The Fairy Tale of the Average Consumer: Why We Should Not Rely on the Real Consumer When Assessing the Likelihood of Confusion

The present article considers whether the confusion analysis in trademark law is at risk of being used strategically as a self-servicing mechanism by the industry to obtain trademark rights to descriptive, cultural and non-traditional signs. In this context, two features of the trademark system are particularly worrisome. First, trademark owners can strengthen the distinctive character of their marks by investing in marketing and branding campaigns. Second, trade- mark owners can afford expensive surveys demonstrating the high distinctiveness of their marks and strengthening their position in an infringement case. A study of 189 Dutch cases reveals that in the likelihood-of-confusion assessment, the degree of a mark’s distinctiveness and corresponding empirical ﬁndings can have a deep impact. However, the study also shows that descriptive, cultural and non-traditional signs were not often disputed. In practice, the risk of misappropriating these signs might therefore be limited. In some cases, however, trademark owners did succeed in achieving a broad scope of protection regarding descriptive, cultural and non-traditional signs. Therefore, national courts should have more room to make normative corrections in favour of freedom-of-competition interests. This would allow judges to prevent trademark owners from achieving unjustiﬁed economic advantages ﬂowing from these signs.


I. Introduction
A central question in trademark infringement cases is whether the defendant's sign is identical or similar to the claimant's trademark and therefore likely to confuse the average consumer as to the origin of the goods or services. 1 If the judge answers in the affirmative, the trademark owner can prohibit the defendant from using the sign that has been found to be confusingly similar. Because of the crucial role of consumer perception and its deep impact on the infringement analysis, it is of utmost importance to be clear about how we assess a likelihood of confusion.
Should we assess consumer behaviour factually (empirical approach), or should we leave room for assumptions about how consumers ought to behave in the marketplace (normative approach)? 2 In either approach, national courts have considerable freedom to substantiate their decisions with normative considerations and empirical findings. After all, according to the CJEU, the judge must assess the likelihood of confusion globally, taking into account all factors relevant to the circumstances of the case. 3 This article addresses the potential risk of overreliance on empirical findings. Without normative corrections, EU trademark law may become a self-servicing mechanism for the industry. The infringement analysis may especially become imbalanced if judges were only interested in empirical studies seeking to demonstrate the existence of a likelihood of confusion or the reputation of the trademark. 4 Trademark owners often have the financial resources necessary to invest in expensive marketing campaigns to educate consumers to perceive a sign as a source identifier. The more a trademark owner invests in advertising presenting a sign as an identifier of commercial source, the broader will be the scope of protection. Consumers will view a trademark more positively after repeated exposure. 5 The more the trademark is seen as a symbol of a desirable lifestyle, the more willing consumers will be to pay extra for the brand image that the trademarked product evokes. 6 The stronger a trademark, the easier it is to prove a likelihood of confusion. 7 Research suggests that the success of a well-known mark is primarily related to its familiarity and not so much to the characteristics of the product itself. Besides, consumers are also inclined to interpret information in line with their expectations. If a consumer has positive feelings towards a particular trademark, she/he will also interpret following exposures to the trademark positively. Familiarity is therefore a strong predictor of consumer decisions. 8 Hence, there is the prospect of an attractive return on investment. Furthermore, trademark owners can invest in expensive empirical studies to demonstrate a highly distinctive character and the existence of a likelihood of confusion.
A system biased towards trademark owners' investments could harm competitors' freedom of competition. Exclusive trademark rights on fantasy marks and other fanciful forms of marks need not create substantial entrance barriers for competitors, but descriptive, cultural and non-traditional signs do. 9 These signs give trademark owners economic advantages that go beyond their own investment in the sign, and should be kept freely available for competitors on the market. 10 It is, however, unclear to what extent trademark owners strategically use the current trademark system to strengthen their own gains in this way. To bring light into the darkness, several cases will be analysed to explore how precisely national courts assess likelihood-ofconfusion questions in practice and whether they follow a more normative or empirical approach to consumer perception.
I first discuss the rationales of trademark law in the following Section II. In Section III, I offer an overview of previous empirical studies of infringement cases. Section IV describes the methodology that was followed. In Section V, I present the results of the empirical analysis of 189 Dutch cases. A more detailed analysis of the results follows in Section VI. I also discuss an exemplary study of UK cases (Section VII) to find out whether similar conclusions can be drawn from other trademark systems relying on CJEU jurisprudence. Section VIII concludes.

II. Trademarks and free competition
The main function of trademark law is to ensure market transparency. In a transparent market, distinctive signs exclusively indicate goods or services from a single source of origin. In such a marketplace, trademark proprietors can clearly identify their offer, and consumers can quickly identify and choose their preferred goods or services. Signs that deceive consumers reduce the efficiency of the market because consumers inevitably purchase the wrong product. From an economic perspective, trademarks reduce consumers' search costs. 11 In addition, distinctive marks encourage trademark owners to keep investing in product quality. If trademark owners lower their products' quality, consumers are likely to be disappointed and may subsequently purchase another product. In particular with regard to unobservable product features, consumers are dependent on trademarks. Without trademark protection competitors would mainly invest in product features that consumers can observe and consumers would only pick by chance their preferred product. Furthermore, a transparent market offers the public's preferred goods and services and does so in sufficient supply. 12 In order to ensure market transparency, we therefore seem to need strong trademark protection that prevents competitors from producing confusingly similar signs. Broad trademark protection encourages trademark proprietors to massively invest in marketing and branding campaigns resulting in lower search costs for consumers, and sufficient quality of available products. From this perspective, granting strong trademark protection seems entirely unproblematic.
However, to achieve undistorted competition, trademark protection must always be balanced against other core values, such as free expression that enhances consumer information and consumer choice, 13 and free competition that prevents unnecessary market entry barriers. The recognition of a need to keep signs freely available offers competitors equal access to communication tools that can be used to inform consumers about product characteristics and compete for market shares. 14 The fundamental principle of the need to keep free is reflected in several provisions in the Trade Mark Directive and the EU Trade Mark Regulation and recognized by the CJEU in several cases. 15 For example, signs that fall under the grounds for refusal are excluded from trademark protection. 16 Furthermore, it must be shown that the use of the sign is made 'in the course of trade' and 'in relation to goods or services'. These protection requirements ensure that trademark rights do not affect social and cultural forms of use. The likelihood-ofconfusion provision is directly related to the preservation of fair and undistorted competition. 17 Trademark limitations also aim to strike a balance between trademark proprietor's interests and free movement of goods and services within the internal market, and the competitor's interests in using a sign for legitimate purposes in accordance with honest practices. For instance, third parties are allowed to use descriptive signs identical or similar to protected trademarks or their components, where the trademark proprietor's legitimate interests are not adversely affected. 18 In principle, trademark law and free competition do not conflict with one another. Trademarks provide information about the commercial source of goods or services and enable consumers to repeat satisfactory purchases. Trademark owners are stimulated to keep the quality of their goods or services high and to invest in their goods and services. Competitors can choose a different sign if they want to freely offer their goods and services on the market. 19 However, in practice, competition can only be neutral if the appropriation of the sign as such does not grant trademark owners a competitive advantage from which third parties are excluded. This requirement is largely fulfilled with respect to (fantasy) word marks and other traditional forms of trademarks. These signs are often available in sufficient supply, 20 such that third parties do not experience entrance barriers in competing with similar products and services. 21 By contrast, descriptive, cultural and non-traditional signs are not available in unlimited numbers. Exclusive trademark rights may affect the availability of these signs on the market and generate obstacles to competition.
A trader who has acquired exclusive rights in a word that describes a positive feature of a good obtains an unfair economic advantage because consumers easily understand what this trademark stands for. 22 The costs for competitors, by contrast, are higher. They must inform 12 Kur and Senftleben (n 10) 7-9; Economides (n 11) 526; Landes and Posner (n 11) 166 consumers of the fact that their goods have this same feature while not enjoying the freedom of using the descriptive word as a mark in their brand name or advertising. 23 Appealing signs, such as colours, patterns and shapes contribute to the style and appearance of a specific product and therefore also give trademark proprietors a competitive advantage that goes beyond the sign as such. 24 Cultural signs may evoke positive associations which they have as a result of their evolution in the literary and artistic domain. Adopting such a sign, the trademark owner can benefit from these pre-existing connotations. 25 Trademark rights on these signs are likely to hinder innovation and encourage standardisation. 26 Competitors might be afraid of risking a legal procedure and as a result avoid using particular signs on their products. 27 In addition, brand owners have been obliged to use the same signs to keep the recognition level among consumers high. This could consequently lead to fewer investments in long-term product quality and the development of new products. 28 As long as descriptive, cultural and nontraditional signs can acquire distinctiveness through use in trade, 29 trademark owners will invest in these appealing signs which give them a competitive advantage.
In light of this, it is worrisome to see trademark owners strategically using the current trademark system to achieve the desired degree of protection. This has been particularly true since the Adidas/Marca case, in which the CJEU explicitly refused to consider the need to keep signs freely available for competitors on the market in the context of the confusion analysis. 30 As a result, national courts are not allowed to make normative corrections in favour of freedom-of-competition interests in the likelihood-of-confusion assessment. If trademark owners succeed in making descriptive, cultural and nontraditional signs a source identifier, they achieve an overbroad competitive advantage over third parties. We must therefore be cautious when giving weight to empirical findings. Trademark owners can strengthen the distinctive character of their marks by investing in marketing and branding campaigns. If the infringement analysis automatically rewards these investments by granting broader protection, we are quite close to a selfservicing mechanism that allows trademark owners to shape consumer perception in a way that makes it possible to claim trademark rights to descriptive, cultural and non-traditional signs. Moreover, trademark owners can also afford expensive surveys demonstrating the high distinctiveness of their marks. In this way, they can strengthen their position during an infringement case. Again, this can create a self-servicing mechanism within trademark law.

III. Previous studies
Previous studies have shown that the most important factors were similarity of signs and goods or services. Consumer surveys were not used on a regular basis so the risk of biased decisions favouring trademark owners' investments might be limited. Nevertheless, it cannot be excluded that trademark owners' investments did influence the likelihood-of-confusion assessment. The distinctiveness factor might have played an indirect role in the assessment of similarity of signs.
For example, in an empirical study conducted in the United States, Barton Beebe analysed 331 cases of federal district courts opinions during a five-year period from 2000 to 2004, using a multifactor test for the likelihood of consumer confusion. Beebe found that a number of factors were decisive, while the rest was redundant. 31 This study revealed that similarity of marks was by far the most influential factor. Two other factors were decisive: the defendant's intent factor, but only when it favoured likelihood of confusion, and the proximity of the parties' good factor, but only when it disfavoured likelihood of confusion. 32 By contrast, survey evidence played a minor 23 Nevertheless, third parties are allowed to use descriptive signs identical or similar to protected trademarks or their components, where the trademark proprietor's legitimate interests are not adversely affected (art 14(1)(b) TMD 2015; art 14(1)(b) EUTMR). However, the use of the sign is only justified if it is in accordance with honest commercial practices; a principle which constitutes in substance the expression of a duty to act fairly in relation to the legitimate interests of the trade mark owner. Since the limitations are substantially informed by factors reflecting the infringement criteria, such as consumer confusion, the use of similar signs is more likely to contrast with honest commercial practices the more has been invested in a particular sign. See also Annette Kur, 'Yellow Dictionaries, Red Banking Services, Some Candies, and a Sitting Bunny.  role: it was only submitted in 20% of court decisions. 33 Furthermore, the inherent strength factor, i.e. the distinction between fanciful, arbitrary, suggestive, descriptive and generic marks was not important in the courts' opinion. On this point, legal doctrine diverged remarkably from the investigated district court cases.
Nevertheless, where courts did address the strength factor, in particular inherent strength, there was a surprisingly strong correlation between inherent strength and success in the multifactor test. In addition, in order to win the multifactor test, it seems that the plaintiff must not lose the strength factor. In most cases where the plaintiff lost this factor, it also lost the overall test. This also seems to apply in reverse: in 90% of the 102 judgments that found a likelihood of confusion, the strength factor favoured this result. 34 In an empirical analysis of likelihood-of-confusion factors in European trademark law, Ilanah Fhima and Catrina Denvir analysed 136 General Court decisions during a three-year period from 2009 to 2012. Fhima and Denvir found that the most important factors were similarity of marks and goods: overall confusion was never found unless both the parties' marks and their goods were similar. 35 On the contrary, the distinctiveness factor was generally ignored, or deemed irrelevant. Inherent distinctiveness was not considered in 68% of the cases, and acquired distinctiveness was not addressed in 82% of the cases. Inherent distinctiveness only favoured confusion in 7% of the cases where likelihood of confusion was found, as compared to 6% regarding acquired distinctiveness. Nevertheless, the authors suggested that the distinctiveness factor may have had a limited impact because it also played a key role in the similarity of mark analysis via the dominant elements of both parties' signs. In the authors' opinion, it may seem odd to displace the factors of similarity of signs and goods or services with an additional factor that is not fully understood. 36 In an empirical study conducted in Australia, Huang, Weatherall and Webster found that survey evidence was rarely used in Australian case law, namely in 10.4% of cases from 1990 to 2010. 37 Their analysis pointed out that surveys had only limited impact on the results of the cases surveyed. 38 Graeme Austin also indicated that survey evidence rarely carried weight in trademark infringement cases in the United States. 39 A Dutch survey among judges of preliminary relief 40 further showed that in practice, the judges' own perception was the most important basis for the infringement decision, while expert opinions and public studies did not rank very high. Actually, these judges were more likely to have a negative attitude towards empirical evidence because they believed this evidence might be manipulated and contained leading or poorly formulated questions.
In other words, the risk of overreliance on empirical findings might be limited. In previous studies, market analyses were rarely used and seldom decisive to the outcome of the decision. Furthermore, the distinctiveness factor seems to have played a modest role. In these studies, similarity of marks, and goods or services, were the core factors in the likelihood-of-confusion assessment. Nevertheless, the finding that the factor of distinctiveness was not considered in the majority of cases need not imply that it did not impact the outcome of the case. It may be that the distinctiveness factor was not mentioned because the high degree of distinctiveness was self-evident. Furthermore, the distinctiveness factor might have played an indirect role in the assessment of similarity of signs. The global appreciation of the visual, auditory or conceptual similarity of the marks in question must be based on the overall impression given by the marks, bearing in mind, in particular, their distinctive and dominant components. 41 The more distinctive the trademark, the more similar the marks will be, in particular if the defendant uses the same highly distinctive elements. Therefore, we cannot rule out that trademark owners' high investments in marketing and branding campaigns may yet have a major impact on likelihood-of-confusion cases.

IV. Methodology
To uncover the impact of the degree of a mark's distinctiveness and corresponding empirical findings on the likelihood-of-confusion assessment, I performed a data analysis of Dutch cases. In comparison to other EU countries, Dutch cases are of particular interest because of their regional character. The trademark legislation applied by Dutch courts is drawn from a single Treaty -the Benelux Convention on Intellectual Property (Trademarks and Designs) -which has direct effect in three EU Member States.
Furthermore, Benelux law has traditionally been an important element in harmonising EU trademark law. The impact of the Benelux tradition can still be seen in the anti-confusion provision in the EU Trade Mark Directive, which states that 'the likelihood of confusion includes the likelihood of association between the sign and the trade mark'. 42 This phrase was the result of a political compromise.
Traditionally, the Benelux countries adopted a requirement of association when assessing likelihood-ofconfusion cases, which meant that likelihood of confusion was also assumed when consumers merely associated the use of a third party's sign with a trademark. The final phrase in the EU Trade Mark Directive was the result of the efforts of the Benelux countries to include their association requirement in the anti-confusion provision. Nevertheless, the CJEU ruled in the Puma/Sabel case that  The Fairy Tale of the Average Consumer risk of association was a relevant factor, but not sufficient to conclude likelihood of confusion. 43 Considering this interaction between harmonised Benelux and EU law, it can be deemed an important first step to look at court decisions following from harmonised Benelux law, as reflected in Dutch case law. Providing this groundwork, the current analysis can encourage further studies relating to the situation in other EU Member States.
A total of 189 cases were used in this study. The infringement cases were found in the Dutch databases 'Kluwer Navigator', 'Legal Intelligence' 44 and 'IE-Forum.nl'. The data for the study comprise cases decided by district courts and courts of appeal over the period For an initial selection, generic search terms were entered, such as 'verwarringsgevaar' (English: 'likelihood of confusion'), 'inbreuk' (English: 'infringement'), 'verwarring' (English: 'confusion') and so forth. However, this did not lead to a practicable result. Subsequently, a more refined search method was chosen. The search terms used were 'artikel 2.20 lid 1 sub b BVIE' 45 , 'artikel 9 lid 1 sub b GMVo', 46 'artikel 9 lid 2 sub b UMVo' 47 and variants (hereafter 'sub b'). It can be presumed that courts mention these search terms when assessing the likelihood of confusion, which makes it highly likely that these search terms can be found in similarity cases.
After entering search terms, double cases were filtered out, as were infringement cases that did not contain likelihood-of-confusion questions. For example, the judge did not discuss whether a likelihood of confusion existed. If preliminary questions were referred to the CJEU, the case was also not included. Moreover, Supreme Court cases were not included because the Dutch Supreme Court (Hoge Raad) is not a court that establishes new facts. Finally, bibliographic data were noted, such as the date and citation of the judgement, and the court.
The claimant won when the infringement claim based on sub b was successful. A loss was when the infringement claim had no success. If the trademark owner lost the case because of a limitation of trademark rights, such as freedom of comparative advertising, this was also termed a loss. A win was also coded if the trademark owner won some part of the issues. For example, this scenario would arise where two aspects were claimed to be infringing under sub b, but only one aspect was found to be infringing. Furthermore, when considering court of appeal cases, I looked at the party who was the trademark owner. This was not per definition the same person as the claimant. It was only considered whether or not the anticonfusion claim was successful. The outcome did not say anything about the rest of the claim.
The factors analysed in the present study were: comparison of signs, comparison of goods and services, relevant public and degree of attention, the distinctiveness of the earlier mark, and consumer studies.
Signs were comparatively categorised as 'identical', 'almost identical', 'similar', 'insufficiently similar', and 'not mentioned'. 'Identical' was used when the signs were considered to be identical. 'Almost identical' was used when the court found a very strong similarity between signs. 'Similar' was used when the signs were found to be similar. 'Insufficiently similar' was used when the signs were dissimilar or when the degree of similarity was low. This category was also selected when the judge found the sign to be insufficiently similar to cause confusion. 'Not mentioned' was used when the degree of similarity was unclear. When several signs were claimed to be infringing, and only one sign was considered sufficiently similar, this degree of similarity was coded. When several signs claimed to be infringing, the highest degree of similarity was coded.
The type of mark was also assessed. The analysis only includes trademarks that were assessed under sub b. If the defendant infringed on both the word mark and the figurative mark, but only the word mark was considered in the assessment, only the word mark was coded.
In addition, the guideline that the global appreciation of the visual, aural or conceptual similarity of the marks in question must be based on the overall impression given by the marks, bearing in mind in particular their distinctive and dominant components, 48 was researched. A 'yes' was coded when the court referred to this ruling in the likelihood-of-confusion assessment. A 'yes' was also coded when a reference was made to the distinctive or dominant elements of the signs, or when a similar expression was used. I also investigated the guideline that the average consumer only rarely has the opportunity to make a direct comparison between the different marks and must therefore place his trust in the imperfect image of them that he has in his mind. 49 Furthermore, the guideline of the Dutch Supreme Court stating that similarities between signs are more important than differences was explored. 50 With regard to the factor of relationship of the goods and services involved, this was categorised as 'identical', 'almost identical', 'similar' and 'slightly similar', 'dissimilar' and 'not mentioned'. An 'identical' was selected when the goods and services were considered as identical. An 'almost identical' was selected when the goods and services were regarded as 'identical or similar' or when the goods and services were defined as very similar. A 'similar' was coded when the goods and services were found to be similar. A 'slightly similar' was coded when the goods and services were less similar. A 'dissimilar' was coded when the goods and services were argued to be dissimilar. An 'unspecified' was selected when the judge did not explicitly reason whether there was sufficient similarity of goods and services.
The 'relevant public and degree of attention' factor was also explored. When a reference was made to the average consumer a 'yes' was selected. Furthermore, it was explored whether the court used the measurement invented by the CJEU 51 of the average consumer who is reasonably well-informed and reasonably observant and circumspect, or another consumer, for example an expert who is highly attentive or a consumer who is less intelligent. When the heuristic of the CJEU concerning the average consumer or a similar qualification was found in the reasoning, this was coded as a 'yes'.
Furthermore, the degree of distinctiveness was analysed. The degree of distinctiveness was categorised as 'highly distinctive', 'distinctive', 'slightly distinctive', 'not distinctive', and 'unspecified'. The category 'highly distinctive' referred to signs with a highly distinctive character or signs that were well-known. A trademark was coded as 'distinctive' when there was a normal or sufficient degree of distinctiveness. A 'slightly distinctive' was coded when the judge explicitly considered that the scope of protection was restricted or, for example, that the degree of distinctiveness was not so great. An 'unspecified' was coded when the judge did not state whether or not the sign was distinctive. When several trademarks were invoked in respect of sub b, the highest degree of distinctiveness was coded. The CJEU guideline stating that the more distinctive the earlier mark, the greater the likelihood of confusion was also explored. 52 A 'yes' was coded when the court referred to this ruling in the likelihood-ofconfusion assessment.
When the judge referred to consumer studies in the assessment of likelihood of confusion, a 'yes' was coded. When consumer studies played a role somewhere else, for example to support the assessment of trademarks with a reputation or when there were no consumer studies, a 'no' was selected. These consumer surveys were used to assess likelihood-of-confusion questions. Consumer surveys can prove that a trademark is well known and therefore has a highly distinctive character. A high score regarding the factor 'distinctiveness of the trademark' makes an allocation for the infringement claim more likely. These consumer studies were possibly also used to assess whether an infringement was made under sub c, but this was not coded. A reference to consumer studies was only coded when this reference was found in the assessment of sub b.

V. Results
In the sample studied, trademark owners more often won (65%) than lost (35%) anti-confusion cases falling under sub b of the infringement provision at the EU and Dutch level. 53 Moreover, trademark owners were more successful when they filed cases before specialised courts in The Hague (74%) 54 than elsewhere in the Netherlands (59%) (see Table 1). 55 As expected from previous studies, similarity of signs was indeed the most important factor. The degree of similarity of signs directly influenced the decision.
In most cases, the signs were sufficiently similar (127 cases: 67%). 56 In 35% of the cases, the degree of similarity was even high. 57 In the majority of cases a word mark was assessed in the sample used for the present study (see Table 2). 58 Figurative marks were also frequently assessed. 59 But again, in the majority of cases this concerned a figurative mark containing word elements. 60 Purely figurative marks 61 and shape marks 62 were seldom assessed (see Table 3). 63 Trademark owners had a stronger claim not only when the defendant's sign was partly descriptive, but also when the trademark was highly distinctive. 64  53 The confidence interval of winning 123 out of 189 cases is 58.5<p<70.4%. With 95% reliability, it can be assumed that trademark owners won slightly more cases than they lose. This means we can assume with 95% reliability that the odds of trademark owners winning a case were between 58.5% and 70.4%. In this sample, a percentage of 65% was found (123 cases). This percentage could thus also have been anything between 58.5% (111 cases) and 70.4% (133 cases), eg if a bigger sample had been chosen. See also Fhima and Denvir (n 35) 327. In 60% of cases the General Court found a likelihood of confusion and in 40% of cases no likelihood of confusion was found. See also Lotte Anemaet, 'Beslissen in sub b zaken: hoe beoordelen rechters of een consument in verwarring is of niet?' (2019) 3 Berichten Industriële Eigendom 122. 54 The confidence interval of winning 59 out of 80 cases is 63.2%<p<80.6% (with 95% reliability). 55 The confidence interval of winning 64 out of 109 cases is 49.5%<p<65.9% (with 95% reliability). 56 The confidence interval of selecting 127 cases with (sufficient) similar signs from 189 cases is 60.6%<p<72.4% (with 95% reliability). 57 The confidence interval of selecting 67 cases with identical or highly similar signs from 189 cases is 29.2%<p<41.1% (with 95% reliability). 58 The confidence interval of selecting 90 word marks from 189 cases is 41%<p<53.4% (with 95% reliability). 59 The confidence interval of selecting 56 figurative marks from 189 cases is 23.7%<p<35.1% (with 95% reliability). 60 The confidence interval of selecting 49 figurative marks containing word elements from 189 cases is 20.3%<p<31.2% (with 95% reliability). 61 The confidence interval of selecting seven purely figurative marks from 189 cases is 1.4%<p<6.2% (with 95% reliability). 62 The confidence interval of selecting six shape marks from 189 cases is 1.1%<p<5.5% (with 95% reliability). 63 It must be noted that the number of cases where several trademarks were claimed was in fact slightly higher. The analysis only included trademarks that were assessed under sub b. 64 Cases where the distinctiveness of the trademark directly influenced the classification of the degree of similarity of signs, in a way that similarity was found: in reverse: if the trademark consisted of descriptive elements, the judge was more likely to conclude that the signs were not confusingly similar, particularly if the defendant had also added their own distinctive elements to the sign. 65 Nevertheless, trademark owners did not always lose when the trademark consisted of descriptive elements. In some cases, the judge argued that the signs were similar because both signs referred to the same conceptual meaning. 66 The court often referred to the CJEU ruling stating that the global appreciation of the visual, auditory or conceptual similarity of the marks in question must be based on the overall impression given by the marks, bearing in mind, in particular, their distinctive and dominant components (124 cases: 66%). 67 In these cases, the claimant frequently won when the trademark was distinctive or highly distinctive. 68 However, claimants also won cases when the trademark was slightly distinctive, i.e. in cases where the application of the rule ought rather to have led to a loss because the trademark did not actually consist of highly distinctive elements. 69    67 In 82 cases the ruling was explicitly mentioned by the court. In the other cases a reference was made to the distinctive or dominant elements, or a similar expression was used. 68 The confidence interval of selecting 60 win cases from 74 cases that concerned a (highly) distinctive trademark and a reference to the CJEU ruling was found is 70.6%<p<88.1% (with 95% reliability). 69 In 13 cases (out of 29 cases), the claimant of a slightly distinctive trademark won, even though a reference to this rule was found. The Another guideline of the CJEU about the average consumer was that he only rarely had the chance to make a direct comparison between the different marks but must place his trust in the imperfect picture of them that he holds in his mind. 70 In some cases where this rule was found, 71 the claimant won. 72 In addition, judges sometimes applied the Bigott-Batco/ Doucal rule of the Dutch Supreme Court (43 cases: 23%). 73 According to this rule, the similarities between the trademark and the sign are more important than the differences. 74 When courts referred to this rule, it seems to have strengthened a trademark owner's claim: in almost all cases in which this rule was found, the trademark owner won. 75 Similarity of goods and services was also a relevant factor. In the majority of cases, the goods and services were sufficiently similar (137 cases: 72%), 76 while being dissimilar in only a few cases (10 cases: 5%) 77 (see Table 4). 78 The claimant did not go to court with an anticonfusion claim if the goods and services were clearly dissimilar. In many cases, similarity of the goods and services was even high or identical. Nevertheless, trademark owners did not necessarily win more often when the goods and services were highly similar. 79 In most cases, the judge referred to the average consumer with a standard degree of attention. In 87 cases (46%) (see Table 5), 80 the judge referred explicitly to the average consumer, while in 29% of the cases, the judge mentioned that the average consumer was reasonably well informed, observant and circumspect. 81 Only in some cases did the judge explicitly refer to a consumer with a higher or lower degree of attention. 82 Judges might be unwilling to label consumers as being of limited competence. 83 The distinctiveness of the trademark, however, was an important factor. In most cases, the trademark was sufficiently distinctive (148 cases: 78%) (see Table 6). 84 The degree of distinctiveness of the trademark was often high or average (107 cases: 57%), 85 while in 22% of the cases the trademark was slightly distinctive. 86 In a third of the cases, the trademark was considered to be highly distinctive, 87 a factor that helped the trademark owner win significantly more often. 88 Conversely, trademark owners lost (slightly) more often when the trademark was averagely 89 or slightly distinctive. 90 Furthermore, courts did not often explicitly refer to the rule that consumers are more easily confused when confronted with a highly distinctive trademark. 91 However, if the trademark was indeed highly distinctive, this strengthened the trademark owner's position. Out of 25 cases, 92 the trademark owner won 21 cases 93 and lost only four. 94 Moreover, judges frequently concluded that there was a likelihood of confusion by emphasising the highly distinctive character, for instance in their final conclusion ('Given the high degree of distinctiveness, it can be concluded that consumers are likely to be confused'). When confidence interval of selecting 13 win cases with slightly distinctive trademarks from 29 cases is 25.9<p<58.3% (with 95% reliability). 70 See Case C-342/97 Lloyd v Loints ECLI:EU:C:1999:323, para 26. 71 The confidence interval of selecting 29 cases with a reference to the CJEU rule from 189 cases is 10.8%<p<19.8% (with 95% reliability). 72 The confidence interval of selecting win 20 cases with a reference to the CJEU rule from 29 cases is 48.6%<p<80.0% (with 95% reliability). 73 The confidence interval of selecting 43 cases with a reference to the Bigott-Batco/Doucal rule from 189 cases is 17.4%<p<27.8% (with 95% reliability). Because of the similar parts of the signs, the average consumer will conclude that the products are connected with each other. 74 Dutch Supreme Court, 16 April 1999, IER 1990, 161 -Bigott-Batco/ Doucal. 75 The confidence interval of selecting 37 win cases with a reference to the Bigott-Batco/Doucal rule from 43 cases is 71.6%<p<92.2% (with 95% reliability). 76 The court did not consider the degree of similarity of goods and services in 42 cases. The judge may have found it self-evident and thus not relevant in assessing the degree of similarity of goods. Another reason could be that the court had already decided that there was no likelihood of confusion based on other factors (the confidence interval of selecting 42 cases without a classification of goods and services from 189 cases is 16.9%<p<27.3%, with 95% reliability). 77 The confidence interval of selecting ten cases with dissimilar goods and services from 189 cases is 2.6%<p<8.2% (with 95% reliability). 78 In the Canon v Cannon case the CJEU ruled that a lesser degree of similarity between goods and services can be compensated for by a greater degree of similarity between marks (Case C-39/97 ECLI:EU:C:1998:442, para 17). The CJEU ruling was sometimes mentioned, but odds were slim that a small degree of similarity of goods and services would be compensated for. The judge considered the goods and services to be highly similar in many cases but slightly similar in only a few cases. 79 The confidence interval of selecting 77 win cases with identical or highly similar goods and services from 96 cases is 71.2%<p<85.8% (with 95% reliability). The confidence interval of selecting 35 win cases with (slightly) similar goods and services from 41 cases is 70.4<p<91.8% (with 95% reliability). 80 The confidence interval of selecting 87 cases with a reference to the average consumer from 189 cases is 39.4%<p<51.8% (with 95% reliability).

81
The confidence interval of selecting 54 cases with a reference to the average consumer who is reasonably well informed, observant and circumspect from 189 cases is 22.7%<p<34% (with 95% reliability). 82 This is in line with CJEU case law: Case C-342/97 Lloyd v Loints ECLI:EU:C:1999:323, para 26. The confidence interval of selecting 27 cases with a reference to a consumer with a higher of lower degree of attention from 189 cases is 9.9%<p<18.6% (with 95% reliability). The confidence interval of selecting 148 cases with sufficiently distinctive trademarks from 189 cases is 72.3%<p<82.7% (with 95% reliability). In 38 cases the court did not mention the degree of distinctiveness. However, the degree of distinctiveness was not always unclear. When infringement was ruled, it can be presumed that the trademark was sufficiently distinctive. The judge apparently found it not necessary to assess the degree of distinctiveness in all cases. 85 The confidence interval of selecting 107 cases with (high) distinctive trademarks from 189 cases is 49.9%<p<62.2% (with 95% reliability). 86 The confidence interval of selecting 41 cases with slightly distinctive trademarks from 189 cases is 16.4%<p<26.7% (with 95% reliability). In a few cases the trademark was defined as not distinctive. 87 The confidence interval of selecting 57 cases with highly distinctive trademarks from 189 cases is 24.2%<p<35.6% (with 95% reliability). 88 The confidence interval of selecting 50 win cases with highly distinctive trademarks from 57 cases is 76.1%<p<92.9% (with 95% reliability). 89 The confidence interval of selecting 35 win cases with averagely distinctive trademarks from 50 cases is 55.6%<p<78.8% (with 95% reliability). 90 The confidence interval of selecting 18 win cases with slightly distinctive trademarks from 41 cases is 28.5%<p<55.5% (with 95% reliability). 91 Case C-251/95 Puma v Sabel ECLI:EU:C:1997:528, paras 24-25; Case C-342/97 Lloyd v Loints ECLI:EU:C:1999:323, para 28. 92 The confidence interval of selecting 25 cases with a reference to the CJEU rule from 189 cases is 9%<p<17.4% (with 95% reliability). 93 The confidence interval of selecting 21 win cases with a reference to the CJEU rule from 25 cases is 62.5%<p<91,8% (with 95% reliability). 94 Of the 25 cases in which the judge explicitly referred to this rule, this rule was also applied. In the four cases in which the claimant lost, the trademark was not (sufficiently) well known, or despite the notoriety of the trademark, likelihood of confusion could not be presumed because the defendant's sign was insufficiently similar to the trademark. judges did so, they actually followed the CJEU rule 95 without explicitly mentioning it. 96 The number of cases where empirical evidence was used was 24, i.e. 13% of the cases (see Table 7). 97 In most cases the claimant presented a consumer study to the court. The consumer studies were related to the market strength of the trademark as well as to likelihood-ofconfusion questions. In this sample, trademark owners more often won cases when they presented consumer studies to the court. 98 VI. Analysis

Risk of overbroad grants protection
As the results show, some patterns underpinning the studied cases indeed confirm that judges possibly favour trademark owners' investments in anti-confusion cases.
In the sample studied, trademark owners more often won than lost anti-confusion cases falling under sub b of the infringement provision at EU and Dutch level. Moreover, not all cases ended up in court because the overwhelming majority were settled amicably. 99 In  practice, wealthier litigants had more resources for sustaining a dispute, 100 so that the odds of winning might have been even higher had the settled disputes also been included. Furthermore, trademark owners won more cases when they invested in a particular sign. In a third of the cases, the trademark was considered to be highly distinctive, a factor that helped the trademark owner win significantly more often. Conversely, trademark owners lost (slightly) more often when the trademark was averagely or slightly distinctive. By comparison, trademark owners did not necessarily win more often when the goods and services were highly similar. 101 CJEU case law also illustrates the impact of the trademark's market strength on the confusion analysis. In some cases, courts explicitly referred to and applied the rule that consumers are more easily confused when confronted with a highly distinctive trademark. 102 If the trademark was indeed highly distinctive, this strengthened the trademark owner's position.
Furthermore, not only did a high degree of distinctiveness play a role as a separate factor in the global assessment, it also impacted the assessment of similarity between signs. The CJEU namely prescribed that the global appreciation of the visual, auditory or conceptual similarity of the marks in question must be based on the overall impression given by the marks, bearing in mind, in particular, their distinctive and dominant components. 103 Because of the notoriety of the trademark, the trademark naturally consists of distinctive and dominant elements. When judges referred to this rule, they often labelled signs as confusingly similar when the trademark was (highly) distinctive.
If judges also applied the Bigott-Batco/Doucal rule of the Dutch Supreme Court, 104 stating that the similarities between the trademark and the sign were more important than the differences, this further strengthened trademark's infringement claim. In almost all cases in which this rule was found, the trademark owner won.
In addition, in the majority of cases judges referred to the average consumer with a normal level of attention. 105 It may have been easier to conclude that such an average consumer was more likely to be confused when seeing signs similar to a highly distinctive trademark than a more experienced consumer with a higher degree of attention. Experts who are highly attentive are likely to notice even slight differences if the stakes are high.
Furthermore, survey evidence also strengthened the trademark claim. Trademark owners more often won  than lost when consumer studies were presented to the judge. Although the number of studies was limited, it helped trademark owners win those cases.

Counterbalances
The analysis also suggests that the situation is not alarming. In practice, in the majority of cases the assessment concerned a word mark. Figurative marks were also frequently assessed, but again, in the majority of cases this concerned a figurative mark containing word elements.
Purely figurative marks and shape marks were seldom assessed. 106 The same applied to descriptive 107 and cultural signs. 108 Granting broad protection to fanciful word and figurative marks also seems consistent. A strong and firm protection encourages trademark proprietors to maintain the high quality of their product, thereby protecting their reputation. Strong trademark rights ensure that consumers are protected against confusion and have lower search costs. In addition, the appropriation of fanciful signs and figurative marks, as already discussed above, does not as such grant trademark owners a competitive advantage over third parties. From this perspective, a broad protection might be justified.
In addition, in some cases, both parties had enough financial resources to afford the often expensive studies. For example, in the G-Star/H&M case, 109 both parties presented consumer studies to the court, such that there was no imbalance between the litigants. Nevertheless, we do not know how many cases were settled amicably. It may well be that parties with enough financial resources were more likely to go to court.
Furthermore, the mechanism of granting more protection the higher the trademark owner's investments did not necessarily apply. For example, in the Red Bull/ Leidseplein Beheer case 110 concerning energy drinks, the court did not apply the rule that well-known marks have a broader scope of protection. This also influenced the question of whether the signs were similar. The court argued that there was no similarity between the RED BULL trademark and 'The Bulldog' sign.
Firstly, the court argued that there were no major visual similarities between the signs. According to the judge, the RED BULL trademark also included a picture of a bull and the word elements 'Krating-Daeng' and 'Red' which were elements that could not be neglected even though they were less distinctive than 'bull'. 'The Bulldog' sign did not have these elements.
In the case of 'The Bulldog' sign, the court argued that 'dog' was as important as 'bull', while the element 'The' could not be ignored from a visual perspective. The RED BULL trademark did not have the elements 'dog' and 'the'. 111 In the court's opinion, the fact that the words 'Krating-Daeng' were displayed in a smaller font in the RED BULL trademark did not alter the fact that these words were spoken. Therefore, the court argued, the emphasis placed on the element 'bull' was less auditory than visual. The court argued that the auditory similarity was even less relevant than the already fragile visual similarity. 112 Furthermore, the court ruled that the word 'Bulldog' had a fixed and clear meaning, namely referring to a dog breed. A bull and a dog were, it said, different species of animals. In the court's opinion, the average consumer would not think of a bull when confronted with the word 'bul(l)dog' or a picture of a bulldog. This was also supported by market studies by the claimant. Therefore, the court said, the difference from a conceptual perspective was such that the minimal visual and auditory similarities between the signs were entirely suspended. Furthermore, the court argued, the considerable conceptual difference would ensure that the average consumer would not be confused by the visual and auditory similarities. 113 Although the goods were identical (energy drinks) and the degree of distinctiveness was high, consumers were not likely to be confused because the signs were not, or not sufficiently similar.
Interestingly, the court did not refer to the CJEU guideline stating that the more distinctive the earlier mark, the greater the likelihood of confusion. This would have favoured Red Bull because the trademark was very wellknown amongst average consumers. Remarkably, another CJEU guideline was also not mentioned, namely that account should be taken of the fact that the average consumer only rarely has the opportunity to make a direct comparison between different marks but must place her/his trust in the imperfect picture of them kept in her/ his mind. 114 If the court had applied these guidelines, it might have concluded that the 'Krating-Daeng' and 'The' elements were not noticed even though the 'Krating-Daeng' element was part of the trademark as registered. From an empirical perspective, the odds were even higher that this element would not be noticed by the average consumer. The fact that the court itself referred to the RED BULL trademark versus the 'Bulldog' sign in its reasoning may suggest that indeed such elements as 'Krating-Daeng' and 'The' can be neglected. The degree of similarity between signs would have been higher. Given the highly distinctive character of the trademark and the identity between goods, the judge could even have ruled that consumers were likely to be confused. That consumer studies pointed to a different conclusion could have been declared irrelevant in this respect by emphasizing, in the case of a highly distinctive mark, a likelihood of confusion had to be assessed normatively.
This case illustrates that judges did not always grant more protection when the trademark was very well known. When necessary, judges were able to block the automatism of granting more protection the more distinctive the trademark was. When judges found that the freedom of third parties was jeopardised, they could simply say that the signs were not similar.

Risk areas
This does not mean there is no risk of overbroad grants of protection. The study of Dutch case law indicates that trademark owners do sometimes succeed in achieving a broad scope of protection with regard to descriptive, cultural and non-traditional signs. 115

a) Non-traditional signs
For example, in the De Zuivelhoeve/Lidl case, 116 the judge found that the defendant Lidl with their product packaging for yoghurts had infringed the trademark rights of De Zuivelhoeve. The claimant's consumer study played a key role. The consumer study showed that 78.6% of respondents had used or seen claimant's transparent bucket with yoghurt and fruit. In answer to the question whether respondents knew what was normally mentioned on the transparent bucket, 66.7% of respondents (51.7% of all respondents) answered De Zuivelhoeve. 117 According to the judge, the shape mark consisting of yoghurt and fruit separated in different layers in a transparent bucket had acquired distinctiveness through use in trade and was to a certain extent well known. 118 The judge argued that Lidl's product packaging was highly similar to the shape mark of De Zuivelhoeve. The most striking elements of De Zuivelhoeve's shape mark were, also in view of the consumer study, the shape and image of the transparent bucket filled with yoghurt. These specific elements were also found on the Lidl packaging. Combined with the fact that Lidl used their product packaging for identical goods, i.e. yoghurt, the judge argued that a likelihood of confusion existed. 119 De Zuivelhoeve therefore not only succeeded in educating consumers in perceiving the shape of a transparent bucket with yoghurt and fruit in separate layers as a source identifier, they also succeeded in getting protection against a similar transparent bucket because they had taught their consumers so well. The court strongly based its decision on empirical findings. The prospect of an attractive return on investment was fulfilled. Competitors who also wanted to present their yoghurt in a transparent bucket had to come up with something else, even though the shape mark was, in actual fact, a mere property of the product concerned. Arguably, a transparent bucket as product packaging is functional and should not have been registered at all. Consumers could easily see the content of the bucket since it was transparent. Therefore, De Zuivelhoeve did in fact obtain an unfair competitive advantage.
Interestingly, in the Dyson case, the CJEU used similar reasons to reject Dyson's request for trademark registration on a transparent collecting bin for a vacuum cleaner. 120 Admittedly, the Dutch case scenario was slightly different from the Dyson case, where the application concerned not a particular type of a transparent collecting bin, but rather, in a general and abstract manner, all conceivable shapes of such a collecting bin. 121 De Zuivelhoeve's trademark, on the other hand, concerned a particular shape 122 of transparent bucket with yoghurt and fruit in separate layers depicted with their word mark and several figurative elements. Nevertheless, although De Zuivelhoeve did not register a transparent bucket in a general and abstract manner, in fact, the registration of a specific transparent bucket consisting of yoghurt and fruit in separate layers with a word mark and print on the cover led to a similar result -a monopoly on a mere property of the product concerned.
If the judge had ignored the fact that both buckets were transparent, it remains to be seen whether he would also have ruled that this was a case of infringement. The yoghurt layers were different as was the size of the buckets. Furthermore, Lidl's bucket had no print on the side. Another important difference was that the buckets had different word marks and different prints on the cover. 123 If the judge had considered these elements in isolation, he would probably not have concluded that this was an infringement. Therefore, although De Zuivelhoeve did not register the transparent bucket as such, in the infringement proceedings we see that registering a combination of elements as a trademark still allows a trademark owner to monopolise a single element.
In other words, it can be risky to follow a generous approach towards registering non-traditional signs. Once a trademark has been registered, trademark owners can invest substantially in marketing and branding their trademark and then claim a broad scope of protection. Since the CJEU rejected the need to keep signs freely available as a factor in the confusion analysis, 124 judges are not allowed to make normative corrections in favour of thirdparty interests. If consumers perceive the transparent bucket as a source identifier, the judge must take this into account. Other CJEU rulings further encourage the judge to grant a broad scope of protection. According to the CJEU, judges must focus on the distinctive elements and not so much on the less distinctive elements. Since consumers were specifically educated to perceive the transparent bucket as a source identifier, this was the most distinctive part. From this perspective, the other elements, such as different sizes, word marks and prints were less distinctive and therefore less important. The Bigott-Batco/Doucal ruling of the Dutch Supreme Court, stating that similarities are more important than differences, further contributed to the judge focusing on the fact that both buckets were transparent. 125 Furthermore, colours, patterns or shapes can also be claimed via a figurative mark. An example is the IKEA/ Serboucom case. 126 Serboucom exploited a chain of hardware stores where construction materials, tools and decoration articles could be purchased. IKEA had a figurative mark with blue and yellow elements and a similar mark but with 'IKEA' in blue letters. The court of appeal ruled that given the strong reputation, both of the figurative elements as words were equally distinctive.
Serboucom used a similar logo but with the words 'multimate' and '(service)bouwmarkt'. According to the judge, these words had little distinctive character. The consumer would interpret the signs 'multimate' as referring to an assortment of several devices and thus not as a badge of origin. Therefore, the average consumer would perceive the blue and yellow elements of Serboucom's logo as most distinctive. Because of the similar colours, Serboucom's sign was similar to the IKEA trademark. This outcome was not changed by the fact that the colours were not exactly the same (the colours had a different PMS number). 127 The judge used empirical findings, such as the high degree of distinctiveness of the IKEA trademark on the one hand, and the lack of distinctiveness of the defendant's sign on the other hand to argue that this was a case of infringement. It is questionable whether real consumers were confused when seeing the 'Multimate' logo. Because of the strong reputation of the IKEA logo real consumers would know that 'Multimate' was not connected with IKEA, but with a third party. Although IKEA did not register a colour mark, they could still prevent Serboucom from using similar colours because of their trademark rights on the figurative mark with blue and yellow elements.
Again, this case illustrates how trademark owners can avoid normative restrictions regarding the need to keep signs free during the registrations process. Via the backdoor of a figurative mark, trademark owners acquired protection on particular elements that would otherwise have been rejected had trademark owners asked for a single registration of an abstract colour mark or a shape. 128 Similarly, in the TUC/Apéro case, 129 the trademark owners only had a trademark right on a logo for salted crackers. According to the court, the defendant's sign, which was also used for salted crackers, was confusingly similar to the trademark owner's logo because of the similarity of the colours on the packaging. Given the high investments in marketing campaigns, market share, brand recognition and the long-term use of the trademark, the court argued that it was more likely that consumers were confused, even though the words were completely different. 130 The court also stated that the product concerned was a supermarket product that consumers would purchase without thorough evaluation. Therefore, the consumer would more easily be influenced by the visual impact of the trademark they were looking for. 131 In addition, the court emphasised that the similarities were more important than the differences and that the differences (such as deviation in colour, letter type, layout, and word marks) were of minor importance. 132 The district court further rejected the argument that the trademark owner was unduly monopolising colours with regard to salted crackers. The colours used were not common for salted crackers. 133 The judge also attached some value to the trademark owner's consumer study. Although the trademark owner TUC seemed to have had a market leadership position, such that this actually should have been filtered out by a control group, the judge nevertheless found the results gave some indication that consumers were likely to be confused. 134 Trademark owners have not always succeed in claiming colours via a figurative mark. For example, in the Reckitt/Action case, 135 the court ruled that the VANISH In recent years, the CJEU has taken a more cautious approach to protecting non-traditional marks. It could be argued that this case originated from a more generous period and the trademark might not have been registered if the current CJEU rulings were applied. However, because the trademark consists of a combination of several elements the refusal grounds probably still do not apply. The sign does not consist exclusively of a shape, or another characteristic which results from the nature of the goods themselves, or which is necessary to obtain a technical result or which gives substantial value to the goods. For example, the transparent bucket also has essential decorative elements that are not inherent to the generic function of the goods. Furthermore, it is questionable whether consumers would specifically buy this transparent bucket just because of the substantial value of the shape.  trademark for detergents was very distinctive and well known among the relevant public because of massive marketing campaigns.
However, the court also argued that the trademark owner Reckitt could not claim any rights on the colour pink for detergents. This particular colour was only one element of their figurative marks. According to the judge, colours are normally not more than a characteristic of the goods and are used on a large scale because of their attractiveness. In the court's opinion, people are generally aware of the fact that detergents are presented in multicoloured packaging. Consequently, the average consumer would not perceive the colour pink on the label as a badge of origin, in spite of the fact that they had been consistently presented with the appearance of the product to which the VANISH trademark was attached.
Given this fact, the judge argued that only a minimal degree of distinctiveness could be assigned to the colour pink. Accordingly, the fact that the colour pink dominated the trademarks could not have a significant impact on the overall impression. According to the judge, the average consumer who was used to the fact that detergents were presented in multi-coloured packaging would not pay special attention to the colour pink and therefore would not remember this particular colour as a badge of origin. 136 It is, however, doubtful whether these assumptions are true. Consider, for instance, the widespread recognition of the lilac colour of Milka and the colour magenta of Deutsche Telekom; these examples show us that consumers are able to recognize an abstract colour as an identifier of commercial source after intensive use in trade. 137 There are more reasons to believe that consumers do not by definition experience more difficulties in perceiving colours. For example, colours are easy to perceive, and they are clearly visible from a distance, in contrast to words. Colours are a good tool for remembering objects. 138 Furthermore, with regard to services, consumers quickly recognize colours as an identifier of commercial source because by nature services do not have colour. 139 In addition, colours can better function as identifier of commercial source than word marks if the trademark is addressed to a public that has difficulties with reading or that cannot read at all. 140 However, the court probably used these assumptions, i.e. consumers will not recognize the colour of the label as an identifier of commercial source, to reach a normative purpose: colours should remain available to other users on the market. If the court had empirically assessed the perception of the consumer, the trademark owner Reckitt would have expanded their figurative mark rights to a right on an abstract colour (pink) used for detergents. Because of the attractiveness of a colour and the fact that the number of colours is limited, an expansion must be justified. For example, the threshold to achieve a trademark right on an abstract colour is much higher simply because of these normative considerations. The trademark owner must prove that the sign has acquired distinctiveness through use in trade if it wants to register an abstract colour as a mark. Conversely, a figurative mark does not have to pass this high threshold. 141 This case therefore illustrates that judges sometimes did make normative corrections when the filtering mechanisms of the registration process did not function properly. In light of the CJEU case law, 142 judges could argue that the colour pink was not particularly dominant and could be ignored. High investments in the VANISH brand were not enough to overrule the normative objections towards keeping colours freely available for competitors. 143 However, the fact remains that judges do have a range of rules at their disposal that favour trademark owners' investments. Even though the number of cases involving non-traditional signs was limited and judges did not always take a generous approach towards granting trademark protection, it was possible to circumvent normative restricting regarding the need to keep signs free during the registration process via the registration of a figurative mark.

b) Descriptive and cultural signs
Similar problems could arise with regard to descriptive signs. Although there were only a few cases in which the court ruled that a trademark was not distinctive, this did not mean that a trademark could not consist of descriptive elements.
For example, in the IT-Staffing/Staffing it case, 144 the court found that the third party's sign 'Staffing it' had infringed the trademark rights on the figurative mark 'IT-Staffing'. Both parties used their signs for employment services. Considering the many similarities, the court said, the small differences (such as another colour and font, a hyphen and several tiny figures surrounding the letter 'I') failed to give weight to the similar overall impression. 145 In addition, the court argued that the trademark had a highly distinctive character which was supported by several consumer studies, so that it was more likely that consumers were confused. 146 The fact that the public consisted of professionals specialised in the IT field did not alter this conclusion. In addition, the relevant public not only consisted of companies who outsourced IT work to self-employed parties but also of self-employed parties operating in the IT field. These self-employed parties could not be expected to be more cautious because they did not generally have a particular interest in knowing who was responsible for assigning the contract. 147 Despite the fact that IT-Staffing was quite descriptive, the trademark owner succeeded in getting trademark protection. Empirical findings concerning the highly distinctive character of the trademark mainly determined the outcome of the case. Again, a figurative mark (containing word elements) was used to register elements that were not particularly distinctive and could have been refused if only the single words were registered.
In addition, the 't IJs van Columbus case 148 illustrates that a cultural meaning does not necessarily overrule the mechanism that grants more protection if a trademark has a highly distinctive character. Instead of limiting scope of protection, the judge argued that the name 'Columbus' contributed to the trademark's distinctive character because the name would be more easily brought to mind owing to the associations the cultural name evoked. The defendant's sign also contained the name 'Columbus'. Given the dominant position of the name 'Columbus' in the figurative mark, the sign could therefore be seen as similar. 149 Moreover, trademark owners did not even always have to invest heavily in their trademark to receive protection. 150 For example, in the LIEF!/LIEFDIER case, 151 the court found that the defendant IJsvogel c.s. infringed the trademark rights of KHF c.s. on LIEF!, which was registered for pet accessories by using the sign LIEFDIER also for the same goods and services. Firstly, the court argued that the LIEF! trademark had a distinctive character.
The fact that (the adjective) LIEF (English: 'sweet') referred to the standard meaning of 'sweet', the court said, did not mean that the trademark was descriptive of the registered goods, namely dog and cat pillows. 152 Secondly, the court argued that there was a likelihood of confusion. 153 According to the judge, the most obvious similarity between the trademark and the sign was the identical element LIEF. According to the judge, because LIEF in LIEF DIER was placed in front, and DIER (English: animal) must be categorised as descriptive for pet accessories, LIEF was a dominant element in LIEF DIER, or at least, it was an element with an independent place in the word combination. In the court's opinion, this suggested a high degree of auditory and visual similarity between the trademark and the sign. LIEF in LIEF DIER also had a specific meaning, the court said, because it usually referred to (a characteristic of) the animal for which the accessory was intended. According to the judge, the trademark LIEF! could be interpreted in a similar way. Therefore, the court argued, there was also a conceptual similarity between the trademark and the sign.
The registration of a quite descriptive term could therefore also have serious consequences for competitors on the market. Because of its descriptive character, the trademark indirectly informed consumers about a characteristic of the animal for which the accessory was intended. This could give trademark owners KFH c.s. an unfair advantage on competitors on the market. Defendant IJsvogel c.s. was no longer allowed to use a similar term for their product. Furthermore, the specific meaning that the trademark LIEF! evoked by its descriptiveness was used against the defendant when assessing likelihood of confusion. Because the trademark LIEF! usually referred to a characteristic of the animal for which the accessory was intended -pets are usually associated with such terms as sweet, soft and fluffy -there was also a conceptual similarity.
While the court found the link between pet accessories and sweet to be insufficiently direct to deprive the trademark owner of his rights, this link was, however, direct enough to conclude that the trademark and sign were conceptually similar. Instead of limiting the scope of protection by emphasising the differences between the trademark and the sign, such as the interspace between LIEF and DIER, the different look and feel, and the exclamation mark, the judge looked at the similarities instead of the differences and labelled these as dominant. Because judges need to consider the overall impression created by 152 According to the judge, the fact that a dog basket can be described as 'sweet' if it is soft and fluffy does not provide a sufficiently direct and concrete link. 'Sweet' is not a general characteristic of dog and cat pillows, but perhaps only of those that are soft and fluffy. The court argued that soft and fluffy was also not a direct meaning of sweet and it was not recorded under that meaning in the dictionary. The judge ruled that the Caffè Nero case (Case T-29/16 Caffè Nero ECLI:EU:T:2016:635) did not apply in this case. The sign 'caffè nero' was found to be descriptive of coffee-related products and services. In the court's opinion, the link in this case was, however, more direct and concrete than the link between pet accessories and 'sweet'. On this basis, the trademark was found distinctive and not descriptive (paras 4.6-4.7). 153 Although the sign LIEFDIER was used as a single word, the judge ruled that the average consumer would perceive the sign as LIEF DIER (with two words). It argued that this could be derived from the logo and from a Facebook post in which the defendants had written about the 'LIEF DIER collection' (para 4.14). marks, slightly distinctive trademarks could be considered dominant and distinctive if the defendant's sign did not contain particularly distinctive elements. 154

Summation
As shown above, in practice, the automatism of granting more protection in response to heavier investments in marketing and branding did not often appear when dealing with descriptive, cultural and non-traditional signs. From this perspective, the situation was not alarming.
Nevertheless, in some cases, trademark owners still succeeded in achieving broad protection for descriptive, cultural and non-traditional signs. In particular, figurative marks formed a risk. When trademark owners invested heavily in their logo, the elements concerned, such as particular colours, automatically became very well known to consumers. When a defendant used similar colours in their logo, this could lead to infringement even if the defendant added a different word mark. Since judges are not allowed to make normative corrections towards third-party interests and trademark owners can avoid normative restricting during the registrations process, trademark owners with deep pockets may have a free hand in achieving broad trademark protection with regard to descriptive, cultural and non-traditional signs. 155 It is true that in some cases courts blocked this automatism of granting more protection to highly distinctive trademarks. Courts simply argued that the signs were not similar or emphasised the fact that consumers were not used to perceiving colours as a badge of origin. Nevertheless, there were also cases in which courts did not deviate from the pattern of granting more protection to highly distinctive signs. Trademark owners sometimes did not even have to invest heavily in their trademark to achieve broad protection. In some cases, the CJEU rulings were enough to conclude that the signs were similar, for example because both signs started with the same and therefore eye-catching elements.
VII. An exemplary study of UK case law To check whether similar conclusions could be drawn for other EU Member States, an exemplary study was done of UK case law. A small sample of UK cases 156 seems to confirm these findings. In the UK cases, claimants also won more often than they lost. 157 Similarity between signs played a crucial role in the likelihood-of-confusion assessment. This factor was discussed in all cases, in contrast to factors of similarity of goods and services and degree of distinctive character. In most cases the consumer had an average degree of attention, and only in a few cases the UK courts deviated from the benchmark of the average consumer. 158 Consumer surveys also played a modest role in the UK cases. 159 Instead, witness statements were frequently used as evidence of actual confusion.
Furthermore, in the majority of cases, the degree of distinctive character was average or high. A high degree of distinctive character was in some cases even explicitly rewarded with a winning claim. 160 Furthermore, the rule stating the more distinctive the earlier mark is, the greater will be the likelihood of confusion, was well known among judges and regularly found in the beginning of the decision, where judges outlined the legal framework. In other words, in UK case law, infringement factors also seemed to favour trademark owners' investments.
Nevertheless, in practice, in UK case law, the risk of misappropriating descriptive, cultural and nontraditional signs also seems to be limited. The majority of cases concerned word marks and figurative marks containing word elements. 161 In one case, involving the shape of a London taxi, the trademark owner even lost. The court found that the trademark was invalid. Consumers would have difficulties recognising shapes as a source of origin and because of the Regulation for London taxis, consumers would think that there was only one kind of London taxi. 162 However, in some UK cases, trademark owners succeeded, via a figurative mark. For example, in the Enterprise/Europcar case, the trademark owner's high 154 A more restricted approach would involve looking at the various elements in isolation. Courts could deny any similarity between signs if the elements coinciding in two conflicting signs were not protectable as such. Although the infringement claim could also have been successful under this approach since the LIEF trademark was found to be distinctive, in some other cases it would have led to the claim being rejected. For example, in the Connect/Connect Professionals case involving employment agencies, the trademark was slightly distinctive. If the judge had assessed the elements 'Connect' and 'Professionals' in isolation, the judge might have argued that the trademark CONNECT was slightly distinctive and therefore the slightest difference between the marks (like 'Professionals') could have been enough to reject the infringement claim. District Court of Gelderland, 4 April 2017, ECLI:NL:RBGEL:2017:2433 -Connect/ Connect Professionals. See also Kur (n 23) 99.
155 Even under the new limitations, the use of the sign would probably still not be justified since it would have to be in accordance with honest commercial practices. Given the notoriety of the colours, i.e. as in the IKEA/Serboucom case, using similar colours would probably be seen as unfairly competing with the trademark owner. Because of the notoriety of the colours, consumers will automatically think of IKEA when confronted with similar colours; irrespective of the question whether the defendant used their own trademark or not. The use of similar colours can also not be seen as intended to give an indication concerning one of the specified characteristics of the goods. See Case C-63/97 BMW v Deenik investments in marketing and branding campaigns were awarded. Both parties used the letter 'e' in their logos. The judge considered that the letter 'e' was an extremely commonplace letter, and it was widely used in both descriptive and denominative contexts. 163 However, the judge also argued that the letter 'e' was not widely used in logos to denote providers of vehicle rental services in the UK, thus the logo had inherent distinctive character. Furthermore, the court argued that the 'e' logo had an enhanced distinctive character as a result of its use for vehicle rental services. 164 This high distinctive character and the fact that both parties used their logos for identical services supported a likelihood of confusion even though a low similarity between the logos was found and in fact granted the trademark owner a monopoly on a single letter with regard to vehicle rental services. 165 Finally, in the Thomas Pink/Victoria's Secret case, the claimant had registered the word 'PINK' with pink lines making up the letters for clothing and other accessories. 166 The defendant, Victoria's Secret, launched a subbrand called 'PINK', also used for clothing. The court found that the trademark had acquired a distinctive character through use, but argued that pink was a colour, 167 so it did not have an enhanced, but only a normal, level of distinctive character. 168 Nevertheless, the use of PINK emblazoned on the defendant's items of clothing still created a likelihood of confusion among consumers since the goods were identical and the signs were very similar. The same applied to the use of PINK in slogan and device format. 169 Although the claimant did not use a colour mark, he still managed to (partly) claim the colour pink via the logo.
As these UK cases show, it can pay off to invest heavily in marketing and branding campaigns. In the majority of cases, trademark owners won the case while the degree of distinctiveness was often average or high. Nevertheless, the risk of misappropriating descriptive, cultural and non-traditional signs seems to be limited since in most cases, this concerned a word mark or a figurative mark. However, the UK cases also illustrate that this type of mark may be a risk for the trademark system. Via the registration of a figurative mark, trademark owners received protection for the letter 'e' for vehicle rental services and the colour pink for clothing and other accessories. Without these investments and additional consumer studies it is doubtful whether the claimants would have won.

VIII. Conclusion
This article considers whether there is a risk of trademark law being strategically used as a self-servicing mechanism by the industry. In this respect, two features of the trademark system are particularly worrisome. First, trademark owners can strengthen the distinctive character of their marks by investing in marketing and branding campaigns. Second, trademark owners can afford expensive surveys demonstrating the high distinctiveness of their marks thus strengthening their position in an infringement case.
As shown above, the EU trademark system indeed encourages and rewards trademark owners' investments in several ways. An empirical study of Dutch case law demonstrates that trademark owners more often win than lose. Furthermore, trademark owners more often win the more they invest in their trademark. In addition, when a court applies the CJEU rule stating that consumers are more easily confused when confronted with a highly distinctive trademark, trademark owners also win most cases. Furthermore, judges are more likely to label signs as similar if a trademark is highly distinctive. Moreover, consumer surveys further strengthen trademark claims.
The analysis also illustrates, however, that cases involving attempts to monopolise descriptive, cultural and non-traditional signs (where trademark protection may endanger freedom of expression and freedom of competition) are not widespread. Granting broad protection to fantasy marks and other fanciful forms of marks does not seem inconsistent in light of the rationales underlying trademark law. In addition, in some cases both parties had enough financial resources to afford the often expensive studies. Besides, judges did not always give more protection when the trademark was very well known. When necessary, judges simply stated that the signs were not similar.
Nevertheless, this does not mean that there is no risk at all of granting overbroad protection against confusion when it comes to descriptive, cultural or non-traditional signs. The study of Dutch case law indicates that trademark owners indeed sometimes succeed in achieving a broad scope of protection with regard to descriptive, cultural and non-traditional signs. In particular, figurative marks may form a risk. Since they allow trademark owners to circumvent normative restrictions when registering a figurative mark, and judges, in line with the CJEU decision in Adidas/Marca, 170 are not allowed to make normative corrections towards third-party interests on the basis of a need to keep signs freely available on the market, trademark owners with deep pockets can go very far in infringement cases. Consequently, this may lead to an unbalanced market situation with, on the one hand, trademark owners enjoying broad trademark rights and competitive advantages from the sign as such, and on the other hand, third parties being excluded from using these appealing signs and being forced to use less effective trademarks because these signs are not available in sufficient supply.
As long as the CJEU persists in refusing to consider the need to keep signs freely available to competitors on the market as a relevant factor in the likelihood-of-confusion assessment, trademark owners will invest heavily -and sometimes be rewarded -with a broad scope of protection. If we want to further reduce the risk of descriptive, cultural and non-traditional signs no longer being freely available to competitors on the market, we must recalibrate the confusion analysis and allow judges to make normative corrections in favour of third parties' freedom of expression and freedom of competition. In this way, judges will be better equipped to weigh trademark owners' interests against the interests of competitors and the general consuming public. Accordingly, judges should be able to permit unauthorised uses of trademarks even when this might lead to likelihood of confusion among consumers. Consumers would not be confused about the commercial origin of goods or services if enterprises were not able to invest in these non-distinctive signs in the first place. 171 Therefore, the fact that consumers are likely to be confused should not prevent judges from refraining from a cause of action that may lead to an imbalanced anti-confusion assessment and generate dysfunctional incentives.