants. 8 As levels of trademark depletion increase, it becomes more difficult for market entrants to find marks that are not identical or closely similar to already-claimed marks. Ever fewer unclaimed marks remain available. Trademark crowding is a related phenomenon. If an entity has registered a particular mark in a class of goods or services, that registration will likely significantly hinder but not necessarily prevent other unrelated entities from registering the same or a closely similar mark in that class. But when unrelated entities succeed in obtaining such parallel registrations, the result is trademark crowding, in which increasing numbers of identical or closely similar, if not confusingly similar, marks registered by unrelated entities coexist in the marketplace. 9 As trademark crowding levels increase, consumers face greater challenges in differentiating marks, which impairs both the source-indicating and advertising functions of the marks and degrades the overall integrity of the trademark system.

*p. 8*
In this article, we seek to understand the dangers that the processes of globalization and intensifying market integrationparticularly across jurisdictions that speak different languagespose to the viability of the emerging global trademark system. To do so, we undertake an empirical case study of the transnational trademark system of the European Union. Consisting of 450 million rich-world consumers, 10 the EU marketplace accounts for 15% of the global economy. 11 An EU trademark registration establishes exclusive rights over the entirety of this marketplace and is among the most potent trademark registrations in the world. But the EU trademark system is distinctive not simply because of the magnitude of the European Union's gross domestic product. 8 Barton Beebe & Jeanne C. Fromer, Are We Running Out of Trademarks? An Empirical Study of Trademark Depletion and Congestion, 131 Harv. L. Rev. 945, 978 (2018). 9

*p. 8*
As discussed further below, in our previous study we focused on a special case of trademark crowding, which we referred to as "trademark congestion." Id. at 1012. Congestion refers to the increasing number of identical marks used by unrelated entities in a particular class of goods or services. Id. Trademark crowding refers more generally to the increasing number of identical or closely similar marks used by unrelated entities in a particular class of goods or services. We previously focused on the special case of trademark congestion largely because of limitations in computer processing capacity available to us. We have since overcome those limitations and are able in this study to quantitatively assess the broader phenomenon of trademark crowding, one with which trademark lawyers are familiar, though one that to our knowledge has never been studied empirically. World Economic Outlook Database, Int'l Monetary Fund, https://www.imf.org/en/ Publications/WEO/weo-database/2020/October/ weo-report?a=1&c=001,110,163,119,123, 998,200,505,511,903,205,400,603,&s=NGDPD,&sy=2018&ey=2025&ssm=0&scsm=1&s cc=0&ssd=1&ssc=0&sic=0&sort=country&ds=.&br=1 (last visited June 17, 2022).

## Vol. 112 TMR

*p. 9*
Consisting of 27 nations 12 speaking 24 different languages, 13 the European Union is a massively multicultural, multilingual marketplace, one that is composed of once-separate national markets that are now increasingly integrating into a single market. From its formation in 1996, the EU trademark system has both fostered and been forced to cope with intensifying economic, cultural, and linguistic integration. The most expansive and complex transnational trademark system in the world, it is uniquely a microcosm of the global trademark system and the many challenges it faces.

*p. 9*
We use the EUIPO's recently released Open Dataset and an array of other datasets to show that the EU trademark system is already experiencing extreme levels of trademark depletion and crowding, exceeding even those in the U.S. system. To further assess the implications of integrating jurisdictions that use different languages, we also measure the desirability of claiming multilanguage words and the degree to which the reverse Babel problem exacerbates trademark depletion and crowding. We also appraise how the EU trademark system has sought to cope with these trends and compare the Europeans' more permissive approach to the registration of closely similar marks to the Americans' stricter approach. Based on the European example, we conclude that this new stage in the development of the world's major trademark systems into a global multilingual system-one characterized by both severe depletion and crowding-will require a number of new policies and doctrines to maintain these systems' continued integrity. The need for these reforms will only grow as the integration of the various national and regional trademark systems intensifies. 14 As to reforms, we think that it is critical that trademark law finally recognize that there are real costs to granting trademark rights. Since the beginning of modern trademark law, the assumption everywhere has been that there is an inexhaustible supply of potential trademarks available for adoption by market entrants, either in the form of common dictionary words in some language or new coined terms. Because we have assumed that there will always be "enough and as good"foot_2 left for others, we have considered the granting of exclusive rights in such marks to be essentially costless, and registering agencies around the world have 12 Countries, Eur. Union, https://europa.eu/european-union/about-eu/countries_en (last visited June 17, 2022).

## EU

*p. 9*
Languages, Eur. Union, https://europa.eu/european-union/about-eu/eu-languages_en (last visited June. 17, 2022). 14 Graeme B. Dinwoodie, Territorial Overlaps in Trademark Law: The Evolving European Model, 92 Notre Dame L. Rev. 1669, 1672 (2017) [hereinafter Dinwoodie, Territorial Overlaps]; infra section I.B.

*p. 10*
Vol. 112 TMR 909 unrestrainedly granted such rights on that basis. But we have now reached a stage of economic development where that assumption no longer holds. Supply is no longer adequate to meet demand, especially at the transnational or more broadly global level. This "peak trademark" 16 condition urges a rethinking of many costbenefit analyses in global trademark policy and doctrine. In particular, we advocate a rethinking of translational similarity as a basis for a finding of confusing similarity and stronger enforcement of a use-in-commerce requirement as a prerequisite for trademark rights. We further recommend that offices that do not engage in ex officio review for confusing similarity either institute such a system of review or, short of that, at least provide to current registrants better information about applied-for marks that potentially conflict with already-registered marks.

*p. 10*
Our argument proceeds as follows: Part I addresses the challenges of brand-name selection in a multinational marketplace. To set the stage for our case study, this Part also provides background on the EU trademark system and the various relevant ways in which EU trademark law differs from U.S. trademark law. Part II describes the datasets that we use. Part III details our findings on trademark depletion in the EU trademark system. Part IV focuses on trademark crowding in the EU trademark system. Part V discusses the implications of our findings both generally for global trademark policy and more specifically for particular points of trademark doctrine within individual trademark systems.

## I. THE GLOBAL MULTILINGUAL MARKETPLACE

*p. 10*
The fundamental purpose of a trademark system is to promote the communication of accurate and easily intelligible information about the source of goods and services in the marketplace. 17 The primary way in which a trademark system does so is by preventing conduct that causes consumer confusion as to source or affiliation. 18 More specifically, trademark law prevents firms from using trademarks that are sufficiently similar to other firms' preexisting marks that consumers would likely be confused as to the true source of the products bearing the marks. For example, the law would prevent a market entrant from selling its mobile phones under the trademark APELL because of the likelihood that this would confuse a significant proportion of consumers into thinking that those phones originate from or are affiliated with the same source as 16 Cf. Colin J. Campbell & Jean H. Laherrère, The End of Cheap Oil, Sci. Am., Mar. 1998, at 78 (predicting a condition in which the rate of world oil production would peak while demand would continue to rise). 17 Mark P. McKenna, The Normative Foundations of Trademark Law, 82 Notre Dame L. Rev. 1839 (2007). Vol. 112 TMR phones bearing the trademark APPLE. By preventing confusion, trademark law preserves the integrity of an information system that enables consumers to more easily find, in the clamor of the marketplace, the products that they seek. It allows consumers to rely on trademarks to indicate the source and thus the quality and characteristics of those products. In economic terms, it lowers consumers' "search costs." 19 Stated more generally, trademark law defends signals against noise. 20 In this Part, we discuss the constraints that limit the population of brand names that are viable in a multinational or global marketplace. We then set out background information about the EU trademark system.

## A. Brand Name Selection in a Global Multilingual Marketplace

*p. 11*
Largely due to the influence of law and economics, the conventional wisdom in trademark law has long assumed that there is an inexhaustible supply of good, competitively effective trademarks, and that if the trademark a company wishes to use has already been claimed, then that company can easily find an alternative that is just as good. 21 As a theoretical matter, this assumption may be correct; new companies can always coin new words or phrases-or even alphanumeric codes-of ever increasing length in an attempt to distinguish themselves and their products in the marketplace. 22 But as a this-worldly matter of basic marketing know-how, the assumption that there is an infinite supply of competitively effective trademarks, each just as good as the other, is almost certainly wrong-if not ridiculous. 23 Instead, as we have discussed in previous work, 24 competitively effective brand names tend to share certain characteristics, which Richard A. Posner, Trademark Law: An Economic Perspective, 30 J.L. & Econ. 265, 274 (1987). 20 By enforcing exclusive rights, trademark law also encourages producers to maintain consistent levels of product quality by ensuring that they, and not their competitors or counterfeiters, will internalize any gains to reputation from doing so. McKenna, supra note 17, at 1844-49 (describing this theory as "conventional wisdom"). 21 E.g., Landes & Posner, supra note 19, at 274 ("[T]he distinctive yet pronounceable combinations of letters to form words that will serve as a suitable trademark are as a practical matter infinite, implying a high degree of substitutability and hence a slight value in exchange."). See generally Beebe & Fromer, supra note 8, at 962-64 (elaborating on and clarifying this conventional wisdom).

*p. 12*
Vol. 112 TMR 911 can significantly limit their quantity. First, they tend to be unique, both in the sense that no other company anywhere in the economy uses the term as a mark (COCA-COLA is unique in this sense, while UNITED is not 25 ) and in the sense that the mark stands out as distinctive as compared with other marks in the marketplace (GOOGLE is unique in this sense, while NAPSTER, GROKSTER, and FRIENDSTER, with each crowding around the -ster suffix, are not). 26 Second, common dictionary words used in a suggestive or arbitrary manner (for example, JAGUAR used suggestively for cars or APPLE used arbitrarily for electronics goods) tend to be more effective as brand names than coined terms. 27 They impart connotations of familiarity and authenticity and are easier to pronounce and remember. 28 Third, if coined terms are used, they are most effective when they evoke more familiar words that convey the brand's meaning, as with VIAGRA, which simultaneously suggests "vigor," "vitality," "aggression," and "Niagara." 29 Fourth, shorter marks are generally more effective than longer marks. A rule of thumb is that, ideally, a brand name should be no more than two syllables and seven letters in length. 30 Finally, firms strongly prefer brand names that they can register in the .com top-level domain. 31 Indeed, branding consultancies now often recommend searching for new brand names from among terms that are still available for registration as a domain name. 32 For transnational firms doing business with transnational consumers, there are further, quite severe constraints-so severe as to suggest that the number of potential brand names that will be competitively effective globally is not infinite, but closer to zero. These constraints apply nearly universally, because now even most

## Vol. 112 TMR

*p. 13*
small businesses tend to operate across national borders. 33 The most significant constraint is that the brand typically should be the same everywhere it appears. Ideally, as a leading marketing textbook puts it, "the marketing program for a global brand consists of … one package design, one advertising program, one pricing schedule, one distribution plan, and so on." 34 Uniform branding is thought to lower marketing costs, convey credibility and status, offer a consistent brand image to consumers as they traverse jurisdictions, and make market entry into new jurisdictions easier. 35 The result is that all of the characteristics just discussed-uniqueness, familiarity, pronounceability, memorability, and positive associations-should hold in each local market in which the brand seeks market share. Of particular concern is that the mark not carry negative connotations in any relevant foreign language. 36 Thus, IRISH MIST liquor changed its name in German-speaking markets, where "Mist" means "manure" and is used as an interjection; the producer of TEGRO weight-loss pills changed their name in Frenchspeaking countries, where "tegro" can be understood to mean "you are fat"; and-as noted above-Microsoft would prefer that its search engine be known as BIYING in China. 37 In the European Union, for example, with twenty-four official languages and several other major regional languages such as Basque and Catalan, the challenge of finding good marks that satisfy all of these conditions is considerable. Finding such marks that are not already claimed by others is harder still. Engaging in the same search on a global scale is even worse.

*p. 13*
Stated differently, transnational firms optimally choose a mark from the intersection, not the union, of the differing sets of marks that are competitively effective and still unclaimed in each local 33 E.g., Shubhomita Bose, 58 Percent of Small Businesses Already Have International Customers, Survey Finds, Small Bus. Trends, Aug. 26, 2016, https 38 This is one reason why image marks, which can transcend the barriers of text (and literacy), can be so powerful. 39 Yet given the constraints of Internet search and social media and given the importance of word of mouth, text remains the dominant medium of marketing. 40 When expanding businesses do not heed this wisdom from the start, they may be forced to choose different marks for their product in different markets. For example, when clothing retailer TJ MAXX expanded from the United States to Europe and Australia, it was compelled to operate in these new markets as TK MAXX to avoid confusion with already-established UK-based department store TJ Hughes. 41 Similarly, in order to avoid confusion with existing marks, BURGER KING restaurants are called HUNGRY JACK'S in Australia and BUDWEISER beer is called BUD in Europe. 42 As consumers cross borders, the use of localized brands can cause more 38 But cf. Daniel J. Hemel & Lisa Larrimore Ouellette, Trademark Law Pluralism, 88 U. Chi. L. Rev. 1025, 1064 (2021) ("[O]ne side benefit of public investment in foreign language learning would be an expansion of the set of possible marks that Americans could recognize at relatively low marginal cognitive cost.").

*p. 15*
confusion than it otherwise seeks to solve. In an oft-cited example, transatlantic consumers may be hopelessly confused by the similarities and differences among MARS chocolate bars in Europe, MILKY WAY chocolate bars in the United States and Europe, and 3 MUSKETEERS bars in the United States. 43 Specifically, Mars produces one chocolate bar (with caramel) under the brands MARS in Europe and MILKY WAY in the United States. 44 Meanwhile, Mars produces another chocolate bar (without caramel) under the brands MILKY WAY in Europe and 3 MUSKETEERS in the United States. 45 We emphasize two additional considerations that firms must take into account when choosing new brand names. First, within the narrow set of words that are effective in all markets in which a transnational business is operating, there lies an especially limited subset of highly versatile words from which businesses might choose a mark consisting of what we call "multilanguage words." Often taking the form of cognates, borrowings, or onomatopoeia, these are words that are orthographically closely similar across multiple languages and convey roughly similar meanings in each of those languages. 46 For example, the equivalent of the English word FANTASTIC is FANTASTISCH and FANTASTIQUE in German and French, respectively, and FANTASTICO in both Italian and Spanish. Other such words in English are ASPECT, BOOM, IDEA, IDEAL, MODERN, and METHOD, whose equivalents are mutually intelligible among all of the five major European languages. 47 Such words are highly prized as brand names or as parts of brand names because they are comprehensible in multiple markets. 48 Indeed, to consumers who do not speak the word's language, such words may convey an optimal combination of distinctiveness and familiarity in that, like many forms of fashion, they are just foreign enough to be interesting but just recognizable enough to be reassuring. 49 A final, overriding consideration that firms must take into account is that English is by far the dominant language of global commerce, and global branding is no exception. 50 The world's sole "hypercentral language," spoken by approximately two billion people and by approximately 400 million as a native language, English is the one language that all brands that aspire to be global must work with and accommodate. 51 As the EUIPO data suggest, most brands do so by taking the form of English-language word marks. Figure 1 shows the distribution by mark language of trademark applications filed from 1996 through 2018 at the EUIPO for marks containing words. 52 Of such marks, 57.5% consisted of English-language marks. This is particularly notable given that English, even pre-Brexit, was the native language for only 13% of the European Union's citizens, behind German (at 18%) and tied with Italian. 53 Distantly following English in the distribution of mark languages were Italian, French, Spanish, and German, accounting for 4.5%, 3.4%, 3.3%, and 3.2% of mark languages, respectively.

## 49

*p. 16*
Id. ("When developing names for a global audience, it's helpful to draw on universally relevant ideas. So a name that suggests 'happy' will be more relevant to a global audience than one suggesting 'Sycamore' (a type of tree)."). We detected the language of marks by using the Google Translate API's languagedetection feature. For the general reliability of this feature, see infra note 117. This feature was somewhat less reliable in detecting word marks of less than four characters, as they were often acronyms or other irregular words. The distribution of languages is nearly the same, however, even if those word marks are omitted. With this background on brand name selection in the global marketplace, we now turn to the source of our case study, the EU trademark system, given that it is a microcosm of the global marketplace.

## B. The EU Trademark System: A Microcosm of the Global Marketplace

*p. 17*
Like any trademark system, the EU trademark system seeks to promote the efficient communication of accurate information about commercial source. But among the many things that make the EU system so interesting is that from the start, it has been expressly designed to serve a far weightier purpose: to promote the integration of the European single market and thereby promote the European project. 54 In both civil and common law systems, trademark rights have traditionally been understood as territorial in nature; at best,

## 54

*p. 17*
In reaction to World Wars I and II, many thinkers and politicians began campaigning for a more politically and economically unified Europe. European Union, The History of the European Union, https://europa.eu/european-union/about-eu/history_en (last visited June 17, 2022). The European Union was formally established in 1993 to establish a single market throughout member countries with free movement of goods, services, people, and money within that market, and in 1999, a unified currency, the euro. European Union, supra. they extend to national borders, but not beyond them. 55 Before the EU trademark system became operational in 1996, firms wishing to do business in multiple European nations were required to navigate a welter of different national trademark registration systems and trademark laws. 56 The Trade Mark Directive of 1988 sought to harmonize the various trademark laws of the individual EU member states, but it did little more than establish minimum standards in core areas of substantive trademark law and barely addressed procedural rules relating to trademark registration processes. 57 It represented a step forward, but was hardly sufficient. 58 Finally, the Community Trade Mark Regulation of 1993 established the Office for Harmonization in the Internal Market (now called the "EUIPO" 59 ) and a regime for the registration of "trade marks enabling the products and services of undertakings to be distinguished by identical means throughout the entire Community, regardless of frontiers." 60 The Trade Mark Regulation has been amended many times since 61 and now represents a stateof-the-art trademark registration statute.

*p. 18*
In effect, the EU trademark system is a microcosm of a global multilingual marketplace. 62 Many things that are true de facto of Vol. 112 TMR the global marketplace are true de jure for the unified EU trademark system. From the beginning, the EU trademark system has been a stunning success, at least as measured by the number of applications filed and registrations issued. 63 Figure 2 reports the annual number of applications filed at the EUIPO from 1996 through 2018 and the annual proportion of those applications that succeeded to registration through 2016. 64 By the end of 2018, the EUIPO boasted a total of 1,332,601 live registrations on its register. 65 The EUIPO and U.S. Patent and Trademark Office ("USPTO") report substantially different registration rates for the period 1996 through 2016. At the EUIPO, of the 1.7 million applications filed during this period, 89.2% registered. 66 At the USPTO, of the 5.6 million applications filed during the same period, only 55.8% registered. 67 As we explain below, two differences between U.S. and EU trademark law may largely explain this wide divergence in register it as a national-level mark at one or more of the national trademark offices of the member states, or they may do both. Id. As Rebecca Tushnet emphasizes, registration in the EU trademark system is treated as "controlling the scope of a registrant's right in an infringement case," which is quite unlike the U.S. trademark system, which does not treat a registration-or even the existence of one-as controlling the scope of a plaintiff's right in an infringement action. Rebecca Tushnet, Registering Disagreement: Registration in Modern American Trademark Law, 130 Harv. L. Rev. 867, 907 (2017) We do not report the registration-rate data past 2016 because some applications filed after 2016 may not have been fully processed by the end of 2018, when the data on which this figure is based were compiled. For this reason, subsequent figures that report registration rates stop at 2016.

*p. 20*
Vol. 112 TMR 919 registration rates: first, the USPTO engages in ex officio review of all applications for confusing similarity with already-registered marks whereas the EUIPO does not, and second, U.S. trademark law requires that marks be used in commerce before they may register, whereas EU trademark law imposes a lax use requirement. But before turning to these differences, we focus on the important "all or nothing" rule in EU trademark law. Under the "all or nothing" rule, if a mark is disqualified from protection in any part of the European Union, it cannot qualify as an EU trademark. 68 Thus, for example, if an applied-for mark is generic for its goods or services in any of the twenty-four official languages of the European Union (such as MILK or LAIT for milk), the EUIPO will refuse registration. 69 Similarly, if the applied-for mark is immoral or offensive in any official European language, the EUIPO will refuse registration, even if it is perfectly innocent in the 2) (forbidding registration of a mark "devoid of any distinctive character"). The applicant must instead resort to applying to the national offices of those nations, if any, whose residents would not perceive the mark as generic. Supra note 62.

*p. 21*
Vol. 112 TMR applicant's own language. 70 If a mark would be perceived as descriptive of its goods or services by any significant population within the European Union, including within individual member states, then it can qualify for registration in the EUIPO only if the applicant can show that the mark has acquired secondary meaning as a designation of source in the minds of that population. 71 In some ways, this "all or nothing" rule is a legal analogue of the branding maxim that a mark is competitively effective across all marketplaces or none at all. 72

## Confusing Similarity as a Bar to Registration

*p. 21*
The EUIPO will refuse to register any trademark that it determines to be confusingly similar to any mark that is already registered at the EUIPO. 73 But importantly, the EUIPO will make this determination only if a third party files an opposition to the registration of the mark. In significant contrast to the registration process at the USPTO and several other major national registration offices, 74 the EUIPO does not, on its own initiative, engage in socalled "relative grounds" examination of applications for confusing similarity with preexisting registrations. 75 Instead, it relies only on 70 E.g., Case T-526/09, Paki Logistics v. OHIM, 2011 E.C.R. II-0000 (denying registration to the German trademark PAKI for logistics on the ground that the term was a racial slur in English for persons of South-Asian origin).

## 71

*p. 21*
Case C-108/05, Bovemij Verzekeringen NV v. Benelex Merkenbureau, 2006 E.C.R. I-07605 (applying this rule with respect to the mark EUROPOLIS for insurance services, where "polis" in Dutch refers to an insurance agreement); Case T-219/00, Ellos v. OHIM, 2002 E.C.R. II-735 (denying registration to the Swedish word mark ELLOS for clothing, including clothing specifically for men, on the ground that the term in Spanish is the third-person plural pronoun referring to men and would therefore be perceived by Spanish-speaking consumers as descriptive of the goods); Kur, Dreier & Luginbuehl, supra note 63, at 183-84; Annette Kur & Martin Senftleben, European Trademark Law 118 (2017). Otherwise, the applicant must again resort to the national offices. 72 Supra section A.

## 73

*p. 21*
Trade Mark Regulation, supra note 62, arts. 8(1)-( 2). It will also refuse to register any trademark that it thinks is confusingly similar to a mark already registered at any of the national trademark registration offices of the EU member states. Id.

## 74

*p. 21*
In addition to the United States, other countries whose trademark office engages in ex officio review for likelihood of confusion include Australia, Brazil, China, Colombia, and Egypt. Philippe Bhering, Trademark Procedures and Strategies: Brazil, World Trademark Rev., Mar. 29, 2017, https://www.worldtrademarkreview.com/portfoliomanagement/trademark-procedures-and-strategies-brazil; Amir H. Khoury, The Development of Modern Trademark Legislation and Protection in Arab Countries of the Middle East, 16 Transnat'l Law. 249, 288 (2003); Wanhuida Peksung IP Group, Fourth Revision of China's Trademark Law, World Trademark Rev., May 20, 2019, https://www.worldtrademarkreview.com/fourth-revision-chinas-trademark-law; Grounds of Refusal in Australia, WIPO, https://www.wipo.int/export/sites/www/sct/en/ comments/pdf/sct21/ref_australia.pdf (last visited June 17, 2022); Colombia, 99 TMR 449, 463-64 (2009).

## 75

*p. 21*
Trade Mark Regulation, supra note 62, arts. 8(1)-(2). Other trademark offices that do not engage in ex officio review for likelihood of confusion include Germany, France, and Turkey. Ger. Patent & Trademark Office, Trade Marks: An Information Brochure on a third-party opposition process. 76 Upon receipt of a trademark application, the EUIPO will review it for compliance with various administrative formalities, ensure that it is not generic for or merely descriptive of any of the goods or services specified in the application, and then publish the application for opposition. 77 Prior registrants and other earlier rightsholders are then given three months to file an opposition to the registration of the applied-for mark on the basis, among others, that the mark is confusingly similar to an already-registered mark. 78 If no opposition is filed or is successful, the application proceeds to registration.

*p. 22*
Third parties must therefore be willing to spend the resources to continuously monitor applications at the EUIPO and challenge conflicting applications. 79 They cannot rely on the kind of initial ex officio review for confusing similarity that the USPTO undertakes, where a specialized division of trademark examiners make a firstand, as we show, incisive-cut of applications for confusing similarity before any are published for third-party opposition. 80 To be sure, in initially processing an application, the EUIPO generates a brief semi-automated search report listing potentially confusingly similar registrations and sends "surveillance letters" to the owners of those registrations notifying them of the application. 81 But these are of little practical value. Indeed, the search report is often comically inadequate. They nearly always report matches only for identical whole words. For example, the search report for Vol. 112 TMR CREMOLAIT, 82 now registered for foods-including milk products-and non-alcoholic drinks, does not list any potentially confusingly similar marks, including the many registered marks containing the word LAIT (meaning milk in French) in these same classes of goods. 83 In a reversal of the roles traditionally ascribed to the Americans versus the Europeans, the Americans rely on a government agency to do much of the heavy lifting in preserving the integrity of the trademark register while the Europeans leave it entirely to the "market to regulate itself and for applicants to 'have a go' at registering borderline or possibly invalid marks." 84 The EUIPO opposition process itself is straightforward. If an opposition is filed, a cooling-off period of at least two months commences in which the applicant and opposer may settle their dispute 85 -and as discussed below, though oppositions are rarely filed, when they are filed, settlements are common. 86 If the parties fail to settle, the EUIPO Oppositions Division then initiates a proceeding at the conclusion of which a three-member panel issues a reasoned decision either granting or denying the opposition. 87 In determining whether an earlier and a later mark are confusingly similar, the EUIPO assesses whether an appreciable proportion of relevant consumers would mistakenly believe that goods carrying the applied-for mark originate from or are commercially associated with the source of goods carrying the already-registered mark. 88 To make this determination of confusion as to source, the EUIPO considers a number of factors, most 82 EU Trade Mark No. W01009589.

## 83

*p. 23*
Office for Harmonization in the Internal Mkt., Community Search Report for Application IR 01009589, Aug. 12, 2009. Moreover, the EUIPO search report typically contains matches only for a subset of words in multi-word marks. For instance, the search report for GUCCI BAMBOO, EU Trade Mark No. 013688551, now registered, includes matches for marks containing the term GUCCI (happily, all the applicant's own marks) but not for the many marks containing the term BAMBOO. Office for Harmonization in the Internal Mkt., Community Search Report for Application 1149682, Jan. 3, 2013. We have not located information explaining why these search reports match for certain words within a mark and not others. For an example of a potentially conflicting mark consisting only of the word BAMBOO registered in Class 3 (cosmetics), see EU Id. arts. 8, 47(5). In guiding this inquiry, the EU Trade Mark Regulation explicitly provides that "the likelihood of confusion includes the likelihood of association" between two marks, id. arts. 8(1)(b), 9(2)(b) (emphasis added), which can broaden the confusion analysis considerably.

*p. 24*
importantly, the similarity of the two marks and the relatedness of the goods with which they are used. 89 The process evaluates mark similarity along the three dimensions of visual, aural, and conceptual similarity (comparable to the familiar American trinity of "sound, sight, and meaning"). 90 Over time, EU trademark law has developed various doctrines to come to terms with the multilingual nature of the European market. For example, in assessing word similarity, diacritical marks are generally disregarded; thus, the marks UBER and ÜBER would likely be considered to be essentially identical. 91 More significantly, under conceptual similarity, two orthographically different marks from different languages may be considered to be confusingly similar if they convey the same meaning to any significant population of consumers capable of understanding the terms in both languages. 92 We refer to this form of conceptual similarity as "translational similarity," a principle applied in U.S. trademark law as well. 93 The "all or nothing" principle further dictates that if a significant population of consumers anywhere in the European Union would recognize the OHIM Opposition Division Decision No. 3111/2000 (Dec. 21, 2000) (refusing registration of the English-language mark RED BARON with respect to certain goods on the basis that it was confusingly similar with the Spanish-language mark BARON ROJO); OHIM Opposition Division Decision No. 131/1999 (March 25, 1999) (refusing registration of the English-language mark 5 OCEANS on the basis that it was confusingly similar with the Spanish-language mark CINCO OCEANOS); Examination Guidelines, supra note 91, § 3.4.3.1; Fhima & Gangjee, supra note 89, at 58-61; cf. Case C-603/14, El Corte Inglés v. OHIM (Dec. 10, 2015) (holding that the English-language mark THE ENGLISH CUT was not sufficiently similar to the Spanish-language mark EL CORTE INGLÉS to confuse consumers, but it may be sufficiently similar to result in the dilution of the latter mark); Case T-534/10, Organismos Kypriakis Galaktokomikis Viomichanias v. OHIM (June 13, 2012) (finding conceptual similarity between the mark HALLOUMI in Greek and HELLIM in Turkish on the ground that because Turkish and Greek are official languages of Cyprus, Cypriots will recognize the common meaning of both terms as referring to a type of cheese). But see Case T-437/11, Golden Balls Ltd v. OHIM (Sept. 16, 2013) (finding "at most, a weak conceptual similarity" between GOLDEN BALLS and BALLON D'OR); Examination Guidelines, supra note 91, at § 3.4.4.2 ("As it is the actual understanding of the relevant public that matters, the mere fact that one term is objectively the foreign-language equivalent of the other may not be relevant at all in the conceptual comparison."). Interestingly, the EUIPO examination guidelines recognize that in certain instances, "a significant part of the relevant public may have only a limited command of the relevant foreign language and, therefore, might not be able to distinguish the difference in meaning between two expressions." Examination Guidelines, supra note 91. In such instances, consumers may be more likely to confuse the two terms because of their lack of sophistication in the language. Id.

## Vol. 112 TMR

*p. 25*
translational similarity between an earlier and a later mark and the two marks are used on related products, then the later mark cannot qualify for EU-wide registration. The result is that the registration of a word or phrase in any of the major European languages may conflict with any later application for any word or phrase that conveys the same or a similar meaning in any other European language, when the two marks' goods are related. This is especially true for English, which is widely spoken throughout Europe. 94 For example, if there exists an earlier registration for DOG in connection with apparel, that registration would almost certainly conflict with any later application for the equivalent term in any European language in connection with apparel. The same is likely true if the registration were for HUND, CHIEN, CANE, or PERRO in connection with apparel. The reasoning is that in each case there exists some significant population of consumers somewhere in Europe who would associate the mark in the major European language with the equivalent term in their first language. By contrast, a registration for ΣΚΎΛΟΣ might not conflict with a registration for SUNS ("dog" in Greek and Latvian, respectively) because there may be no significant population of consumers in Europe who understand even basic terms in both languages. Registrations of terms in the major EU languages-again, especially English-are thus quite powerful. They potentially block visually, conceptually, and aurally similar words in other languages, including translationally equivalent words in those languages.

## The Use Requirement

*p. 25*
EU trademark law fundamentally differs from U.S. trademark law in another respect: EU law allows the registration of marks that the registrant is not actually using in commerce. 95 An EU trademark registrant enjoys a five-year grace period from the date of registration to make a "genuine use" of its mark for the goods or services specified in the registration. 96 After this grace period has elapsed, third parties (but not the EUIPO itself) may challenge the registration on the basis of non-use. 97 If no third party institutes or is successful in such a challenge, the registration will remain in effect and may be renewed indefinitely, even if the registrant never in fact makes a genuine use of its mark. The American system, by 94 In 2012, 51% of EU citizens spoke English, and the current percentage is likely much higher. Dave Keating, Despite Brexit, English Remains the EU's Most Spoken Language by Far, Forbes, Feb. 6, 2020, https://www.forbes.com/sites/davekeating/2020/02/06/ despite-brexit-english-remains-the-eus-most-spoken-language-by-far. contrast, is emphatically a use-based system in which a firm's actual use of the mark in commerce forms the basis for trademark rights and registration. 98 Except in the case of International Registration designations, the USPTO insists on proof of use for a registration to issue and every time the registration is renewed. 99 In principle, registration at the USPTO is merely a recordation of the preexisting rights created by use. 100 As one study shows, third-party challenges at the EUIPO on the basis of non-use are "exceedingly rare." 101 They arise in opposition proceedings, when the opposed applicant will challenge the opposer's earlier registration on the basis that the opposer has not made a genuine use of its earlier-registered mark. But even in this context, a recent study has shown that as many as one-third of opposers base their oppositions on earlier registrations that are still within the five-year grace period and thus cannot be challenged on grounds of non-use. 102 Indeed, though the EUIPO has sought to limit the practice, sophisticated filers still often maintain a series of temporally overlapping new registrations of their mark, sometimes in classes in which they have no intention to use the mark, in order to benefit from the grace period and prevent others from adopting the mark. 103 The combined result of EU trademark law's lenient use requirement and registrants' exploitation of the five-year grace period is that the EUIPO register has a significant problem of "trademark clutter." 104 "Clutter" refers to registrations for marks that the registrants are not using in commerce. 105 Such 98 15 U.S.C. § 1051(a)(1) (providing for registration of a mark "used in commerce"); id. § 1051(d) (providing for registration of a mark filed on an intent-to-use basis upon filing of a statement that the mark is "used in commerce").

## 99

*p. 26*
The USPTO insists on proof of use before registration, id. § 1051(a)(1); 37 C.F.R. § § 2.34(a)(1)(iv), 2.56(a), in the sixth year of registration, 15 U.S.C. § 1058, and in every tenth year of registration, id. 103

*p. 26*
Id. at 3 (referring to "a stream of follow-on registrations that exist primarily to ensure that their core brands are always linked to a registered mark falling within the grace period").

*p. 26*
104 von Graevenitz, Ashmead & Greenhalgh, supra note 101. Almost certainly contributing to clutter on the EU trademark register is the flat fee until 2016 for filing a trademark for three classes, which has since been replaced with a graduated fee system to discourage prolific filings. Infra note 111.

## Vol. 112 TMR

*p. 27*
registrations represent a barrier to the adoption of the unused but registered marks by other entities.

## The Nice Classification Scheme

*p. 27*
As with the USPTO, each trademark application to the EUIPO must indicate the goods and services for which the applicant seeks to register its mark. 106 The applicant must provide a written description of these goods and services and further indicate where they are classified among the forty-five classes of the Nice Classification. 107 The Nice Classification is intended in theory to serve "only administrative purposes"; 108 EU trademark law explicitly states that the classification scheme should have no bearing on the office's evaluation of the relatedness of any goods or services. 109 Nevertheless, trademark lawyers routinely use the Nice Classification as a heuristic for evaluating relatedness and it remains the standard index used by researchers seeking to understand the operation of the global trademark system. 110 Figure 3 shows the number of active EU trademark registrations in each Nice Class in 2018. Consistent with data from the USPTO, certain classes are heavily populated, such as Class 9 (electronics goods), covered by 26.3% of all active registrations, and Class 35 (general business administrations services), covered by 23.3% of all active registrations. Also of significant interest is Class 25 (apparel), covered by 12.2% of all registrations. In part because of the lax use requirement and relatively low per-class registration fees, individual EU trademark registrations commonly cover multiple Nice classes. 111 Live registrations in 2018 covered an average of 2.7 Nice classes (SD=2.3), with 66.1% covering more than one class. Of Vol. 112 TMR 929 two goals, and their pursuit of the first-of an effective European trademark-has increasingly made the pursuit of the second-of an effective European trademark that has not yet been claimed-all the more difficult. We turn now to quantitative measures of just how difficult this pursuit of an unclaimed but commercially effective EU mark has become. We begin with a brief description of the datasets that are the basis of these measures.

## II. THE DATASETS

*p. 30*
We rely primarily on the EUIPO Open Dataset, which the EUIPO first made available in May 2017 and updates regularly. 115 The dataset consists of partially anonymized information on each of the 1,860,561 trademark applications submitted to the EUIPO from January 1, 1996, when the EUIPO began to accept applications, through 2018.

*p. 30*
We developed two additional datasets to fill gaps in the Open Dataset. First, the Open Dataset does not specify the language and meaning of any typographical words or characters appearing in an applied-for mark. Applicants are not required to indicate the language, if any, of the mark, nor are they required to provide translations of the mark into any languages. The EUIPO does not itself add this information. Because of the importance of translational similarity to our understanding of trademark depletion and crowding, we used the Google Translate API 116 (application programming interface) to detect the source language of any mark that included typographical characters and to translate, where possible, the mark into English, German, French, 115 EUIPO, Anonymized Dataset (Open)

## Vol. 112 TMR

*p. 31*
Italian, and Spanish, 117 which are the five major European languages. 118 Second, the Open Dataset lacks detailed information on the opposition history of the 255,825 trademark applications submitted to the EUIPO from 1996 through 2018 that received oppositions. We therefore developed a dataset consisting of opposition data on the subset of 88,798 trademark applications filed from 1996 through 2018 that were opposed and that led to a decision by the EUIPO Oppositions Division. 119 We did so by systematically searching the EUIPO's online database of Oppositions Division decisions. 120 These data include the opposing mark, the statutory bases for the opposition, the classes with respect to which the opposition was filed, and the outcome of the decision.

*p. 31*
To develop our word-frequency data in the five major European Union languages, we primarily relied on the corpora listed in Table 1. For each language, we limited our analysis to the 20,000 most frequently used non-proper-noun words in the language, both because of the significant computational resources required to conduct word-similarity analyses among millions of words and because in each language, the 20,000 most frequently used words accounted for a very high proportion of overall word usage, on the order of 85% to 95%. As Figure 5 Where we rely on these data, we explain how we have adapted our research methods to these limitations. 118 These languages are five of the six most widely spoken by mother tongue in the European Union (the other being Polish). Eur. Comm'n, Europeans and Their Languages 10 (June 2012), https://op.europa.eu/en/publication-detail/-/publication/f551bd64-8615-4781-9be1-c592217dad83. Additionally, English, German, French, and Spanish are the four most widely spoken foreign languages of those in the European Union. Id. at 19-20. English, German, French, Spanish, and Italian are also thought by Europeans to be the most useful European languages in that order. Id. at 69. Finally, the EUIPO has adopted these languages as its five working languages. Trade Mark Regulation, supra note 62, arts. 146(1)-(2). 119 Oppositions to the remaining 167,027 opposed applications were apparently resolved before the Oppositions Division issued a decision. Of these 167,027 applications, 68.9% proceeded to registration and the remainder failed to register. consistent with Zipf's law, in which word frequency follows a power law distribution. 121 For example, the Lexique 3 corpus indicates that the ten most frequently used non-proper-noun words in French account for 19.0% of overall word usage; the 100 most frequently used account for 51.1% of usage; the 1,000 most frequently used account for 71.2% of word usage; and the 10,000 most frequently used account for 85.6% of word usage. The other corpora each yielded strikingly similar results. As these data suggest, exclusive rights in marks that consist of high-frequency words are especially powerful and can exert an outsized impact on competition. To conduct a sentiment analysis of those frequently used English words that are registered as trademarks and those that remain unclaimed, we used the Harvard General Inquirer dataset of words coded for positive and negative affect. 122 For purposes of studying the proportion of frequently used words in the five major European languages that are already registered as domain names in the .com top-level domain, we gained access to and Subtitles, 32 Psicológica 133 (2011). As with the SUBLTEX-DE data, supra note 125, the SUBLTEX-ESP data did not include partof-speech information, so it was not possible based only on the SUBLTEX-ESP data to filter out proper nouns. We therefore used the TenTen Spanish Web Corpus 2018 to establish a list of the 20,000 most frequently used words in Spanish according to that corpus. Sketch Engine, esTenTen-Spanish Corpus From the Web, https://www.sketchengine.eu/estenten-spanish-corpus (last visited June 17, 2022). We then combined this list with the proportion-of-usage data from the SUBTLEX-ESP corpus to create the frequency data underlying our results. We compared our main results based on word-usage data from the SUBTLEX-ESP corpus to results based on word-frequency data drawn from the TenTen Spanish Web Corpus 2018. Id. We found no substantial differences.

## III. TRADEMARK DEPLETION IN A GLOBAL MULTILINGUAL ECONOMY

*p. 35*
It is a common refrain in the popular press that nearly all the good brand names are already taken and that creativity in branding is now largely an exercise in finding and making the best of whatever is left. 132 Of course, new brands emerge every year that seem to belie this conventional wisdom, making it appear obvious in retrospect that the previously unclaimed term TWITTER is an ideal name for a social media network or WHATSAPP for an online messaging service. Meanwhile, law-and-economics orthodoxy insists that the supply of competitively effective marks is theoretically and therefore practically inexhaustible. Yet despite the periodic success of new superstar brands (which through salience bias are mistaken as representative examples) and despite law-andeconomics dogma, the prevailing view in the marketing world is that all the most fertile land has already been claimed and the new settler is operating at a disadvantage from the start. The frontier is effectively closed.

*p. 35*
In previous work, we empirically studied the degree of trademark depletion in the U.S. trademark system, finding astonishingly high and worsening levels of depletion for many categories of competitively effective trademarks. 133 Showing that the media reports are more accurate than the law-and-economics view, we recommended decreasing-or at least slowing downtrademark depletion for the harms it poses to the trademark system. 134 In this Part, we confirm that our previous findings of severe depletion in the United States extend also to the European Union. We then move well beyond that work and show how the multinational and multilingual nature of the EU trademark system leads to yet further depletion than might otherwise exist.

*p. 35*
Combined with the wealth of the European market, the massively multinational and multilingual nature of the EU trademark system sets it apart from all other trademark systems in the world. 135 This makes the study of the EU system uniquely There are a few other, less developed regional trademark systems, ranging in their degree of harmonization and integration, including two in Africa, one in Southeast Asia, and one in South America. Irene Calboli & Coenraad Visser, Regional Trademark Protection: Comparing Regional Organizations in Europe, Africa, South East Asia, and South America, in The Cambridge Handbook of International and Comparative Trademark Law 103 (Irene Calboli & Jane C. Ginsburg eds., 2020). On the linguistic front, the Indian constitution recognizes twenty-two official languages in addition to English. India Const. sched. 8. The Indian trademark register also boasts a large number of applications and registrations. For example, in 2018, it counted a total of 1,904,698 interesting for purposes of understanding the global trademark system. But it also makes the study of the EU system uniquely difficult. We first set out a framework for analyzing and quantifying trademark depletion in such a system. In this connection, we address what we call the "denominator problem": to estimate the proportion of competitively effective marks that have already been claimed, we need some measure of the total population of such marks. We also consider how trademark depletion works in a trademark system that recognizes translational similarity across numerous languages. We then turn to our results. We first focus on the extent of trademark depletion among frequently used English words and possible coinages pronounceable by English speakers. We focus first on English because it remains by far the most important commercial language in the EU market. It is also the most severely depleted-at levels comparable to the results we reported in our study of the U.S. trademark system. Our results also show severe depletion across the other four major European languages. Translational similarity significantly exacerbates levels of trademark depletion across the five major European languages. Finally, we analyze the depletion of "multilanguage words," the shift in applicant behavior at EUIPO toward applications for coined terms, and the extent of domain name depletion in the five major languages.

## A. A Framework for Assessing Word-Mark Depletion in a Multilingual Trademark System

*p. 36*
In previous work, we defined trademark depletion as the process by which a decreasing number of competitively effective trademarks remain unclaimed by any trademark owner. 136 For purposes of this study, we define a potential mark as unclaimed when it is not identical or closely similar to a mark that is the subject of a registration at the EUIPO. As explained above, EU trademark law assesses similarity phonetically, orthographically, and conceptually, including translationally. 137 The result is that a single trademark registration will, in effect, deplete the broader set of marks closely similar to the registered mark in sound, sight, meaning, or translation. Trademark depletion is a slippery concept, and the study of it in the multilingual context raises a number of challenges, three of which we address here.

## The Effects of Trademark Depletion on Market Entrants

*p. 37*
The first challenge is specifying what the consequences are for other firms when a registrant claims a mark and thus depletes it and closely similar marks from the stock of unclaimed marks. The most likely consequence is that the registration will deter most other firms from using any of the marks covered by the registration on goods or services closely related to those specified in the registration, and if a firm nevertheless does so, trademark law may enjoin its use. This is because such a use would likely lead consumers mistakenly to believe that the two firms' closely related products bearing closely similar marks come from the same source. Trademark law is designed to prevent precisely this form of consumer confusion as to source. Even the mere possibility of legal action on confusion grounds may drive risk-averse firms to seek an alternative mark. Furthermore, regardless of the threat of legal sanction, entrants may have a genuine interest in avoiding the possibility that their products would be confused with the registrant's and may on that basis choose a different mark. Finally, as discussed above, a firm may wish to adopt a mark that is unique and maximally distinctive as compared with all other marks in the marketplace or at least as compared with all other marks in the firm's particular product sector. 138 Though a trademark registration thus imposes a significant barrier to the adoption by others of any of the marks covered by the registration, it is important to emphasize that this barrier is not insurmountable. Even if a firm has registered a word mark in a particular Nice class, it is possible that another firm may register the same or a closely similar mark in the same or another Nice class. Courts may find no likelihood of confusion or the registrant may simply not bother to assert its exclusive rights. The result would be multiple firms using closely similar marks on closely similar products. Indeed, we explore this phenomenon of trademark crowding in Part IV. But depletion deters and often prevents other firms from adopting any mark in the set of marks covered by the registration on goods or services related to those specified in the registration. For this reason, our results focus here on depletion within particular Nice classes of goods or services. Depletion may also deter or prevent firms from adopting marks even in situations in which doing so would not cause consumer confusion. Firms may be concerned instead about choosing a unique mark. For this reason, we also report our findings on depletion regardless of Nice classes, with respect to the overall marketplace.

## The Denominator Problem

*p. 38*
A second challenge that the study of word-mark depletion poses is determining the magnitude of the overall population of all possible competitively effective word marks. Depletion is important only when a sufficiently high proportion of such marks has been claimed that market entrants, lacking viable alternatives, are put at a significant competitive disadvantage. 139 Determining this proportion requires some estimate of the denominator, namely, the number of plausible competitively effective marks.

*p. 38*
The problem is that there is no good way to establish the sum total of all possible competitively effective word marks, including all already-existing words and all viable coinages. The primary difficulty is that there is no clear standard for determining which words qualify as competitively effective and for which categories of goods or services. APPLE may be a good trademark for hightechnology products, but not at all good for a business selling pears.

*p. 38*
Assessing competitive effectiveness across multiple languages and cultures adds additional complexity. 140 Based on our discussion above about brand selection in a global multilingual marketplace, even if one could establish the sum total of all possible competitively effective marks in any one language, it is even less straightforward to do so across multiple languages. That is, one cannot simply sum up the competitively effective marks in each language to derive the number of competitively effective marks globally. In particular, a mark might be competitively effective in one or more languages, but particularly ineffective-and even pernicious-in another language and thus entirely ineffective for a business choosing a mark to use across multiple jurisdictions. 141 Therefore, one would need to remove all marks that are competitively ineffective in any relevant jurisdiction from a list of globally competitively effective marks. The same holds true for a mark that is competitively effective and available in some jurisdictions but unavailable (because it is already claimed or because it is not protectable) in at least one jurisdiction.

*p. 38*
Our approach to this denominator problem has been to focus primarily on major categories of desirable trademarks, principally frequently used words and short, pronounceable neologisms. 142 We also look at probable symptoms of depletion, such as trademark crowding rates, refusal rates, and opposition rates. Admittedly, this is not ideal, and is akin to looking only under the streetlights

## Vol. 112 TMR

*p. 39*
because it is easiest to look there. 143 In that sense, it both understates and overstates the denominator: understates by looking only to these well-defined categories and overstates by including terms that would never be considered competitively effective. It also does not address the more complex problem raised by global multilingual trademark systems by overstating the number of competitively effective marks, as just discussed. However, the common characteristics of effective brands that we reviewed above suggest that this is the right place to start. 144 In general, firms prefer to use familiar, frequently used words, and in a multilingual marketplace, they especially prefer words that are frequently used in multiple languages. If they resort to neologisms, they prefer that their marks be easily pronounceable and perhaps evocative of and thus similar to more common words. These factors have guided our approach. If we can show that there is significant depletion of words that share these characteristics, we believe that is strong circumstantial evidence of a broader problem.

*p. 39*
Finally, it is worth emphasizing that trademark depletion is a chronic rather than a critical condition in a trademark system. 145 It is an incremental process that proceeds along a continuum. We should not expect depletion to reach some point where all viable marks have been claimed and market entry under a unique brand name becomes impossible. Nor should we look for some clear qualitative shift where the extent of depletion, though not yet total, has suddenly become unmanageable. 146 Instead, trademark depletion gradually makes it more difficult for firms to find a competitively effective mark that has not yet been claimed. Their clearance searches grow longer and costlier. The marks they eventually settle upon seem as a whole ever less compelling, and media reports and sharp-eyed consumers increasingly begin to notice new brands with names that are inexplicably ridiculous-like BLIND PIG and PERMANENT FUNERAL for beersfoot_10 and YERVOY, VIIBRYD, and ZYTIGA for pharmaceuticalsfoot_11 -or banal-such as TRUIST FINANCIAL for the merger of BB&T Corp. and SunTrust Banks 149 and PEARL MILLING COMPANY for the rebranding of the racist AUNT JEMIMA breakfast foods. 150 This continuous process resists categorical measurement. But we think it is enough to offer some estimate of the current level of depletion in the EU trademark system and assess whether the problem is worsening and at what rate.

## Cross-Language Depletion

*p. 40*
A final challenge of studying depletion in a multilingual trademark system like the European Union's is assessing how depletion operates across multiple languages. As explained above, the registration of a single word mark in, say, English can result in a scope of protection beyond the word in question to all English and non-English words that look and sound like that word or are closely similar to it in appearance, sound, meaning, or translation. 151 Thus, the mark JOLLY in English would likely cover, for example, JOLLIFY in English, JOLIE in French ("beautiful"), and perhaps also JULI in German ("July"), at least when the two marks are used on related goods or services. The registration may also extend to any non-English words that convey a closely similar meaning in a language, a significant number of whose users would recognize the similarity in meaning, such as FRÖHLICH in German or ALEGRE in Spanish.

*p. 40*
For this study, we rely primarily on orthographic and translational similarity across languages to determine which potential word marks qualify as identical or closely similar. With respect to orthographic similarity, we use the Jaro-Winkler measure of similarity, which is based on the edit distance between two strings of characters-that is, the number of edits by character required to transform one string into another. 152 Importantly, the Jaro-Winkler algorithm places more weight, as the EUIPO does in its assessment of mark similarity,foot_12 on the similarities among the initial characters of the strings being compared. 154 The algorithm produces a score that is normalized to range from 0 (no similarity) to 1 (identical similarity).

## Vol. 112 TMR

*p. 41*
between JOLLY and JOLLIFY is 0.943, between JOLLY and JOLIE is 0.813, and between JOLLY and JULI is 0.670. As in previous work, we coded two terms as closely similar if they produced a Jaro-Winkler similarity score equal to or higher than 0.875. 155 This is a conservative measure of similarity, as the examples just offered suggest and our EUIPO Opposition Division data confirm. 156 To address translational similarity across languages, we also took a conservative approach to minimize false positives. We proceeded from the assumption that with respect to any two of the five major European languages, there exists a significant population in Europe that speaks both of them. We therefore assumed for purposes of assessing translational similarity that our baseline average consumer spoke all of these five languages (and only these five languages). Thus, the registration of a word mark in any of the five languages would likely also claim the translationally equivalent words in each of the other major languages. On this basis, we classified a word in one of the five languages as depleted through translational similarity if it was identical to the translation, if any, into the word's language of a mark registered at the EUIPO when that registered mark also came from one of the five major languages. This approach significantly underestimates depletion through translational similarity because it omits languages outside of the five major languages and thus omits translational similarity between a major language (for example, English) and any minor language or among minor languages. 157 But as we show in a moment, even this approach reveals extraordinarily high levels of depletion through translational similarity in each of the five major languages.

## B. General Trademark Depletion

*p. 41*
We study general trademark depletion in the EUIPO in English, French, German, Italian, and Spanish. 158 We find that very high proportions of word usage in each of these languages consist of 155 Beebe & Fromer, supra note 8, at 991. 156 By comparison, of the 47,991 EUIPO Opposition Division decisions that involved opposing word marks and in which the Division sustained the opposition at least in part, the mean Jaro-Winkler score of the opposing marks was 0.796 (SD=.213). Of the 30,098 decisions in which the Division denied the opposition in full, the mean Jaro-Winkler score of the opposing marks was 0.699 (SD=.282).

## 157

*p. 41*
Our approach is also conservative because it finds translational similarity only when there is an identical match between the translation into a particular language of the registered mark and the frequent word in that language. Thus, a trademark such as SCHNELL in German is arguably translationally similar to potential English word marks such as FAST, QUICK, RAPID, and SWIFT. But the Google Translate API returns only a single result, the word FAST. On that basis, our algorithm would indicate that SCHNELL in German depletes only FAST in English but not the alternative translations QUICK, RAPID, and SWIFT.

*p. 42*
Vol. 112 TMR 941 words that are identical or closely similar to marks registered at the EUIPO. We find comparable results for coined terms that are possible in the English language. We then study the particular issues depletion raises in a global multilingual trademark system. We focus primarily on translational depletion and depletion of multilanguage words.

## Identical Matches

*p. 42*
Consider first English. A very high proportion of English word usage consists of words registered as single-word marks at the EUIPO. As Table 2 indicates, of the 1,000 most frequently used English words, 756 were the subject of active registrations in 2017, and these 756 high-frequency words account for 69.4% of all word usage. 159 Of the 20,000 most frequently used words, 8,133 were registered, accounting for 77.3% of all word usage. 160 In effect, when we use English, more than three-quarters of the time we are using a word that identically matches a registered trademark at the EUIPO. The solid black line in Figure 6 shows how quickly this state of affairs came to pass.

## Vol. 112 TMR

*p. 43*
Proportion of Word Usage Consisting of Words Identically Matching Registered Trademarks at EUIPO, by Language, 1996-2017 What English words are actually left? Focusing on the subset of the 1,000 most frequent English words offers some insight into the general characteristics of those English words that remain unregistered as single-word marks. Of the 244 words that are still unclaimed among the 1,000 most frequent, many remain unregistered for obvious reasons. It is hard to imagine that any of the following words would make effective brand names: AFRAID, BEHIND, DIFFICULT, KILL, KILLED, LOSE, POOR, PROBLEM, PROBLEMS, SORRY, TRYING, USED, WEAKEST, WORRIED, WORRY, WORSE, and WORST. Indeed, Table 2 reports the results of a simple sentiment analysis of frequent words that are already registered as compared with those that have not been registered (with words carrying positive affect coded as 1, those carrying negative affect coded as -1, and neutral words coded as 0). These results are consistent with a more impressionistic perusal of the lists of unclaimed words. Few are commercially viable brand names because of their unfavorable meanings or associations. Those words that are viable are already taken.

*p. 43*
Figure 6 also shows for each of the other four major European languages the proportion over time of word usage consisting of words that identically matched a mark registered at the EUIPO. 161 The major Romance languages each report substantial levels of depletion.

## Vol. 112 TMR 943

*p. 44*
By 2017, 55.4% of French word usage consisted of words identically matching a registered mark. The results for Italian (65.7%) and Spanish (62.8%) are higher. And as with English, an admittedly impressionistic perusal of those words that remain unclaimed in the three languages often shows why they have not been registered as trademarks for any goods or services. For example, of the 511 words that remain unregistered among the 1,000 most frequently used French words, the following are typical: DÉSOLÉ (sorry), DIFFICILE, FAIM (hunger), FATIGUÉ (tired), HONTE (shame), PAUVRE (poor), PEUR (afraid), PRESQUE (almost), PROBLÈME, and TUER (kill). German, by comparison, is in better shape, with only 46.2% of word usage consisting of words identically matching a registered mark in 2017. 162 We additionally studied trends over time in the length of newly applied-for marks. We found clear increases over time in mark length as measured by character count (from a mean character count of 10.4 characters in 1996 to 12.15 in 2017) and word count (from a mean word count of 1.6 words in 1996 to 1.9 in 2017). Increases were more pronounced for English-language marks, where mean character count increased from 10.9 characters in 1996 to 13.2 in 2017 and mean word count increased from 1.7 words in 1996 to 2.0 in 2017. For English-language marks, syllable count also increased substantially, from a mean syllable count of 3.5 syllables in 1996 to 4.2 in 2017. (Consistent with Mark Twain's observations, German-language marks tend to be longest, but overall they have not increased appreciably in length over time. Mark Twain, The Awful German Language, in A Tramp Abroad 538, 546 (1889) ("Some German words are so long that they have a perspective.").)

## Vol. 112 TMR

*p. 45*
The data show that in each of the five languages, registrants initially rushed in to register the highest-frequency words, which explains why the proportion of word usage covered by registered marks initially increased steeply. Registrants then continued each year to newly register frequently used words, but because the highest-frequency words tended already to be taken, they increasingly resorted to settling for lower-frequency words. This explains why, in Figure 6, the proportion of word usage covered by registered marks flattens over time for each language. Figure 7 depicts this process in more detail for English. For comparison purposes, the solid black line in Figure 7 replicates the solid black line in Figure 6. The bars show the number of the 20,000 most frequently used English words identically matching a live registration over time. By this measure, there is a relatively steady increase over time in the absolute number of the 20,000 most frequently used English words claimed by registrants. In other words, there was initially something akin to a land rush at the EUIPO, but it took the form of a rush to claim the highest-frequency words in English. We find similar trends in the other four major European languages.

*p. 45*
We have focused so far on the proportion of frequent words that are registered in any class of goods or services. We have done so because, as explained above, a firm would ideally prefer to be the sole user in the marketplace of a particular term, thus significantly enhancing the uniqueness of its brand. 163 But as also explained, trademark law will allow parallel uses by different firms of the same term as a trademark provided that in doing so they do not confuse consumers as to source. 164 Thus, AAA as registered by the American Automobile Association at the EUIPO in various Nice classes 165 can coexist with AAA as registered by Whirlpool in other classes. 166 To address the possibility of parallel uses of the same term by different firms in different classes, Figure 8 shows by major language for each Nice Class the proportion of word usage in the language made up of words that match a term actively registered in that class in 2017. Even when breaking down the data by class, we see in each language except German high levels of depletion in important classes, such as Class 9 (electronics goods), Class 25 (apparel goods), Class 35 (general business services), Class 41 (cultural services), and Class 42 (high-technology services). 167 Comparable class-specific results for active registrations at the USPTO in 2017 were lower in each class when measured by proportion of word usage. Across all 45 Nice classes and limited only to identical non-translational matches, the average difference in the proportion of word usage covered by marks registered in the class at the EUIPO words that remain unclaimed in particular classes tend as a general matter to be those that would be less competitively effective as trademarks. For example, in Class 25, of the 5,000 most frequent English words, 1,264 were actively registered in 2017, accounting for 51.7% of word usage and with a mean sentiment score of 0.043, as compared with the 3,736 that remained unregistered, with a mean sentiment score of -0.032. 168 To get some sense of the degree to which even coined word marks are depleted at the EUIPO, we focused on coinages pronounceable at least by English speakers. We further focused on potential single-syllable coinages in part because firms prefer shorter word marks. We compared phonetic representations of all unique syllables used in English to phonetic representations of all English-language single-word word marks registered at the EUIPO from 1996 through 2018. We found that by 2018, 80.0% of all syllable usage in English consisted of syllables that were the subject of single-word English-language trademark registrations at the EUIPO. Certain Nice classes were especially depleted. For example, 66.6% of English syllable usage is claimed by single-syllable Englishlanguage marks registered in Class 9 (electronics goods) and 58.3% in Class 25 (apparel goods). Our approach focuses only on English-language syllables. We anticipate that a more rigorous analysis that incorporates frequently used syllables common to at least the four other major European languages would show substantially more advanced levels of depletion.

## Close Similarity

*p. 47*
New market entrants may face significant barriers to adopting a particular word as a trademark not just if it identically matches an already-registered mark, but also if it is closely similar though non-identical to an already-registered mark. 169 To estimate the proportion of frequently used words in each of the five major languages that are closely similar to already-registered marks, we calculated for each language the Jaro-Winkler similarity scores between each of the 20,000 most frequently used words in the language and each of the 1,247,549 marks registered at the EUIPO in 2017.

*p. 47*
Focusing first on English, the results of our Jaro-Winkler similarity analysis indicate severe depletion of the stock of common English words that are not closely similar to an already-registered mark. Of the 20,000 most frequently used English words, only 59 were not closely similar to a registered mark at the EUIPO in 2017.

## 169

*p. 47*
The Court of Justice of the European Union has ruled that mark similarity-visual, aural, or conceptual-is a critical component of assessing the likelihood of confusion. Case C-251/95, SABEL BV v. Puma AG, Rudolf Dassler Sport, 1997 E.C.R. 528. Because consumers might pay more or less attention to the subtleties of marks in particular contexts, close similarity might often result in confusing similarity, but not always. Fhima & Gangjee, supra note 89, at 17-66.

*p. 48*
The remaining 19,941 words that were closely similar account for 94.8% of English word usage. In effect, it is essentially no longer possible for an entity to adopt a commonly used English word as a trademark and expect to be the only user of that mark in the European marketplace. Moreover, in those cases in which the mark would be confusingly similar to another registered mark, the entity might face significant barriers to registering it if it wanted to do so. Even when focusing on particular classes of goods or services, the availability of sufficiently dissimilar marks is severely limited. As Figure 9 shows, a significant number of Nice classes show levels of depletion amounting to over 90% of word usage. For example, 94.6% of word usage consists of words closely similar to a mark already registered in Class 9 (electronics goods), and the statistic is 94.5% for Class 35 (general business services) and 93.7% for Class 25 (apparel goods).

*p. 48*
Jaro-Winkler similarity analyses for the other four major languages yield similarly disturbing results. Of the 20,000 most frequently used French words, 19,854 were closely similar to a mark registered at the EUIPO in 2017, accounting for 88.2% of French word usage. For German, 17,914 of the 20,000 most frequently used words were closely similar to a registered mark, accounting for 82.6% of German word usage. Italian and Spanish show similar results. 170 In each of these languages, market entrants will almost certainly fail to find a frequently used word in the language that is not closely similar to an already-registered mark. As Figure 9 shows, for the other four major languages, depletion is found in nearly every Nice class-with the exception of some of the more eccentric Nice classes, such as Class 13 (firearms and explosives), Class 15 (musical instruments), and Class 23 (yarns and threads).

## Vol. 112 TMR

*p. 49*
Figure 10 shows for certain Nice classes the rapid depletion over time in the stock of common words in each of the five languages not closely similar to a mark already registered in that class. By 2010, depletion through close similarity had essentially hit its maximum possible extent in each of the five major European languages, at least with respect to the 20,000 most frequently used words.

## C. Translational Trademark Depletion

*p. 51*
The picture painted in the previous section shows relatively severe general trademark depletion across the five major European languages. Yet it does not account for translational similarity. In this study, we conservatively assume that a significant population of EU consumers is capable of understanding any two of the five major European languages. 171 On that basis, any term that is registered in, say, English would be unavailable for registration were it translated into French, German, Italian, or Spanish. Registration in any of these five languages thus also depletes the translation of those words into the other four languages, even if those translations are not actually registered. Once we account for translational similarity in our study of depletion, a major feature of a global multilingual system, the degree of trademark depletion is significantly worse. We call this increased degree of depletion "translational depletion." This yields the "reverse Babel problem," in that this scenario represents the reverse of the lack of understanding between everyone, each speaking a different language, in the Biblical story of the Tower of Babel. 172 Our findings show that translational depletion is so significant that even languages in which there is less trademark depletion as measured by identical matching, such as German, 173 become severely depleted overall once translational depletion is incorporated into our measure of depletion.

*p. 51*
171 Supra note 118. It is likely that in the European Union, this assumption might further be true of more languages than these five, which is why our assumption is conservative. Consider English first. The solid line in Figure 11 takes into account translational similarity for English. It shows the proportion over time of English word usage consisting of words that were either the subject of an active registration at the EUIPO or that were translationally similar to an active registration. By 2017, 90.8% of English word usage consisted of words already claimed as registered marks at EUIPO either directly, in that the registered term identically matched the English word, or indirectly, in that a translation into English of the registered term identically matched the English word.

*p. 52*
When translations into English are taken into account, the number of frequent words that remain unclaimed declines significantly. Only 30 of the 1,000 most frequently used words in English fail to match either a registered term or a translation into English of a registered term. Such words as WEAKEST, WORRIED, and WORRY are still available. But other English words trigger conflicts: for example, SORRY because of the registered marks SCUSI 174

## Vol. 112 TMR

*p. 53*
DIED because of the registered mark STARB 176 (meaning "died" in German).

*p. 53*
As to the other major European languages, Figure 11 shows that when translations are taken into account, each of them, even German, shows severe levels of depletion. Frequently used words that would appear to be unclaimed because they do not identically match an already-registered mark have nevertheless effectively been claimed because they match an already-registered mark as translated. For example, with respect to German, recall from Figure 6 that by 2017 46.7% of German word usage consisted of words that identically matched a term registered at the EUIPO. When translational similarity is taken into account, however, that statistic jumps to 80.1%. Viewed differently, of the 1,000 most frequently used words in German, 628 are orthographically different from any registered mark. Of these, many might make viable brand names, especially because, at least as a matter of orthographic uniqueness, the owner would be the only user of the word in the marketplace. Yet after translated similarity is considered, only 218 of these words remain unclaimed both orthographically and as translated. Among the 410 German words that are claimed only as translated are SPAß, which matches the registered mark FUN (English); 177 EINFACH, which matches the registered marks SIMPLE (English), 178 SIMPLICE (Italian), 179 and SIMPLESTA (Spanish); 180 GERNE ("with pleasure"), which matches the registered marks AVEC PLAISIRS (French), 181 CONGUSTO (Spanish), 182 and CONPIACERE (Italian); 183 and GEFÜHL, which matches FEELING (English). 184 Though each of these words, if adopted as a mark, would be orthographically unique, they would be conceptually equivalent to many other registered marks.

*p. 53*
Figure 12 Figure 13 shows the extraordinary impact of translational similarity in particular Nice classes for each of the five major languages. To aid in comparison, the black bars are the same as those shown in Figure 8 and indicate the proportion of word usage consisting of words identically matching a mark registered in the class. The gray bars indicate the proportion of word usage either identically matching a mark registered in the class or a translation of the mark into the indicated language. For example, for English, in Class 1 (chemical products), 21.3% of English-language word usage consists of words that identically match a registered mark in the class, yet translational depletion raises the level of depletion to 65.1% in the class. In Class 25 (apparel), 52.4% of English-language word usage consists of words identically matching a mark registered in the class, but translational depletion raises the level of depletion to 78.6% in the class.

*p. 56*
The other four languages show similar results. The results for German are especially striking. In Class 35 (general business administrations services), 25.8% of German-language word usage consists of words identically matching a registered mark, yet translational depletion raises the level of depletion to 72.1% in the class. In Class 25 (apparel), the level of depletion rises from 17.3% to 64.8%. Overall, the data make clear that translational similarity accounts for a large portion of the depletion of frequently used words in the four major languages other than English, especially German.

## D. The Depletion of Multilanguage Words

*p. 56*
We also studied the number of trademark registrations at the EUIPO that consist of or include what we call "multilanguage words." As discussed above, such words often take the form of cognates or borrowings, are orthographically closely similar across multiple languages, and convey roughly similar meanings in each of those languages (such as FANTASTIC and its close variations across the five major European languages). 185 The data indicate that a strikingly high proportion of multilanguage words have already been registered as single-word trademarks at the EUIPO. Figure 14 shows that, of the 1,000 most frequently used words in English, 38 qualify as universal words across the five major European languages. Of these, all but eight were registered as single-word marks. 106 of the 1,000 most frequently used words in English are mutually intelligible across at least four of the major European languages, and all but 22 of these are registered as single-word marks.

*p. 56*
185 Supra section I.A. We identified a frequently used English word as a multilanguage word if it was closely similar (that is, with a Jaro-Winkler score greater than or equal to 0.875) to a word in one or more of the four other major languages and that word's translation into English matched the English word. The data further show the depletion in particular of the subset of multilanguage words that convey positive affect. This should not be surprising. Just because a word like VIRUS is mutually intelligible across the five major European languages does not mean that it would make a good brand name. Instead, firms tend to seek universal words that also carry positive connotations. Of the 20,000 most frequently used English words that are mutually intelligible in the five major European languages, those that were the subject of live single-word registrations in 2018 yielded a mean sentiment score of 0.049, while those that remain unregistered yielded a mean sentiment score of -0.149. Similarly, of the 20,000 most frequently used English words that are mutually intelligible in at least four of the major European languages, those that were the subject of a single-word registration yielded a mean sentiment score of 0.035, while those that remain unregistered yielded a mean score of -0.137.

## E. The Shift to Coined Terms

*p. 57*
In our work on trademark depletion in the U.S. trademark system, we noted a steady increase over the past decades in the proportion of trademark applications for single-word word marks that are for coined terms. 186 We proposed that this trend was Vol. 112 TMR 957 consistent with the hypothesis that new applicants were progressively shifting away from dictionary words because such a high proportion of such words are already claimed. 187 We find similar trends in the EU trademark system. To estimate if an applied-for single-word mark is an invented word rather than an already-existing word from some language in the world, we took advantage of a feature of the Google Translate API. For any word that the API was unable to identify as a word in one of the 109 languages that it recognized, it would return that word unmodified or sometimes with slight modifications. 188 We identified an appliedfor word as invented if the API returned the identical word (or a closely similar word as measured by a Jaro-Winkler score equal to or exceeding 0.875) into at least four of the five major European languages, did not match an already-existing word in any of those languages, and did not match a surname appearing on the U.S. census list of the 151,671 most commonly occurring surnames in the United States. 189 As with our other measures, our goal was to minimize false positives and err toward underestimating the magnitude of any shift toward coined terms in the EU trademark system.

*p. 58*
Figure 15 shows the proportion of single-word word-mark applications submitted to the EUIPO from 1996 through 2018 that were for marks consisting of coined terms across all classes and in Class 25 (apparel goods) in particular. As in the U.S. trademark system, Class 25 shows an especially clear shift toward invented words. Recall that coined terms are generally less preferred as brand names for a variety of reasons discussed above. 190 We think that the trends shown in Figure 15 are symptomatic of a trademark system in which significant levels of trademark depletion are pushing new applicants toward invented words.

## F. Domain Name Depletion in the Major European Languages

*p. 59*
We noted above that companies launching new brands strongly prefer terms that they can register in the .com top-level domain. 191 Yet, as Table 3 reports, large proportions of the 20,000 most frequently used words in each of the five major European languages have already been registered in the .com top-level domain. Any company that wishes to adopt such a word as a brand name will likely be unable to register it in the .com domain. At best, it will need to acquire the domain name from a preexisting registrant or choose country-specific top-level domains for each country in which it operates. This may help to explain why so many trademark applicants in the EU trademark system are shifting to lesspreferred coined terms. potential brand names consisting of or closely similar either orthographically or translationally to a frequently used word in one or more of the five major European languages. To be sure, a business can conceivably seek to coin a new term dissimilar to any frequently used word in these languages and our data suggest that businesses are increasingly resorting to this expedient. But even here, they are finding that others have preempted them. A business can also abandon the search for a mark that is unique across the entire European marketplace and seek out one that is unique at least within a particular sector of goods or services. But the data show that class-specific depletion within the major Nice classes is just as severe. The stock of potential competitively effective trademarks in the EU marketplace is clearly under significant pressure. This is particularly true because the European Union is a global multilingual economy, which aggravates depletion of competitively effective marks via translational depletion to yield the reverse Babel problem.

## Vol. 112 TMR 959

*p. 60*
We now turn to how the EU trademark system seeks to relieve this pressure. The system has sprawled across the five major languages-and undoubtedly many of the other European languages as well-to such an extent that there is little unclaimed space left. Given this limit, the system appears to be maintaining its growth by allowing increased density. Having built out, it is now building up. The next Part addresses this phenomenon of trademark crowding.

## IV. TRADEMARK CROWDING AND OPPOSITIONS IN A GLOBAL MULTILINGUAL ECONOMY

*p. 61*
We emphasized above that even if an entity has already registered a particular mark in connection with particular goods or services, it is still possible for other, unrelated entities to register the same or a closely similar mark even in connection with the same or related goods or services. 192 The existence of an earlier registration will make any subsequent registration of a conflicting mark more difficult, but not impossible. The market entrant that becomes aware of a conflict with an earlier registration faces a choice between two alternatives: it may seek out a different mark that no one has yet claimed (thus worsening depletion) or it may "'have a go '" 193 at applying to register a mark that conflicts with an already-registered mark and hope that the application registers. The EUIPO data suggest that entrants are increasingly turning to this second alternative and are increasingly succeeding in registering conflicting marks. The result is "trademark crowding," in which numerous unrelated entities own registrations of closely similar or even identical marks for closely related products.

*p. 61*
In this Part, we focus on the worsening problem of trademark crowding in the EU trademark system. While our discussion of trademark depletion above occasionally benchmarked EU results against those in the United States, here we bring the American comparison to the fore. We do so to convey just how bad conditions have become in the EU trademark system. In 2018, the 1.3 million trademark registrations on the EUIPO register were only about half as many as the 2.4 million registrations on the USTPO's Principal Register, yet as we show, rates of trademark crowding on the EUIPO register have been increasing rapidly over the past two decades and are now far higher than those at the USPTO. We argue that one leading cause of this difference is that, unlike the USPTO, the EUIPO does not engage in ex officio examination of applied-for marks for confusing similarity with already-registered marks. Like many other trademark systems around the world, the EU trademark system relies only on a third-party opposition process to filter out confusingly similar registrations. We present data showing that, on its own, this opposition process is simply not adequate to prevent trademark crowding. The result is that the EUIPO risks becoming little more than a rubber-stamping agency for ever more crowded fields of marks. As we suggest here and develop more fully in Part V, this rise in trademark crowding represents, we think, one potential, but preventable future for the global trademark system.

## A. Trademark Crowding

*p. 62*
The EUIPO registry is rife with crowded fields, so much so that in many classes of goods and services, market entrants face a very real challenge in finding any areas that are not crowded with already registered marks. For a simple example of a crowded field, consider the well-established French fashion brand SANDRO,194 which is highly successful in the marketplace and highly sophisticated. Yet in 2017 in Class 25 (apparel goods), other entities owned live registrations for the marks SANDRA, 195 SAND,196 SANJO,197 SANO,198 SANRIO,199 and SKANDO,200 to name just a few of the approximately 35 closely similar marks registered in the class by unaffiliated entities. For another, extreme example, consider the trademark LOVE. In 2017 just in Class 25, at least 10 unaffiliated entities owned active registrations in the word itself, while at least an additional 75 unaffiliated entities owned active registrations in close variations on the word, such as LOVER, 201 LOVME, 202 LOVEDO, 203 LOVERS, 204 LOVLEE, 205 LOVECHILD, 206 and LOVERBIRD. 207 And when translational similarity is taken into account, 208 the crowded field around the trademark LOVE expands to 95 different trademark owners, including for the marks AMORE 209 (Italian for "love"), AMOUREUSE 210 (French for "in love"), and LIEBESKLEID 211 (German for "love dress"). See Examination Guidelines, supra note 91, at § 3.4.3.1 (citing "love" as one example of "very basic words, which will be understood in all Member States because they have become internationally used").

## Vol. 112 TMR

*p. 63*
Such crowded fields of marks impair the workings of a trademark system in several ways touched upon above. 212 First, closely similar marks may cause consumer confusion as to source. Some proportion of relevant consumers may believe that unrelated marks originate from the same commercial entity. For example, even consumers who notice the difference between SANDRO and SANDRA may assume that the latter is affiliated with the former, perhaps as a spin-off brand. Second, closely similar marks may increase consumer search costs. Even consumers who are not confused as to source will need to attend more carefully to the slight differences among the trademarks they confront. 213 Thus, consumers may be well aware that SANDRO and SANDRA originate from unrelated entities. Still, they must be careful not to mistakenly choose the one rather than the other. Third, in a process akin to a tragedy of the commons, crowding reduces the distinctiveness of any trademark in a crowded field from other marks in that field. All marks in the field suffer the resulting loss in brand differentiation and "selling power." 214 As a brand name, variations on LOVE and their equivalents in the other major European languages are clearly hackneyed, overused signifiers.

*p. 63*
The EUIPO data indicate that trademark crowding has been increasing at the EUIPO at a far faster rate than at the USPTO. The result is that after only two decades since the establishment of the EUIPO register, levels of crowding at the EUIPO already exceed those at the USPTO. For each of the leading Nice classes at the EUIPO and USPTO, Figure 16 estimates a mean crowding score by year (the solid dots) and, for the EUIPO, a mean translationalcrowding score by year (the hollow dots). We developed these scores as follows: For any individual registered mark in a class, that mark's crowding score is the number of other registered marks in the class that are orthographically closely similar to the mark and are owned by unaffiliated entities. A class's mean crowding score is the estimated average of the crowding scores of all marks in the class, based on a random sample. 215 An individual mark's translational- Frank I. Schechter, The Rational Basis for Trademark Protection, 40 Harv. L. Rev. 813, 819 (1927).

## 215

*p. 63*
To develop these data, for each year from 1998 through 2018, we randomly sampled 500 actively registered single-word trademarks from each leading class and calculated Jaro-Winkler scores between each of the 500 sampled marks and all other single-word marks registered in the class in the particular year. We then counted, for each of the 500 sampled registrations, the number of trademark registrations closely similar to the sampled mark (JW ≥ 0.875) but owned by entities different from the owner of the registration of the sampled mark. Figure 23 shows, by year and Nice class, the mean number of registrations closely similar to each sampled mark, as well as a second-order polynomial trendline of that mean over time.

*p. 64*
crowding score is the number of unaffiliated marks in the mark's class that are orthographically closely similar to the original mark or, if the mark is intelligible in one of the five major EU languages, to the mark as translated into any of the five major European languages. A class's mean translational-crowding score is the estimated average of the translational-crowding scores of all marks in the class, again based on a random sample. 216 The results across the six classes detailed in Figure 16 all show similar trends over time. Crowding levels at the USPTO are flat or suggest only relatively modest increases. By contrast, crowding at the EUIPO has been increasing rapidly, especially when translational similarity is taken into account. For example, in Class 5 (pharmaceutical goods), crowding at the USPTO increased over the 21-year period from a mean crowding score of 3.93 closely similar marks in 1998 to 7.46 in 2018. In comparison, the EUIPO data report more significant increases, with a rise in the mean crowding score from 0.58 in 1998 to 10.71 in 2018 and in the mean translational crowding score from 0.64 to 13.45. These trends in Class 5 at the EUIPO are particularly revealing because we would expect to see levels of crowding in the class that are low and relatively stable over time, as they are at the USPTO. For pharmaceutical names, regulatory agencies employ their own highly restrictive tests of similarity. 217 But even in this class, the EUIPO system appears unable to control crowding.

*p. 64*
We used a random sample because this is a computationally intensive process. Our sample is sufficiently large to yield good estimates of the underlying quantities of interest. For example, the largest subpopulation of EUIPO registrations from which we sampled consisted of 349,934 active registrations in Class 9 in 2018. A random sample of 500 of such registrations yields a confidence interval under 5% at a 95% confidence level.

## 216

*p. 64*
We used an analogous approach to calculate this score as described above in note 215, taking a different random sample from that used for our estimates of non-translational mean crowding scores. Other classes show even higher levels of crowding at the EUIPO. In Class 16 (published goods) and Class 35 (business services), crowding scores at the USPTO have remained flat over time. At the EUIPO, they have been increasing dramatically. Admittedly, like Class 9 (electronics goods), Class 35 covers such a broad range of goods or services that identical marks used by unaffiliated firms can coexist in the classes without necessarily confusing consumers as to source. But even if closely similar marks are not confusing as to source, their similarity nevertheless increases consumer search costs and reduces the distinctiveness of each mark from other marks. The latter is especially a concern in a class like Class 25 (apparel goods), where the perceived uniqueness of a mark may greatly contribute to the marketability of the goods to which it is affixed. Indeed, the mark's uniqueness as against other marks may often be the only "unique selling proposition" 218 that the goods' producer has to offer.

## B. Trademark Oppositions

*p. 66*
In theory, the EUIPO's third-party opposition process should on its own be sufficient to prevent trademark crowding. Incumbent registrants are well placed to evaluate applied-for marks for confusing similarity with their own registered marks and have strong incentives to do so. Our data show, however, that in practice the EUIPO opposition process is not widely used and opposition rates have been declining even as crowding has been increasing. This is in stark contrast to conditions at the USPTO. There, ex officio refusal rates during trademark examination have been steadily increasing while third-party opposition rates have remained flat. Combined, these trends in the U.S. data indicate that ever higher proportions of trademark applications at the USPTO are failing on the basis of confusing similarity with already-registered marks. This helps to explain why crowding rates at the USPTO have remained relatively stable.

## The Limited Population of Users of the Opposition Process

*p. 66*
The EUIPO opposition process is dominated by a relatively small set of very frequent opposers from an even smaller set of countries. 219 Of the 460,441 corporate entities that filed a trademark application at the EUIPO from 1996 through 2017, only 11.6% ever filed an opposition during that period. Just 1% of those 460,441 corporate applicants accounted for 58.4% of all corporate oppositions, and 5% accounted for 73.4% of all corporate oppositions. A list of the five most frequent corporate opposers reveals the usual suspects, at least to those familiar with the European market: The EUIPO dataset anonymizes the identity of non-corporate applicants filing under their personal names, so it is not possible to study the opposition practices of such applicants. The data suggest that such applicants account for 25.1% (or 76,089) of the total of 303,607 oppositions filed against applications filed from 1996 through 2017.

*p. 67*
Vol. 112 TMR oppositions process. German entities filed 28.5% of all corporate oppositions but were responsible for only 15.4% of all trademark applications. Spanish entities filed 14.7% of all oppositions and accounted for only 7.4% of all applications. Finally, unlike German and Spanish entities, American businesses were responsible for a greater proportion of all applications, at 15.6%, than all oppositions, at 11.5%.

## Declining Opposition Rates

*p. 67*
That so few entities actually engage in the opposition process may explain why the data show a steady decline over time in the EUIPO opposition rate (the annual proportion of EUIPO trademark applications that are opposed). As Figure 17 shows, the annual number of EU trademark applications that are opposed has been increasing (the bars and right y-axis). But this increase has not kept up proportionally with the even greater increase in the number of applications filed each year. The result is that the annual opposition rate has been steadily declining (the line and left y-axis), so that by 2017, only 11.1% of applications were opposed. Indeed, across all the major Nice classes, and filtering for application characteristics such as country of origin or language of application, opposition rates have been declining. The declining opposition rate at the EUIPO is especially surprising when compared to trends at the USPTO. As Figure 18 indicates, for the period 2003 through 2017, there has been a clear increase over time in the proportion of applications receiving a refusal from a trademark examiner based on confusing similarity with an already-registered mark (the solid black line), from 10.6% of all applications filed in 2003 to 16.0% of those filed in 2017. Opposition rates at the USPTO over the same period have been steady (the solid gray line). It appears that the USPTO's in-house refusal process serves as the primary device to filter out confusingly similar registrations. These USPTO data paint a picture of a registration system that is experiencing severe trademark depletion and in which applicants are responding to depletion in part by seeking to register ever higher proportions of closely similar marks. But the registering agency is refusing to relieve the pressure depletion is placing on the supply of viable trademarks by allowing trademark crowding.

## The Role of the Cooling-Off Period

*p. 68*
As explained above, the filing of an opposition at the EUIPO initiates a cooling-off period of at least two months in which the applicant and opposer are given the opportunity to settle the Vol. 112 TMR opposition. 220 The cooling-off mechanism plays an important role in the opposition process. The data show that a large proportion of oppositions are resolved before leading to an Oppositions Division decision. For the 1,570,264 applications filed from 1996 through 2016, 216,195 (or 13.8%) were the target of at least one opposition on the basis that the applied-for mark was identical or confusingly similar to an already-registered mark. Of these opposed applications, 64.3% (or 139,054) were never the subject of an Oppositions Division decision, and of these, 72.6% (or 101,011) registered. The EUIPO can justifiably claim that the cooling-off period is a successful mechanism for allowing opposing parties to settle their disputes without recourse to an administrative tribunal. 221 Oppositions that never reach an Oppositions Division decision are typically settled with the applicant's agreement to narrow the applied-for goods or services or a coexistence agreement, such as to operate in distinct regions. 222 The parties otherwise agree to coexist in their use of (often closely) similar marks on goods or services that are also likely to be closely related. Courts have generally upheld such agreements, 223 and the EUIPO allows an applicant's mark to register alongside those marks that served as the basis for the opposition when there is such a governing coexistence agreement. The Open Dataset does not provide the data that would allow a quantitative study of these settlement practices, as coexistence agreements are generally kept private. 224 But qualitative accounts of the cooling-off period indicate that these kinds of settlement agreements are the most common outcome. 225 It is also an outcome that greatly contributes to trademark crowding. While the cooling-off period facilitates settlement, its capacity to do so is diminishing. The proportion of oppositions that are resolved before resulting in an Oppositions Division decision has been declining. Figure 19 focuses on applications that received an opposition on the basis that the applied-for mark was identical or confusingly similar to an already-registered mark. It reports the proportion by filing year of such applications whose oppositions resolved before resulting in an Oppositions Division decision. The downward trend is slight but unmistakable. It is consistent with a condition in which parties to oppositions are finding it more difficult to find space in which to coexist.

## Opposer Win Rates

*p. 70*
Our primary goal has been to present evidence of the high levels of trademark crowding at the EUIPO and the apparent inefficacy of the third-party opposition process to control crowding. But we note two additional trends in the opposition data, specifically relating to opposer win rates when no settlement is reached, that may further reflect on both of these phenomena in the EU trademark system.

*p. 70*
The first is the increase over time in the win rate of parties that do bother to file an opposition and pursue that opposition through to a decision by the EUIPO Oppositions Division. Figure 20 focuses on applications that were the subject of an Oppositions Division Vol. 112 TMR decision on the basis that the applied-for mark was identical or confusingly similar to an already-registered mark. The figure reports by filing year the proportion of such applications that were refused registration either in whole or part. 226 We think these data represent further evidence of the severity of crowding at the EUIPO. At least among those who pursue oppositions through to decisions, their increasing success reflects the increasing strength of those oppositions that are filed. A second trend relates to applications that received multiple oppositions from multiple different opposers. Figure 21 reports the registration rate of applications by the number of different entities who filed one or more oppositions to the applications on the basis that the applied-for mark was identical or confusingly similar to their already-registered mark. As expected, unopposed applications enjoy a very high registration rate of 0.922. Also as expected, the 226 There was no substantial difference in the trend lines when the data are broken out into two trend lines, one for applications that were opposed on the basis of "double identity" under article 8(1)(a) of the Trade Mark Regulation and another for applications that were opposed on the basis of confusing similarity under article 8(1)(b) of the Trade Mark Regulation.

## 227

*p. 71*
We know of no factors that might exert a selection effect on oppositions that are pursued to decision that would explain the significant rise in opposer win rates. registration rate then declines with the number of opposers who arrayed themselves against any particular application. But interestingly, at seven opposers, the registration rate bottoms out at 0.527 and then begins to increase. At this inflection point, the crowded field is apparently occupied by so many different registrants of closely similar marks that the scope of each registrant's property right is critically narrowed and incumbents' ability to prevent entrance into the crowded field declines. At the extremes, the data indicate that the 2008 application for the mark PLUS ONE was opposed by 20 different entities, 228 the 2012 application for the mark STUDIOLINE was opposed by 32 entities, 229 the 2005 application for the mark BONOLOTO was opposed by 33 entities, 230 and at the outermost extreme, the 2010 application for the mark Ö was opposed by 103 different entities. 231 All of these applications overcame their many opposers and succeeded to registration. These crowded fields represent the final breakdown of signification in the most overpopulated areas of a trademark system, where individual tokens cease to exist and every mark is merely a type.

## Vol. 112 TMR

*p. 73*
The EUIPO's third-party opposition process has proven incapable of controlling crowding. If businesses even become aware of an application for a closely similar mark and then go so far as to file an opposition to it, they often settle with the applicant in a manner that allows the coexistence of similar marks. Already, pockets of the EU trademark system have collapsed into supercrowded fields of essentially indistinguishable and indistinctive marks. The lack of oversight provided by examiner review for confusing similarity is taking its toll. As their settlement behavior suggests, many firms may not see a problem with trademark crowding as a means of coping with trademark depletion. But its effects on the integrity of the trademark system, competition, and consumers is another matter. With these effects in mind, we turn now to the legal and policy implications of our findings for the global multilingual economy.

## V. LEGAL AND POLICY IMPLICATIONS

*p. 73*
In this Part, we focus on what the EU example may teach the rest of the world, including the United States. The costs of trademark depletion and crowding on competitors and consumers are especially severe in a multinational multilingual trademark system like the EU system. Our primary point is simply to urge recognition of the fact that trademark systems have ecological limits, and globally integrating multilingual trademark systemswhether de jure or de facto-have especially pronounced limits. At these limits, the costs of granting new trademark rights may outweigh the benefits in ways not previously appreciated when the supply of new trademarks seemed inexhaustible. We then turn to potential reforms that may minimize these costs. To reduce trademark depletion and crowding, we advocate that trademark law significantly curtail the application of translational similarity as a basis for confusing similarity. We also evaluate the clear benefits of ex officio review for confusing similarity, but we recognize that if a trademark system will not impose this form of review, it should at least improve the information it makes available to existing registrants of potentially conflicting marks who may wish to oppose the registration of an applied-for mark. Additionally, we recommend reformed fee structures calibrated to the costs that the claiming of certain terms, such as multilanguage words, imposes on others. Finally, we advocate, among other measures, stronger enforcement in all trademark systems of a requirement that a trademark actually be used in commerce to qualify for and retain protection.

*p. 73*
We then respond to a potential counterargument that depletion and crowding may yield net benefits.

## A. The Costs of Trademark Rights in a Global Multilingual Economy

*p. 74*
Trademark thinking has long been sensitive to the fact that granting exclusive rights in a trademark to one company can sometimes impose costs on the company's competitors. Indeed, in the very earliest cases in English trademark law, courts expressed great hostility toward the monopoly rights they were being asked to enforce. 232 Trademark law has traditionally operated according to the principle that it will grant exclusive rights in a trademark only if competitors still have access to "a latitude of competitive alternatives," 233 to adequate alternative means of describing and designating the source of their products. This is one reason why trademark law refuses to protect generic terms 234 and functional product features 235 and has special rules for the protection of descriptive terms. 236 This is also why trademark law purports to grant broader rights to marks that qualify as arbitrary (in that they have no semantic connection to their product, such as HORIZON for banking services) or fanciful (in that they are coined terms, such as TONO-BUNGAY for a beverage). 237 The idea is that competitors have no need to use such marks to compete effectively, so exclusive rights in them impose no costs on others. 238 Though trademark thinking thus recognizes the costs of trademark rights, it has traditionally conceived of these costs only in isolation, on a case-by-case basis, not in the aggregate, not as these externalities may accumulate over time. 239 This is understandable. Until recently, the law had no need to worry about such concerns-just as until recently there may have been no need 232 E.g., Blanchard v. Hill, 26 Eng. Rep. 692 (Ch. 1742).

*p. 74*
233 Taco Cabana Int'l, Inc. v. Two Pesos, Inc., 932 F.2d 1113, 1119 (5th Cir. 1991), aff'd sub nom. Two Pesos, Inc. v. Taco Cabana, Inc., 505 U.S. 763 (1992).

*p. 74*
234 Abercrombie & Fitch Co. v. Hunting World, Inc., 537 F.2d 4, 9-11 (2d Cir. 1976); Fromer, supra note 27; Jake Linford, A Linguistic Justification for Protecting "Generic" Trademarks, 17 Yale J.L. & Tech. 110, 133-40 (2015); Lisa P. Ramsey, Descriptive Trademarks and the First Amendment, 70 Tenn. L. Rev. 1095, 1121-24 (2003).

## 235

*p. 74*
TrafFix Devices, Inc. v. Mktg. Displays, Inc., 532 U.S. 23, 32 (2001) Ltd. v. Nawab, 335 F.3d 141, 147-48 (2d Cir. 2003) (Leval, J.). For a critique that these rules are nonetheless not capacious enough to advance the fair competition that trademark law promises, see Fromer, supra note 27.

## Vol. 112 TMR

*p. 75*
to worry about running out of phone numbers 240 or Internet protocol addresses. 241 But as we have sought to show, the cumulative systemic costs of new trademark rights can reach a point where they begin to impair competition within particular economic sectors and even across the economy because competitors no longer have access to competitive alternatives. At this point, market entrants face a dilemma: they must somehow navigate between the Scylla of further depletion and the Charybdis of further crowding. They may either choose a not-yet-claimed mark, which often retains that status because it is inferior to already-claimed marks, or they may pursue a mark that is closely similar or identical to an alreadyclaimed mark. 242 The entrant must choose not the better option, but the less bad between the two. 243 To be sure, there may appear to be a third option, which is to claim a mark that some other entity is no longer using and has left fallow. But abandonment rates in the EU trademark system, like those in the U.S. system, do not come close to meeting the demand for new marks. 244 The systemic costs of trademark rights increase with the size of the trademark system. As our data show, these costs accelerate in a trademark system consisting of once-independent markets that are increasingly integrating, especially when those markets bring with them new major languages and multilingual consumers. Through such consumers, translational similarity exerts a multiplier effect on both trademark depletion and crowding. A claim over a single word will result in claims over the set of words that any significant population of consumers will perceive as orthographically, phonetically, or translationally similar. 245 Because trademark depletion and crowding have both reached such severe levels, we think a multinational trademark system like the European Union's is well beyond trading off between proximity costs (akin to allowing further crowding) and distance costs (akin to allowing further depletion), something Daniel Hemel and Lisa Larrimore Ouellette posit is intrinsic to any trademark system. Hemel & Ouellette, supra note 38. That is, because the proximity costs and distance costs are both so high, they must both be reduced to improve the trademark system's functioning, as discussed in this Part. 244 For a simple example, there were 1,160,540 active trademark registrations at the end of 2016 at the EUIPO. Of these, only 41,159 expired, were cancelled, or were otherwise removed from the register through the course of 2017. Meanwhile, there were 142,150 new applications in 2017, of which 127,997 registered by the end of 2018.

*p. 76*
Vol. 112 TMR 975 but, in effect, transnationally. 246 A global firm may be willing to bypass a spoiler in a single small country but not when that spoiler denies access to a significant market. All of these factors result in a rapid increase in the numerator of already-claimed marks in a trademark system. As for the denominator of potential competitively effective marks, when a market increases in size and diversity, the set of marks that will be effective throughout that market decreases in size. 247 A viable mark must fall within the intersection of the subsets of marks that are viable in the various regions, languages, and cultures of a global market. And all the while, as a trademark system grows in size, more and more companies are vying to claim rights over the same shrinking reservoir of marks. 248 Taking into account all of these processes, it is not difficult to understand how the EU trademark system and, more broadly, the de facto global trademark system could reach a stage where there are no longer a sufficient number of competitive alternatives available to market entrants.

*p. 76*
The costs of trademark depletion and crowding are particularly pronounced in their effect on competition. Trademark depletion impairs competition for at least three reasons. First, market entrants face greater difficulties than incumbents did in searching for an unclaimed competitively effective mark or at least a mark that is not yet overcrowded with multiple parallel claimants. 249 Second, market entrants are often compelled to settle for less effective marks, and studies have shown that such marks hinder the long-term performance of firms saddled with them. 250 Third, even as later entrants struggle to find a competitively effective mark, incumbents are typically able to leverage their preexisting registrations for proven marks across Nice classes into new registrations in classes of goods or services that they have not yet claimed. 251 Trademark crowding similarly impairs competition, but affects both market entrants and incumbents. As new marks crowd around a preexisting mark, all marks in the crowded field suffer from a loss of distinctiveness. The likelihood of consumer confusion as to the sources of the products bearing these marks increases. Investor Recognition, and Firm Value, 109 J. Fin. Econ. 813 (2013).

*p. 77*
Vol. 112 TMR marks declines, and with it the ability of the mark to stand out in the marketplace. This effect can be especially damaging to small businesses, which may lack the resources to compensate for their mark's loss of distinctiveness through greater advertising. Finally, at the extreme, crowded fields of marks can devolve into noise, reducing all trademarks in them to ineffective and indistinguishable empty signs, signifying nothing.

*p. 77*
Equally as important are the costs that depletion and crowding impose on consumers. 252 Depletion damages consumer welfare by pushing entrants toward marks that increase consumer search costs. As depletion eats away at the stock of unclaimed marks, businesses may be forced to turn to marks that are less consumerfriendly, in that they are less memorable and generally less effective in serving as shorthand for the characteristics of the goods or services to which they are affixed. 253 Meanwhile, crowding requires consumers to devote more care to distinguishing closely similar marks. 254 As increasing proportions of consumers engage in transnational consumption, 255 depletion hurts such consumers in additional ways. When a firm is blocked from using the same mark in multiple national markets because a competitor has already claimed the mark in one of those markets, consumers may be confused as to the true source of all the products involved. We see this readily from the example of U.S.-based retailer TJ MAXX adopting the mark TK MAXX in England to avoid a similar mark for clothing stores, 256 leading consumers exposed to both marks to ponder whether the two are the same 257 and the media to emphasize that they are the same even if consumers are likely to be confused. 258 Finally, we suggest an additional effect of trademark depletion and crowding in the global context. The firms of the Global North dominate transnational trademark registrations, so much so that developing-world trademark offices find themselves largely serving foreign clients. 260 It may prove to be a bitter irony that the Anglosphere in particular has imposed the English language on much of the world and is now succeeding, through American companies especially, in asserting exclusive rights in much of that language throughout the world, at least for purposes of distinguishing products and designating their source. The degree to which depletion and crowding benefit rich-world market incumbents in the overall global marketplace to the detriment of developing-world market entrants deserves further study. This process may already be occurring on a smaller scale within the European Union, where firms of the EU core countries may be preempting the rights of periphery-country firms. 261 Given the vital role that trademarks will continue to play in facilitating competition in a global marketplace, 262 we now turn to possible ways to minimize these costs.

## B. Ameliorating Trademark Depletion and Crowding

*p. 78*
Efforts to ameliorate trademark depletion and crowding in any trademark system must come to terms with difficult questions of design. There are, for example, basic questions of tailoring: Should reforms set out uniform, one-size-fits-all rules that apply to all applications and registrations or should they be more narrowly In recent years, owing principally to online product reviews, some have wondered whether trademarks are less necessary to promoting competition and consumer welfare. After all, one can peruse product reviews to learn whether a good is worthy of purchase instead of relying on the goodwill associated with a mark. E.g., Itamar Simonson & Emanuel Rosen, Absolute Value: What Really Influences Customers in the Age of (Nearly) Perfect Information (2014); Hemel & Ouellette, supra note 38; Lisa Larrimore Ouellette, Does Running Out of (Some) Trademarks Matter?, 131 Harv. L. Rev. F. 116, 122-23 (2018). While we think that genuine product reviews are generally a positive contribution, we also believe that trademarks are here to stay, both because they still continue to designate source instead of or in addition to online product information and because they help build a business's identity. E.g., Brad VanAuken, The End of Brands?, Branding Strategy Insider, Feb. 25, 2014, https://www.brandingstrategyinsider.com/the-end-of-brands.

*p. 80*
trademark system without unduly burdening market entrants or incumbent registrants.

## The Reverse Babel Problem

*p. 80*
As the world's markets continue to integrate, the EU trademark system teaches that it may be time for the world's trademark systems to abandon or at least significantly moderate application of the doctrine of translational similarity. Recall that this doctrine holds that two or more marks may be confusingly similar if they convey the same meaning to any significant population of consumers capable of understanding the terms in both languages. 268 We showed how this has created a "reverse Babel problem" in the EU trademark system, in which, in effect, nearly everyone may be presumed to understand everyone else regardless of which language they are using. 269 The result is that any claimed mark in a major language may deplete or crowd translationally equivalent marks in a wide variety of major and minor languages spoken in a multilingual society.

*p. 80*
The doctrine of translational similarity and the reverse Babel problem are also found in the U.S. trademark system, though their effects are less severe. U.S. law employs what it calls the "doctrine of foreign equivalents." 270 As the Trademark Trial and Appeal Board has observed, "the doctrine … extends the protection of the [Lanham] Act to those consumers in this country who speak other languages in addition to English … [when] [a]t least one significant group of ordinary American purchasers is the purchaser who is knowledgeable in English as well as the pertinent foreign language." 271 The result, unsurprisingly, is that U.S. law often treats Spanish-language terms as confusingly similar to their translational equivalents in English, 272 as 41 million people in the United States-13.5% of the population-speak Spanish at home 273 268 Supra text accompanying note 92. grounds, 563 F.3d 1347 (Fed. Cir. 2009). It is applied only when the "ordinary American purchaser" would "stop and translate" the foreign wording in a mark into its English equivalent. Palm Bay Imps., Inc. v. Veuve Clicquot Ponsardin Maison Fondee en 1772, 396 F.3d 1369, 1377 (Fed. Cir. 2005).

## Vol. 112 TMR

*p. 81*
and it is the most studied foreign language in U.S. schools. 274 But the law will sometimes do the same for terms in French-when spoken very well or well by 0.6% of the American population 275and Russian-when spoken by 706,000 people living in the United States 276 -and many other languages. 277 To investigate the effect of Spanish on levels of depletion and crowding in the U.S. trademark system, we translated all applied-for and registered marks at the USPTO from 1998 through 2018 into Spanish. Figure 22 presents representative results from Class 9 (electronics goods) and Class 25 (apparel goods). Though we find that the resulting levels of depletion and crowding in the U.S. system do not reach the levels we reported above for the European Union in Figure 16, the data indicate that even in less polyglot markets translational similarity can exacerbate the problem of trademark crowding. To mitigate the reverse Babel problem in trademark systems going forward, we think it will be necessary to eradicate, or at the very least moderate, the reach of translational-similarity doctrine.

## Vol. 112 TMR 981

*p. 82*
It would be preferable to remove the doctrine from trademark law in recognition of the depletive and crowding effects it has in a global multilingual trademark system already experiencing severe levels of depletion and crowding without it. Even if that it is not possible, at the very least, the doctrine should require a higher showing; not simply that consumers are capable of understanding the common meaning of two marks from different languages, but that this understanding will lead a significant proportion of consumers to believe that the two marks originate from the same source. Admittedly, greater tolerance of translationally similar marks may produce some degree of consumer confusion in the short run. But here trademark law may take advantage of its inherent circularity to train consumers over time to expect that translationally equivalent marks do not necessarily originate from the same source. 278 As consumers become aware of more such marks' coexistence, their expectations in the market may change and they may be less likely to be confused. Indeed, their expectations will need to change as conditions of depletion and crowding intensify.

## The Benefits and Costs of Examiner Review for Confusing Similarity

*p. 82*
As discussed above, trademark registering agencies around the world employ one of two different methods to determine if an applied-for mark is confusingly similar to an already-registered mark. 279 Some agencies provide two stages of review, consisting of review by a trademark examiner followed by a third-party opposition process, while others rely on only one stage of review, consisting solely of a third-party opposition process. The EUIPO is the leading expositor of the opposition-only approach to trademark registration, while the USPTO is the leading practitioner of examiner review. The example of the EU trademark system teaches that examiner review is by far superior to an opposition-only process for purposes of reducing levels of trademark crowding, protecting consumers from confusion as to source, and maintaining the integrity of a trademark system.

*p. 82*
When the EU trademark system was first formed, its designers openly embraced a neoliberal, laissez-faire framework for filtering trademark applications for confusing similarity. As the European Commission explained: 278 Barton Beebe, Search and Persuasion in Trademark Law, 103 Mich. L. Rev. 2021, 2066 (2005) ("The scope of trademark protection is based largely on the law's assessment of the degree of actual search sophistication among consumers in the marketplace, yet the degree of search sophistication consumers bring to the marketplace depends largely on the scope of trademark protection they expect to find there."). bargain are consumers, who must then navigate a more crowded field of marks and overcome any confusion as to source that may occur. 299 Also left out are later entrants that might see reflected on the register a more crowded field than truly exists and consequently and unnecessarily decline to seek marks they prefer. 300 These secret settlements thereby undermine the integrity of the EUIPO register. It is precisely out of these concerns for secret settlements that the USPTO, in contrast to the EU approach, may refuse to register any mark that its examiner review deems confusingly similar with an already-registered mark even if the parties themselves believe that there is no likelihood of confusion and have reached a coexistence agreement. 301 We think that as trademark systems have become increasingly global, it is imperative that trademark offices actively work to decrease crowding and the damage it does to competition and consumer welfare. They can accomplish this, as the USPTO does, by implementing likelihood-of-confusion review by trademark examiners. 302 But to the extent a trademark office insists on adhering to the European view-that market participants are wellplaced to assess confusing similarity whereas examiners are not 303 -trademark systems ought to be structured to do a more effective job than the EUIPO in controlling crowding. For one thing, these offices must provide all registrants-especially small businesses unlikely to be monitoring their trademarks independently-with adequate means of learning about applications for potentially conflicting marks. Specifically, the EUIPO must improve its search reports to identify all potential conflicts and should err toward false positives rather than false negatives, thus empowering prior registrants to decide for themselves whether they will initiate an opposition. In an age of Google, the technology exists to perform this task. In particular, third-party providers of trademark monitoring services have existing firms generally have incentives not to license their marks to producers of competing goods or services. Jeanne C. Fromer, The Unregulated Certification Mark(et), 69 Stan. L. Rev. 121, 129-30 (2017).

*p. 86*
In the case of the EUIPO, the Office could devote part of its enormous surplus to cover the costs of such a review system. On the EUIPO's budget surplus, see European Court of Auditors, EU Intellectual Property Office Should Use Surplus Money Productively, May 15, 2019, https://www.eca.europa.eu/en/Pages/NewsItem.aspx?nid=12100.

*p. 87*
Vol. 112 TMR demonstrated considerable sophistication in automating the production of such reports. 304 Alternatively, offices like the EUIPO should consider adopting what might be termed an "examiner search" approach, in which human examiners combine with automated search to produce reports to prior registrants of potentially conflicting applications. Unlike USPTO trademark examiners, such examiners need not rule on whether an application conflicts with a prior registration. But like USPTO examiners, they can develop expertise in evaluating potentially confusing marks and prepare useful search reports on that basis.

*p. 87*
For another thing, in light of the prevalence of secret settlements whose terms are not made public, it may be useful to encourage or even require the contracting parties to publish those terms in the EUIPO register, especially if those terms limit the goods or services on which the parties will use the marks at issue or where in the European Union they will use them. Such disclosures would ensure that newer entrants could ascertain with greater accuracy the true state of the EU trademark register. Finally, if the EUIPO were to initiate examiner review for confusing similarity, then the EUIPO should also seriously consider following the example of the USPTO by requiring that its examiners review the terms of any coexistence agreement to ensure that it is not inordinately contributing to greater crowding and consumer confusion. 305 If an agreement fails this review, the later-filed mark should not merit registration.

## Reformed Fee Structures

*p. 87*
Most trademark offices around the world, including the EUIPO and the USPTO, impose uniform, one-size-fits-all fee structures on their applicants and registrants regardless of the nature of the marks they are applying for or have registered. 306 Offices also typically do not reduce fees for small-and medium-sized enterprises. 307 Such a crude fee structure, combined with a firstcome-first-served approach to the registration of marks, may have made sense in the earliest years of modern trademark systems. But the problems of trademark depletion and crowding, particularly in the transnational and multilingual context, call for a more refined approach.

*p. 87*
Trademark offices should adjust their fees to compel applicants and registrants to internalize some part of the costs they are imposing on others by registering frequently used words rather than 304 E.g., Markify, Best-in-Class Full Trademark Search and Watch, https://www.markify. com (last visited June 17, 2022) Vol. 112 TMR 987 coined terms. 308 This is not a radical idea in trademark law. U.S. trademark doctrine, for example, already affords a broader scope of protection to coined terms to encourage their adoption and because providing exclusive rights in them is understood to impose lower costs on others. 309 Consistent with this approach, the EUIPO could discount fees for word marks that qualify as coined terms in that they are not identical to or mere misspellings of words (or proper nouns) in any EU official language. More significantly, registering agencies should impose higher fees on multilanguage words to reflect the substantially higher value of these words as brand names in a global multilingual economy. Such a fee structure would of course require judgment calls concerning whether a word mark qualifies as a neologism, but registering agencies make such judgment calls all the time, particularly when they are reviewing applications for absolute grounds of refusal such as whether an applied-for mark is generic or merely descriptive.

*p. 88*
As we have suggested in previous work, trademark offices could also impose congestion pricing schemes on registrations in especially depleted or crowded classes of goods and services. 310 Offices may find such schemes to be less challenging to administer than those based on individualized analyses of the lexicographic or etymological characteristics of applied-for marks. Quantitative assessments of depletion and crowding in particular classes could form the basis for the differential fee schedule. The regressive effects of congestion pricing could be mitigated with discounts offered to small-and medium-sized enterprises. 311

## Enforcing the Use Requirement

*p. 88*
As explained above, the EUIPO will register a mark even if the registrant is not using the mark in commerce. 312 The EU trademark registrant enjoys a five-year grace period from the registration date to make use of the mark, and even after that grace period has expired, the registration will remain in effect unless a third party challenges it for non-use-and third parties rarely initiate such 308 Beebe & Fromer, supra note 8, at 1030-33 (discussing tiered fees generally). 309 Virgin Enters. Ltd. v. Nawab, 335 F.3d 141, 147-48 (2d Cir. 2003) (Leval, J.). Admittedly, Judge Leval also applies this reasoning to arbitrary marks, which consist of dictionary words that have no semantic connection to the product to which they are affixed (for example, BLACKBERRY for mobile phones). But in light of the problems of depletion and crowding that we identify, we think that the reasoning now only properly applies to coined terms. 310 See Beebe & Fromer, supra note 8, at 1031-33. 311 See id. at 1032-33. Trademark offices could also increase fees for registration of a mark in additional classes of goods and services. As we discussed above, supra note 111, the EUIPO abandoned its flat fee for registration in up to three Nice classes in 2016. The data show that this reform has curtailed multiclass registrations.

*p. 89*
Vol. 112 TMR challenges. 313 Other registration-based trademark systems around the world have similarly lax use requirements. 314 This permissive approach to the use requirement may once have been sensible when there appeared to be an inexhaustible supply of trademarks and the granting of trademark rights appeared to be costless. It perhaps mattered little whether a registrant was actually using the mark for any or all of the goods or services it specified because there were so many alternative marks available to others. But as we have sought to show, that era has passed. Empirical studies reveal that the multinational multilingual EUIPO register now suffers from significant levels of trademark clutter of unused marks, particularly as compared to the USPTO register. 315 For example, Georg von Graevenitz and other scholars have found that 6% of EU marks for pharmaceuticals are not being used 316 and that EU marks claim 50% more goods and services than identical marks registered at the USPTO, where use in commerce is required to protect or register a mark. 317 Clearing out this clutter promises to substantially mitigate trademark depletion and crowding in the EU trademark system.

*p. 89*
The most effective way that the EUIPO and other registrationbased trademark systems can reduce clutter is by taking steps to actually enforce the use requirement. Take the EUIPO. A simple first step would be to require EUIPO applicants to submit a declaration at the time of application explicitly stating that they either have a good-faith intent to use or are already using the applied-for mark in connection with all the goods and services specified in the application. That EU trademark law currently requires no such declaration might surprise American trademark lawyers, who must submit such declarations to the USPTO at the time of application. 318 A second simple step after registration would be to require registrants to submit declarations every ten years during the registration period-as USPTO registrants are required to do 319 -attesting to the fact that they are using the registered 313 Supra text accompanying note 99. 314 Indeed, Canada has recently amended its trademark law to eliminate its previous requirement that an applicant must submit a declaration that it is making an actual use of its mark in commerce for a registration of the mark to issue. Josh Gerben, Canada 's June 2019 Trademark Law Changes-"Use" Is No Longer Required and Other Highlights, Gerben, https://www.gerbenlaw.com/blog/canadas-june-2019-trademarklaw-changes-use-is-no-longer-required-and-other-highlights (last visited June 17, 2022). mark in connection with all of the goods and services claimed in the registration. To be sure, many EUIPO registrants will oppose such administrative requirements as unnecessary, 320 and it may once have been unnecessary-just as the regulation of fisheries or carbon emissions may once have been unnecessary. But again, the EU trademark system teaches that trademark systems around the world can no longer assume that they consist of an inexhaustible resource and rules and practices based on obsolete assumptions must yield to new realities. 321 An additional way in which the world's trademark offices can reduce clutter is by initiating their own auditing program of registrations to require registrants to submit specimens of use supporting their claims of use. The EUIPO specifically is not empowered to do so. 322 Even if it becomes aware of registered marks that are not being used, it has no means to cancel the relevant registrations. By contrast, as we discuss in previous work, the USPTO undertook a highly successful pilot program that audited registrations for use and found that about half of all marks registered at the USPTO were not being used in commerce as claimed. 323 The USPTO has since made this auditing program permanent. 324 We expect that a comparable program at the EUIPO and other trademark offices around the world would find even higher levels of non-use. 325

## C. The Virtues of Trademark Depletion and Crowding?

*p. 90*
We conclude this Part by addressing an interesting and important counterargument that may be levelled against our claim that heightened levels of trademark depletion and crowding impair competition by impeding market entry and further harm consumers 320 Max Planck Study, supra note 81, at 88-89 (discussing opposition to a periodic declaration-of-use requirement).

## 321

*p. 90*
In the recent case of Sky Plc. v. SkyKick UK Ltd., Case C-371/18 (Jan. 29, 2020), the Court of Justice of the European Union was given the opportunity, which it spurned, to establish a principle that bad-faith registrations of very broad categories of goods or services will be cancelled either in whole or part. Commentators have since strongly criticized the SkyKick opinion, and justifiably so. E.g., Darren Meale, SkyKick: The Disappointment of the Decade, 15 J. Intell. Prop. L. & Practice 227, 227 (2020) (commenting that in the wake of SkyKick, "the trade mark registers of Europe will keep expanding and specifications will not be getting shorter. Eventually something (perhaps my clients' patience for brand clearance) will break.").

*p. 90*
322 Supra text accompanying note 97. 323 Beebe & Fromer, supra note 8, at 1034-35. 324 Id.

## 325

*p. 90*
Of course, requiring the EUIPO to review marks for use would be costly, but it can pass through this cost to registrants, just as the USPTO does. As explained below, EUIPO fees are currently exceptionally low. Infra section 5.

*p. 91*
Vol. 112 TMR by increasing their search costs. This counterargument asserts that, on the contrary, depletion and crowding may result in net benefits for fair competition and consumers. The assertion takes two very different forms.

*p. 91*
The first is that heightened levels of depletion and crowding may actually produce a net benefit precisely because they impede market entry, and by doing so, they reduce further artificial product differentiation in the marketplace. The underlying assumption that drives this view is that new trademarks do not necessarily represent new goods or services. Rather, they may represent closely similar if not entirely fungible goods and services that are artificially distinguished by multiple different brands. 326 On this view, extreme levels of depletion and crowding are better understood as a sign that the market is already overfull of unnecessary trademarks. Many of these marks may confuse consumers, but not in the way traditionally understood. Instead, they confuse consumers in that they lead consumers to believe that various products are originating from different sources or possess different characteristics when in fact they are all originating from the same source or possess exactly the same characteristics. If this description of the marketplace is accurate, then initiating reforms that facilitate brand entry would unnecessarily increase consumer search costs with no offsetting benefits to consumer welfare. Importantly, the sensible version of this counterargument would hold that the circumstances it describes are not found in all market sectors, but primarily in those where depletion and crowding have reached their most extreme levels-for example, in the apparel fashion sector.

*p. 91*
The second, more speculative form of this counterargument is that the processes of trademark depletion and crowding will eventually result in a net benefit for consumers because these processes will result in a general breakdown of the global trademark system. As certain market sectors become more and more crowded with ever-less-distinctive brands, these sectors may reach an inflection point beyond which consumers come to see all the brands in the sectors, if not all brands, as indistinguishable ambient noise. 327 The result will be "peak trademark," not in the sense of "peak oil," 328 but of "peak Kardashian," 329 after which the consumer economy of factitious distinctions, having reached its extreme, will finally retreat to another incarnation. Perhaps, as is periodically Vol. 112 TMR 991 predicted, the result will be the "end of brands," 330 this time with consumers relying largely on online product reviews or barcodes. 331 A full consideration of these arguments and the cost-benefit analyses underlying them is beyond the bounds of this article, but we find them unpersuasive, at least in the short run. It is important to recognize that trademarks themselves function as products and consumers have long revealed a preference for purchasing such products, even when the underlying material good is little more than an alibi for the consumption of the brand it carries. Artificial product differentiation is not necessarily artificial to consumers who are willing to pay for it. 332 In any case, we find evidence of heightened levels of trademark depletion and crowding not merely with respect to apparel goods or other usual suspects that appear (whether fairly or not) in the artificial product differentiation lineup. We also find them in connection with high-technology goods and services and indeed across all classes of services, where the artificial product differentiation argument is less compelling. Finally, reports of the imminent death of brands are likely exaggerated. As the lingua franca of a global marketplace, they serve primal commercial and social purposes, including conveying information succinctly and colorfully, serving to express social distinction, and personifying businesses.foot_26 * * * In this Part, we have used the example of the EUIPO to propose four main reforms that may aid trademark systems in coping with increasing levels of trademark depletion and crowding in a globally integrating marketplace: (1) elimination of-or at least a more restrained application of-translational similarity doctrine, (2) institution of examiner review of confusing similarity where it is not currently used, or at least the provision of better information to current registrants about conflicting applications and the true status of the trademark register, (3) reformed fee structures calibrated to the claiming of valuable terms, such as multilanguage words, and (4) more effective enforcement of the use requirement. Other reforms may also be promising. In particular, we support an overhaul of the Nice classification scheme to organize the classes of goods and services in a manner that corresponds to the current marketplace rather than the marketplace of the late nineteenth century. We also support the creation of more numerous and more specific classes than the forty-five currently provided by the Nice scheme. At the very least, such reforms would allow registering 330 Supra note 262.

## Footnotes

> John Locke, Second Treatise of Government 11 (Jonathan Bennett ed., 2017) (1690).

> Infra text accompanying notes 270-277.

> von Graevenitz, Ashmead & Greenhalgh, supra note 101.

> This includes registrations in the EUIPO and registrations in any of the national trademark offices.

> EUIPO, eSearch Case Law, https://euipo.europa.eu/eSearchCLW (last visited June 17, 2022).

> Steven T. Piantadosi, Zipf's Word Frequency Law in Natural Language:A Critical Review and Future Directions, 21 Psychonomic Bull. & Rev. 1112 (2014).

> The current website of the General Inquirer may be found here: http://www.mariapinto.es/ciberabstracts/Articulos/Inquirer.htm (last visited December 4, 2022).

> Rex W. Huppke, Craft Beer Makers Running Out of Names. How About Flip Donkey Doodleplunk?, Chi. Trib., Jan. 7, 2015, https://www.chicagotribune.com/columns/rexhuppke/ct-craft-beer-names-huppke-talk-0107-20150107-story.html.

> Luke Timmerman, Why Are Drugs Getting Such Weird Brand Names?, Xconomy, May 9, 2011, https://xconomy.com/national/2011/05/09/why-are-drugs-getting-such-weirdbrand-names.

> Fhima & Gangjee, supra note 89, at 20-22.

> Winkler, supra note 152.

> We present the English results again for ease of comparison.

> Beebe & Fromer, supra note 8, at 999-1000.

> Beebe & Fromer, supra note 8 at 1038-39.

> Smith & Compton, supra note 222, at 39. An important exception is the French IP Office, which lists coexistence agreements as part of the documents to include in a trademark application. Institut National de la Propriété Industrielle, Faire Vivre Votre Marque, https://www.inpi.fr/fr/valoriser-vos-actifs/faire-vivre-votre-marque/transmettre-ouexploiter-une-marque (last visited June 17, 2022).

> Zoom Interview with Imogen Fowler (Aug. 31, 2020).

> But see In reMorton-Norwich Prod., Inc., 671 F.2d 1332, 1336 (C.C.P.A. 1982).

> Beebe & Fromer, supra note 8, at 1022.

> Supra section I.B.2.

> Supra text accompanying notes 95-98.

> 27Id. https://www.ideasbig.com/worst-brand-name-of-2019 ("[Y]ou need a name 'that is also capable of evoking pleasant feelings and hasn't been taken by an internet squatter.'" (quoting Andrew Essex, chief executive of branding consultancy Plan A)). 32 W. Scott Blackmer & Sara Skinner Chubb, Brand Protection Today-Article 1: Choose Wisely, InfoLawGroup, Feb. 3, 2021, https://www.infolawgroup.com/insights/2021/2/3/ brand-protection-today-article-1.

> 50 Tsedal Neeley, 51 Abram de Swaan,

> 116Google Cloud, Cloud Translation Documentation, https://cloud.google.com/translate/docs (last visited June 17, 2022).

> 123Top-Level Domain Zone File Information, Verisign, https://www.verisign.com/ en_US/channel-resources/domain-registry-products/zone-file/index.xhtml. 124 Walter J. B. van Heuven, Pawel Mandera, Emmanuel Keuleers & Marc Brysbaertt, A New and Improved Word Frequency Database for British English, 67 Quarterly J. ExperimentalPsych. 1176 (2014). The SUBTLEX-UK data are available at http://crr.ugent.be/archives/1423. Though we focus on British English in our reported results, we also used the Corpus of Contemporary American English and the wordfrequency data provided by its developer to test for any differences in British as compared to American English. Word Frequency Data, Word Frequency Data: Based on 450 Million Word COCA Corpus, https://www.wordfrequency.info/100k.asp (last visited June 17, 2022); Mark Davies, The Corpus of Contemporary American English as the First Reliable Monitor Corpus of English, 25 Literary & Linguistic Computing 447 (2010). We found no notable differences. 125 Marc Brysbaert, Matthias Buchmeier, Markus Conrad, Arthur M. Jacobs, Jens Bölte & Andrea Böhl, The Word Frequency Effect: A Review of Recent Developments and Implications for the Choice of Frequency Estimates in German, 58 Experimental Psych. 412 (2011). The SUBTLEX-DE data are available at http://crr.ugent.be/archives/534. ; Leibniz-Institut fur Deutsche Sprache, DeReWo-Corpus-Based Lemma and Word Form Lists, https://www.ids-mannheim.de/digspra/ kl/projekte/korpora/ (last visited June 17, 2022). We found no notable differences in our results. 126 Boris New & Christophe Pallier, Manuel de Lexique 3, http://lexique.org/ _documentation/Manuel_Lexique.3.2.pdf. The Lexique 3 data are available at http://www.lexique.org. 127 David Crepaldi,

> 152 William E. Winkler,

> 160Comparable results from the USPTO for active trademark registrations in 2017 are slightly lower in terms of word usage covered by registered marks. For example, of the 20,000 most frequently used words in American English, 10,453 identically matched an active trademark registration at the USPTO in 2017, but these words accounted for 73.8% of total word usage in American English. For further comparative results, seeBeebe & Fromer, supra note 8, at 982.

> 174 EU Trade Mark No. 008545899. 175 EU Trade Mark No. 008590184; EU Trade Mark No. 016871956.

> 177 E.g., 182 EU Trade Mark No. 009647587. 183 EU Trade Mark No. 013098066. 184 EU Trade Mark

> 189We derive the census data from

> 254 258 Mary Hanbury,

> 305Supra text accompanying note 301. 306 E.g., 307 E.g., id.