A Comparative Approach to Addressing Language Discrepancies in Multilingual Treaties

Introduction

In 1889, Ethiopia and Italy both signed the Treaty of Wuchale.1 In just six years, the consequences of the treaty would lead to war between the two nations.2 The reason for this calamitous result was mistranslation.3 The Treaty, drafted in both Amharic and Italian, differed by one verb in the section related to Ethiopia’s status as an Italian protectorate.4 Whereas the version written in Amharic gave Ethiopia the option to become a protectorate, the Italian version required that Ethiopia become one.5

While such a result is extraordinary, discrepancies between the language versions of multilingual treaties are not uncommon nor are issues related to legal translation isolated to treaties. The laws of several nations and international organizations are also multilingual.6 In Canada, laws are published in both French and English.7 The European Union goes further, publishing laws in all twenty‑-four official languages.8 There are also similarities between how the ICJ and those other jurisdictions’ courts are expected to read multilingual texts. Both multilingual laws and treaties have language versions that are 1) equally authoritative and 2) equally authentic.9 In other words, no versions of multilingual texts have primacy; each are authoritative embodiments of the law.10

Yet this leads to problems when equally authoritative versions of the text appear to contradict. These contradictions result in a spectrum of different solutions to the problem of multilingual lawmaking and interpretation. Not all these solutions are created equal. While the ICJ has attempted to answer the challenge presented by multilingualism it ultimately fails to respect the seriousness of the equal authority of language texts. While the interpretative frameworks used in Canada and the EU are not perfect, they offer a path forward for the ICJ.

To this end, this Comment will start by examining the historical context that created the multilingual systems in Canada and the EU. In Part II, it will analyze two reasons why multilingual systems continue to present challenges for courts when interpreting laws. The first challenge involves de facto unilingualism. In both the EU and Canada, unilingualism is the norm rather than the exception. Because of this, language versions are often not read in conjunction with each other. Resolving this requires recognizing that multilingualism not only improves access to language groups but also imposes necessary obligations on litigants and court officials. The next challenge involves interpreting translation discrepancies. Both Canadian courts and the CJEU have adopted different methods of statutory interpretation when it comes to multilingual laws that respect the equal authority of language texts.

Part III turns its focus to multilingual treaties and their interpretation at the ICJ. It begins by looking at the history of treatymaking. Then it argues that the challenges facing national courts are like problems the ICJ faces. These challenges are de facto bilingualism and translation discrepancies. For each of these challenges, this Comment argues that solutions relied upon by Canadian and EU courts provide useful frameworks for solving issues affecting the ICJ’s interpretation of multilingual treaties.

The Roots of Multilingual Legislation

Bilingualism in Canada can be traced back to the nation’s founding in 1867.11 When the Dominion of Canada was formed, the Canadian Charter of Rights and Freedoms established that English and French would both be the official languages of Canada.12 Further, the Supreme Court of Canada, as early as 1891, affirmed the equal authority of the French and English text of laws.13 In affirming the equal authority of both the French and English text sof a law, it established that laws would be made and interpreted using the equal authenticity rule and the shared meaning rule.14 Under the equal authenticity rule, both the French and English versions of the law are equally valid and authoritative.15 The shared meaning rule holds that where one version of a law is ambiguous and the other is clear, the meaning of the clear version is shared with the ambiguous version.16

The EU is unique among large international bodies in that its legislation is written in the languages of all its member nations.17 The reason for this stems from the scope of “EU law.”18 Whereas other international organizations may draft treaties for the governments of member nations, EU law directly regulates the lives of EU citizens.19 This effort to legitimize EU law through access to diverse linguistic communities is similar to Canada’s own efforts. These similarities in purpose may be rooted in the origins of both governments. The Dominion of Canada, like the EU, was an effort to federate and unite disparate political communities that did not share a common language. 20 Further, the EU’s multilingual regime also extends to multiple branches of government, including the European Parliament, European Commission, and the Court of Justice of the European Union [hereinafter CJEU]. In the case of EU legislation, these documents are published in twenty-four languages.21 Like Canada, the EU also has its own equal authenticity and shared meaning rule.22

The Challenges of Multilingual Legislation

De Facto Unilingualism

The languages spoken in Canada vary wildly based on what province one looks at. Although only seventy-five percent of Canadians use English as their first language, in all but two provinces, the rate jumps to ninety-five percent or higher.23 Further, only in Quebec are anglophones in the minority, with eighty-two percent of the population using French as a first language.24 This provincial disparity is similarly reflected in rates of bilingualism.25 As of 2016, 17.9% of the population is bilingual in French and English.26 Yet there are only two provinces, Quebec and New Brunswick, with a bilingual population of over 17.9%.27

In effect, throughout much of Canada, the population is largely unilingual. This has created a system of “legal dualism” in which francophones read only the French text of a law and anglophones read only the English.28 This problem of legal dualism impacts the courts as well.29 In R. v. Mac, the Ontario Court of Appeals interpreted the meaning of the word “adapted” in the Criminal Code.30 However, the Court only looked at the English text. From this, the Court ruled that the law was ambiguous.31 Yet, the Supreme Court of Canada held that while the English version was ambiguous, the French version was clear.32 For this reason, the Supreme Court of Canada ruled the Court of Appeals should have relied on the French text given the shared meaning rule.33 Yet this is not to say de facto unilingualism cannot be, at least partially, resolved. Following Mac, the Supreme Court of Canada changed its rules to require that parties’ filings must use the English and French texts of a law.34 At the same time, there has been increased focus in Canada on appointing justices who speak both French and English.35

This issue of de facto unilingualism is even more extreme within the EU though. Whereas Canada has two official languages, the EU has twenty-four. And whereas Canada may be able to require that court parties consider both English and French versions of a law, it is only sophisticated individuals with access to a staff of translators who could parse through the twenty-four versions of EU laws. According to Article 22 of the Rules of Procedure of the CJEU: “[t]he Court shall set up a translating service staffed by experts with adequate legal training and a thorough knowledge of several official languages of the court.”36

De facto unilingualism presents a challenge for individuals hoping to understand the full meaning of a law. As was the case in Mac, a unilingual English speaker would read the section at issue in the Criminal Code and find that it is ambiguous even though the Supreme Court of Canada expressly ruled that the section, given the French text, was clear.37 Yet it is not the case that discrepancies in translation will cause conflicts in meaning for most applications of a law. Further, just as the Supreme Court of Canada has tried to resolve this issue by requiring that parties’ filings must give provisions in both of their official texts, so too has the CJEU determined in several of its cases that interpretation of statutes requires an inquiry into all the different language versions of a law.38 This issue should not be too minimized though. At least in the Canadian context, many of the notable cases involving translation relate to the Criminal Code.39 In these cases, notice is especially critical and yet for unilingual citizens, such notice may be impossible.

Discrepancies in Legislation

The second challenge faced by multilingual systems comes from discrepancies in translation. This challenge stems from the equal authenticity rule and the shared meaning rule. Because there are multiple versions of a text that each have equal authority, it can be a challenge for courts to resolve discrepancies across texts. Yet given the frequency of this challenge, this is something that courts in all multilingual jurisdictions have had to address. Therefore, both Canadian courts and the CJEU have developed strategies to deal with this challenge.

First, Canadian courts accept that the rules of equal authenticity and shared meaning must be applied before any other rules of statutory interpretation.40 The courts have also developed a way of resolving these rules. First, they determine whether there is a conflict between the French and English texts of a law.41 Second, the courts determine whether there is an “absolute conflict” regarding meaning, a conflict because one version is clear and the other ambiguous, or a conflict because one version is broad and the other is narrow.42 If one version is clear and the other ambiguous, the inquiry is over and the court adopts the meaning of the clear text.43 If one version is narrow and the other broad, typically the court will adopt the narrower version.44 Finally, the third step is to determine whether the common meaning, if it can be found, conforms to the purpose and legislative intent of the statute.45 If no common meaning exists and there is instead an absolute conflict between the language versions, then standards rules of interpretation will be used to determine a shared meaning.46

Daoust provides a useful framework for understanding this process. In Daoust the court examined a section of the Criminal Code which criminalized laundering proceeds of a crime.47 The English text gave eight enumerated offenses as well as a catchall provision.48 The French text also listed eight enumerated offenses but lacked any catchall.49 In this case the Supreme Court of Canada ruled that a conflict resulted from the French text being narrower than the English text.50 Because of this, the Court proceeded to hold that the French text contained the meaning shared by both versions.51 Interestingly, although the Court then determined that the legislative intent was to criminalize all attempts at laundering proceeds of a crime, the Court still held that the catchall provision could not be applied.52

This challenge of discrepancies is equally present in the EU as it also has an equal authenticity rule.53 Between 1960 and 2010, the CJEU has asked whether there are discrepancies across texts in 246 judgments.54 In thirty-one percent of those judgments, no discrepancy was found, in twenty-five percent of judgments, no significant judgment was found, and in forty-four percent of those judgments, a discrepancy “posing interpretation problems” was found.55

Yet the method the CJEU has used to resolve these disputes has not always been consistent. In Régina v. Bouchereau the Court first argued that when texts diverge, the law must be interpreted in such a way as to respect the “purpose and general scheme of the rules.”56 According to Baaij’s analysis, between 1960 and 2010, the Court has referenced a purposivist approach in fifty of its judgments.57 Despite what the CJEU may say in some of its opinions,58 the Court has used a purposivist approach to resolve less than half of the cases that involve significant discrepancies between texts.59

Instead, Baaij argues that the most relied upon approach to discrepancies in translations involves either a “majority argument” or a “clarity argument.”60 The majority argument holds that the meaning of the law is whichever meaning can be found in a majority of the law’s texts.61 The clarity argument holds that the meaning of the law is found in whichever text is less ambiguous.62 Yet, in all but two cases in which the Court used the clarity argument, the Court found that the clearer text was the majority text.63 How much daylight really exists between a purposivist approach, and the majority argument can be questioned though. This is because the majority argument can be said to further the broader purposes of the equal authenticity and shared meaning rules. As the CJEU argues in Codan, “to disregard the clear wording of the great majority of the language versions . . . run[s] counter to the requirement that the Directive be interpreted uniformly.”64 Nevertheless, just as the Canadian courts generally ask whether any common meaning is consistent with legislative intent,65 the CJEU may use a purposivist approach to argue against a meaning consistent with a clear text.66 In fact, Baaij found seventeen judgments in which the CJEU relied on a purposivist approach to go against a meaning supported by a majority of the texts.67

The International Court of Justice & Multilingual Treaties

From the Roman Empire to the Eighteenth Century, treaties in Europe were drafted in Latin.68 This allowed for ease of communication in diplomacy and clarity over the meaning of treaties. Even after Latin ceased to be the dominate language of diplomacy, unilingual treaties continued well into the next two centuries as French came to replace it as the lingua franca.69 However, since the Twentieth Century, multilingual treaties have become more common place.70 Post-World War I, the interpretation of these multilingual treaties has been governed by first the Permanent Court of International Justice and later the International Court of Justice.71 The international community has also helped develop principles of interpretation, incorporating the issue of multilingual interpretation into the Vienna Convention on the Law of Treaties.72 Article 33 of the Convention specifically says:

Article 33

Interpretation of treaties authenticated in two or more languages.

1. When a treaty has been authenticated in two or more languages, the text is equally authoritative in each language, unless the treaty provides or the parties agree that, in case of divergence, a particular text shall prevail.

2. A version of the treaty in a language other than one of those in which the text was authenticated shall be considered an authentic text only if the treaty so provides or the parties so agree.

3. The terms of the treaty are presumed to have the same meaning in each authentic text.

4. Except when a particular text prevails in accordance with paragraph 1, when a comparison of the authentic texts discloses a difference of meaning which the application of articles 31 and 32 does not remove, the meaning which best reconciles the texts, having regard to the object and purpose of the treaty, shall be adopted.73

Paragraph 1 presumes an equal authenticity rule unless parties stipulate otherwise, and Paragraph 3 creates a shared meaning rule. Finally, Paragraph 4 provides a general framework for understanding discrepancies between plurilingual multilingual translations. This framework will also be discussed in the coming sections.

​​​​​​​De Facto Bilingualism

Although it may not appear to be relevant to the interpretation of treaties at the ICJ, issues arising out of de facto unilingualism are nevertheless present, to some extent, at the Court. Although the ICJ does not have issues related to unilingualism, it often expresses a kind of “de facto bilingualism” that favors English and French texts at the expense of other equally authentic languages versions of treaties. For example, at the time of the ICJ Advisory Opinion on Conditions of Admission of a State to Membership in the United Nations, the U.N. Charter had five authentic versions in Spanish, Chinese, French, English, and Russian.74 However, when the ICJ analyzed the U.N. Charter, a majority of judges only considered the French and English texts.75

One attempt at rationalizing this behavior is to argue that the court is motivated by 1) a general principle of efficiency and 2) the shared meaning rule. If treaties have a shared meaning rule, then it may not be necessary to examine every text of a treaty absent some conflict between texts. This was a view expressed by one member of the International Law Commission while the Vienna Convention was drafted.76 Yet, as Canadian Courts have shown, a shared meaning rule should not always lead to a presumption of shared meaning. Otherwise, conflicts between texts will go unnoticed.77

Further, it is possible for the Court to interpret languages other than English or French without sacrificing efficiency. Some contemporary cases refer to other languages of a treaty. In The Gambia v. Myanmar, Preliminary Objections, the Court considers arguments made by the parties that cover versions of the Genocide Convention in English, French and Spanish.78 Yet even as some contemporary cases reference other language versions, many do not. Croatia v. Serbia involves claims made by Croatia under Article II of the Genocide Convention.79 One set of claims fell under Article II(a) which refers to the killing of members of a group.80 The Court held that the French and English versions had the same meaning.81 In looking at only the those two texts though, the Court ignored the three other authentic texts of the Genocide Convention.82 Further, the cases cited by Croatia v. Serbia for the meaning of subparagraph (a) also only reference the French and English texts.83

This reliance on only English and French is even more surprising given the fact that one of those cited cases, Blagojević & Jokić, notes that the English term used in Article II(a), “killing,” and the French term, “meutre,” are in conflict with each other.84 Whereas killing “includes both intentional and non-intentional homicides,” “meutre” refers to intentional homicides.85 Given this possible conflict in meaning, Russian, Chinese and Spanish texts could potentially resolve the ambiguity.

But there are several reasons why in the case of Croatia v. Serbia such an inquiry may not have altered the ICJ’s ruling in this case. First, there is existing court precedent based on only the English and French texts of Article II defining subparagraph (a).86 Further, at least the Spanish translation of Article II does not seem to clarify the meaning of subparagraph (a). The Spanish version refers to “matanza”87 which means “killing” or “slaughter” according to one Spanish-English dictionary.88 The Russian version uses the term “убийство”89 which seems more associated with the word murder than kill.90 Finally, the Chinese version of the Convention uses the term “杀害”91 which can be translated to murder.92 Yet the fact that the final ruling likely would not be altered by this more detailed analysis does not change the opinion’s procedural shortcomings.

Ultimately, de facto bilingualism is not something that should pose any insurmountable challenge to the ICJ. Given the significant diversity of nationalities on the ICJ’s bench, there is no reason why the court cannot interpret in the languages of most treaties.93 Nevertheless, the Court appears to have a general presumption in favor of shared meaning unless a party before the Court raises an issue related to conflict between language versions. Further, and perhaps worst of all, even when there is an acknowledged conflict between language versions, the Court has failed to look at other languages versions of the treaty to try to resolve the discrepancy.94 This is to the detriment of the ICJ. A presumption of shared meaning runs contrary to the equal authenticity rule by ignoring entire language versions of a treaty. Further, it can allow clear conflicts between versions to go unnoticed, just as the lower court did in Mac.

However, any argument against a presumption of shared meaning must explain why Article 33 explicitly calls for such a presumption. Paragraph 3 says, “[t]he terms of the treaty are presumed to have the same meaning in each authentic text.”95 Yet, a presumption towards shared meaning does not mean any inquiry into possible conflicts is prohibited. As a matter of prudence, the ICJ should, as an initial step, question whether the texts contain discrepancies between language versions. After all, “[a] good practitioner would almost automatically compare the different language versions before commencing any process of interpretation.”96 Second, a presumption does not mean that the Court should ignore language versions of treaty if a conflict is found. In his critique of courts’ general presumption of shared meaning, Christopher Kuner rightly argues, “[t]he presumption of similar meaning is nothing more than a rule of convenience designed to reconcile the practice of providing authentic versions of treaties in as many as five or six languages with the general unwillingness to interpret treaties in a truly multilingual fashion.”97

​​​​​​​Resolving Discrepancies in Treaties

The challenges that the ICJ faces in resolving discrepancies between language versions of treaties are the same challenges faced by the courts of Canada, the EU or any other court that must interpret multilingual laws. This is due in part because most multilingual treaties are analyzed under shared meaning and equal authenticity rules.98 One early case before the PCIJ attempting to resolve the issues surrounding multilingual treaties is Mavrommatis Palestine Concessions.99 Greece instituted the case against Great Britain based on a dispute surrounding the meaning of Article 11 of the Mandate for Palestine.100 In the case, the court found that the English version of Article 11 created a narrower scope for the powers of the Administration. The Court went on to hold that the English version supplied the shared meaning of the two versions given its narrower scope. The Court in arguing this wrote,

Where two versions possessing equal authority exist one of which appears to have a wider bearing than the other, it is bound to adopt the more limited interpretation which can be made to harmonise with both versions and which, as far as it goes, is doubtless in accordance with the common intention of the Parties. In the present case this conclusion is indicated with especial force because the question concerns an instrument laying down the obligations of Great Britain in her capacity as Mandatory for Palestine and because the original draft of this instrument was probably made in English.101

On first blush, the PCIJ’s analysis appears to assert a rule similar to that used by Canadian courts.102 In Daoust, the Supreme Court relied on the French rather than English text of a statute because it was the narrower of the two.103 This narrower meaning rule is how some commentators have interpreted Mavrommatis Palestine Concessions.104 Yet, the Court clearly considered more than just which text was narrower. The PCIJ also indicates that the meaning in English should perhaps be favored given it was the “original” language of the Convention.

Another PCIJ case which did not make use of any narrower meaning rule is the PCIJ’s advisory opinion regarding the competence of the International Labor Organization regarding the regulation of agricultural laborers.105 The conflict at issue stemmed from the fact that the French text of Part XIII of the Treaty of Versailles used the word “industrielle,” whereas the English text used the word “industrial.”106 The PCIJ held that “context was the final test and that it had to consider the position in which the words were placed” in the treaty.107 After doing this contextual analysis, the Court held that Part XIII of the Treaty allowed the International Labor Organization to regulate the conditions of agricultural laborers. Although decided prior to the Vienna Convention on the Law of Treaties, this decision comports with Article 31 of the Convention.108

The PCIJ has also relied on drafting history when interpreting multilingual treaties. In an advisory opinion interpreting the 1919 Convention Concerning Employment of Women During the Night, the PCIJ looked at preparatory work related to the Convention.109 Further, drafting history, like the broader context of the treaty, is also recognized by the Vienna Convention on the Law of Treaties.110 Given Articles 31 and 32, one may wonder whether there are any interpretive techniques or challenges that are unique to multilingual laws and treaties. Arguably the narrower meaning rule is one, yet this rule was intentionally ignored by the International Law Commission during the Convention. Germer has gone so far as to argue the lack of unique principles for multilingual treaties in the Vienna Convention reflects the fact that there are no unique challenges presented by multilingual treaties.111

One recent case that furthers this point is LaGrand (Germany v. United States of America). This case involved the alleged deprivation of Germany’s ability to provide consular assistance during the execution of a German citizen.112 During the initial proceedings, the ICJ issued an order providing provisional measures under Article 41 of the Court’s Statute.113 In its order, the Court said that “[t]he United States of America should take all measures at its disposal to ensure that Walter LaGrand is not executed pending the final decision” of the Court.114 Germany went on to argue that the U.S. was obliged to follow this order under Article 41 as provisional measures were, according to Germany, binding.115

Yet Article 41 had a conflict between the French text and the English text. Whereas the French text of Article 41 contained the words “doivent” and “indication,” the English text contained the words “ought” and “suggested.”116 This created a potential conflict given the fact that “doivent” indicated that a provisional measure issued by the court must be followed whereas the English word “suggested” does not. The U.S. used this discrepancy to argue that Article 41 lacked any mandatory effect and therefore the U.S. was not obliged to follow any order issued from it.117 Although the ICJ does not mention it in their judgment, the U.S.’s argument is an appeal to the narrower meaning rule that was arguably laid down in Mavrommatis Palestine Concessions. However, the ICJ did not consider such an argument and instead said that it could also be argued that the French text should hold as it was the original version.118

The Court then turned to Article 33 to resolve the conflict. The Court points out that the “object and purpose” of the Statute is to prevent the court from being hampered in the exercise of its functions due to the rights of parties in a dispute not being preserved.119 It also notes that past ICJ precedent holds that parties “must abstain from any measure capable of exercising a prejudicial effect in regard to the execution of the decision to be given.”120 Germer’s analysis seems like, at least in this case, a correct descriptive analysis of what the court actually does. Rather than focusing on the task of reconciling language versions of treaties, the Court “resolves” language conflicts through standard methods of statutory interpretation. Yet whether this is the correct approach, and whether this is in fitting with the requirements of Article 33 of the Vienna Convention on the Law of Treaties, is another thing.

To begin, there are several issues with the Court’s argument in LaGrand (Germany v. United States of America). First, while there is no reason to think that a narrower meaning rule should hold, the Court never argues why as a matter of law or prudence such a rule should not be used. Second, the Court says that the French text should hold as it was the original version of the Statute. Yet the Court itself acknowledges that the French and English versions of the Statute are equally authentic.121 Any argument based around what the original was ignores the equal authenticity of all language versions and goes against Pparagraph 1 of Article 33. The Court’s analysis also ignores the other language versions of the text and adopts a “de facto bilingual” approach. This is despite the fact that Germany’s memorial to the Court and the U.S.’s counter-memorial analyze every language version of Article 41.122

Finally, while the Court’s efforts to look at the context, past precedent and purpose of the Statute are in line with Articles 31, 32, and 33, it is notable how the Court spends no time trying to reconcile the various language versions of the Statute as required by paragraph 4 of Article 33 which says “the meaning which best reconciles the texts, having regard to the object and purpose of the treaty, shall be adopted.”123

​​​​​​​Applying Canadian and CJEU Solutions to Multilingual Treaty Interpretation

The question remains as to what the Court should do instead. Thankfully, both the CJEU and Canadian courts provide useful frameworks. As referenced in Part B of Section II of this Comment, the Canadian courts rely on a relatively systematic approach to resolving discrepancies across language versions. The courts begin by first determining whether there is a conflict between the French and English texts.124 This acts as a check against de facto unilingualism (or bilingualism.)  Then the courts determine what type of conflict exists between the language versions.125 The types of conflict include: absolute conflicts where no shared meaning exists, conflicts of scope where one version is narrower than the other, and conflicts of ambiguity where one version is clear and the other ambiguous. For conflicts of ambiguity and conflicts of scope, a shared meaning can generally be found. Finally, that shared meaning will be compared to the legislative intent and purpose of the statute, or the statute’s purpose will act as a guide to the meaning of the statute if no shared meaning between the texts could exist.126

Yet how does this approach work in the context of a case like LaGrand (Germany v. United States of America)? The English text at issue in the case is:

1. The Court shall have the power to indicate, if it considers that circumstances so require, any provisional measures which ought to be taken to preserve the respective rights of either party.

2. Pending the final decision, notice of the measures suggested shall forthwith be given to the parties and to the Security Council.127

The French text on the other hand is:

1. La Cour a le pouvoir d’indicluer, si elle estime que les circonstances l’exigent, quelles mesures conservatoires du droit de chacun doivent être prises à titre proviso ire.

2. En attendant l’arrêt définitif, l’indication de ces mesures est immédiatement notifiée aux parties et au Conseil de sécurité.128

The English text is clear that the Court’s provisional measures are only recommendations. The French text’s use of the phrase “devoir” as well as the word “iniquer” implies that the provisional measures are binding. Therefore, there appears to be an absolute conflict between these two texts. Yet what of the other three authoritative language versions? According to the German memorial written to the Court, the Spanish, Chinese, and Russian texts all indicate that measures are binding rather than hortatory.129 The fact that the U.S. counter-memorial largely does not contest that point and instead argues the Court should only rely on the English text gives strong evidence that Germany’s characterizations of the other versions are correct.130 Therefore it appears as if four of the versions say that the measures are binding, whereas only one version says the provisional measures are not. Given that an absolute conflict does exist between these five texts then, under the Canadian courts’ system, the ICJ should look to the objects and purpose of the Statute which it ultimately did.

Yet this method has its flaws. Perhaps the greatest flaw stems from the fact that conflicts become significantly harder to determine when there are more than two language versions. For example, in this case the U.S. indicates that the Russian text of Article 41 is ambiguous.131 Assuming that is true, provisional measures are clearly binding under three language versions, possibly binding under one version, and clearly recommendations under another version. There is a conflict of ambiguity between Russian and any other language whereas an absolute conflict between English and the three remaining languages. It is clear that the Canadian approach works best for resolving bilingual treaties that can be more easily categorized.

Nevertheless, the Canadian approach is admirable for two reasons. First, it attempts reconciliation of, rather than ignores, conflicting language versions. This is important given the fact that language versions of treaties and laws are equally authentic and share a common meaning. To make no attempt to reconcile them is to ignore what makes multilingual treaties unique. Second, in attempting to reconcile language versions, the Canadian approach fulfills the requirements set out in Article 33.

Looking at the less systematic approach taken by the CJEU in the context of LaGrand also reveals alternative approaches. The CJEU has relied on majority and clarity approaches in its past cases involving multilingual laws and treaties.132 Under the majority approach, provisional measures are more likely to be binding as three or four of the five language versions of Article 41 align with that reading of the Statute. Germany’s memorial at least implicitly takes such an approach when it highlights how all but one language version supports the idea that provisional measures are binding. Finally, a clarity meaning depends on whether either interpretation of Article 41 can be said to be ambiguous. Given the fact that this appears to be an absolute conflict between language versions though, such an approach likely does not work.

Conclusion

Regardless of the complications they may create or the alleged benefits of drafting treaties in one language, multilingual treaties are here to stay. Given this, it is important that the principles built into them—the equal authenticity and shared meaning rules—are respected. As former Canadian Supreme Court Justice Michel Bastarache said, “[t]he equal authenticity and shared meaning rules are best seen as the fullest expression possible of the principle of bilingualism in legislation, and the importance of their purpose outweighs the practical difficulties that their application may sometimes present.”133 Although said in the context of bilingual legislation, the sentiment is just as true for multilingual treaties.

Yet the ICJ’s past practices have often done just the opposite. By presuming no conflicts exist, ignoring conflicts when they are mentioned, and relying on standards methods of interpretation, the Court has failed to do just what Article 33 of the Vienna Convention on the Law of Treaties requires. It enables, rather than reconciles, disconsonant language versions. A better path is possible though. Both Canadian and the EU experiences interpreting multilingual legislation indicate just how the ICJ may resolve the challenges facing the interpretation of multilingual treaties. These solutions include an end to the ICJ’s de facto bilingualism and greater reliance on the interpretive principles focused on the reconciliation of language versions.

• 1S.A.P., Towards a Fairer Distribution, The Economist, (Apr. 2, 2013), https://perma.cc/6UEV-5GPT.
• 2Id.
• 3Id.
• 4Id.
• 5Id.
• 6See Lawrence M. Solan, The Interpretation of Multilingual Statutes by the European Court of Justice, 34 Brook. J. Int'l L. 277, 277 (2009); see also Marie-éve Hudon, Library of Parliament, Bilingualism in Canada’s Court System: The Role of the Federal Government 4 (2020).
• 7See Hudon, supra note 6.
• 8See Solan, supra note 6.
• 9Official Languages Act, R.S.C., c 31, § 13 (1985) (Can.); see also Cornelis J. W. Baaij, Fifty Years of Multilingual Interpretation in the European Union, in Lawrence M. Solan & Peter M. Tiersma, The Oxford Handbook of Language and law 217 (2012).
• 10See Baaij, supra note 9.

• 11Michel Bastarache, Bilingual Interpretation Rules as a Component of Language Rights in Canada, in Lawrence M. Solan & Peter M. Tiersma, The Oxford Handbook of Language and law 158, 162 (2012).
• 12Id. at 161.
• 13Bastarache, supra note 11, at 163; see also Canadian Pacific Railway Co. v. Robinson (1891), 19 S.C.R. 292 (Can.) (holding the equal authenticity rule and shared meaning rule applied to the laws of Quebec). It was only in 1935 with the case R. v. Dubois that these two rules were applied to federal legislation. See R. v. Dubois (1935) S.C.R. 378 (Can.).
• 14Bastarache, supra note 11, at 164.
• 15See Ruth Sullivan, The Challenges of Interpreting Multilingual, Multijural Legislation, 29 Brook. J. Int'l L. 985, 1007 (2004).
• 16Id. at 1012.
• 17Karen McAuliffe, Language And Law In The European Union: The Multilingual Jurisprudence Of The Ecj, in Lawrence M. Solan & Peter M. Tiersma, The Oxford Handbook of Language and law 200 (2012).
• 18Id. at 201.
• 19Id.
• 20Government of Canada, Discover Canada–Canada’s History (Oct. 26, 2015), https://www.canada.ca/en/immigration-refugees-citizenship/corporate/publications-manuals/discover-canada/read-online/canadas-history.html.
• 21See Baaij, supra note 9.
• 22Id. at 218; see also Case 296/95, EMU Tabac, 1998 E.C.R. I-1605, ¶ 36 (All languages must be given “the same weight.”).
• 23Statistics Canada, Language–2021 Census Promotional Material, (last modified Sept. 19, 2022), https://perma.cc/M5GP-BJ3N.
• 24Id.
• 25See Statistics Canada, English-French Bilingualism Reaches New Heights 3 (2017).
• 26Id.
• 27Id. In Quebec and New Brunswick, the rates of bilingualism are 44.5% and 33.9% respectively.
• 28See Pierre-André Côté, Bilingual Interpretation of Enactments in Canada: Principles v. Practice, 29 Brook. J. Int’l L. 1067, 1073 (2004).
• 29See Bastarache, supra note 11, at 166. Bastarache has argued that most counsel and most judges in Canada are unilingual.
• 30Bastarache, supra note 11, at 166; see also R. v Mac, (2002) 1 S.C.R. 856 (Can.).
• 31Bastarache, supra note 11, at 166.
• 32Id.
• 33Id.
• 34Id.
• 35Bilingualism rule squeezes selection of a new Supreme Court justice, The Globe and Mail (Sept. 28, 2023), https://perma.cc/A3BQ-JZR5.
• 36McAuliffe, supra note 17, at 211.
• 37Bastarache, supra note 11, at 166; see also R. v Mac, supra note 30.
• 38Case 283/81, CILFIT v. Ministry of Health, 1982 E.C.R. 3415, ¶ 18. See also Case 30/77, R. v Bouchereau, 1977 E.C.R. 1999, ¶ 13.
• 39See R. v Mac, supra note 30.; see also R. v. Daoust (2004) 1 S.C.R. 217 (Can.).
• 40Bastarache, supra note 11, at 164.
• 41R. v. Daoust, supra note 39.
• 42Bastarache, supra note 11, at 165.
• 43Id.
• 44Id .The logic being that the narrow text is more able to incorporate the meaning of the broad text and therefore satisfy the shared meaning rule.
• 45Id.
• 46This can include inquiries into the legislative intent of Parliament at the time of drafting. Id.
• 47Id; see also R. v. Daoust, supra note 39.
• 48Bastarache, supra note 11, at 165.
• 49Id.
• 50Id.
• 51Id.
• 52Id. at 166.
• 53See Baaij, supra note 9; see also CILFIT v. Ministry of Health, supra note 38.
• 54Baaij, supra note 9, at 219.
• 55Id. at 219.
• 56Case 30-77, Régina v. Bouchereau, 1977 E.C.R. 01999, ¶ 14.
• 57Baaij, supra note 9, at 220.
• 58Case 426/05, Tele2 v. Telekom, 2008 E.C.R. I-685, ¶ 25. The court has described this purposivist approach as “settled law.”
• 59Baaij, supra note 9, at 220.
• 60Baaij, supra note 9, at 221.
• 61Id.
• 62Id. at 221–22.
• 63Id. at 222.
• 64Case C-236/97, Codan, 1998 E.C.R. I-8679, ¶¶ 23, 29.
• 65Bastarache, supra note 11, at 165.
• 66Baaij, supra note 9, at 224.
• 67Id.
• 68Dinah Shelton, Reconcilable Differences--The Interpretation of Multilingual Treaties, 20 Hastings Int’l & Comp. L. Rev. 611, 614 (1997).
• 69Id.
• 70John King Gamble et al., Choice of Official Text in Multilateral Treaties: The Interplay of Law, Politics, Language, Pragmatism and (Multi)-Nationalism, 12 Santa Clara J. of Int’l L. 29, 46 (2014).
• 71Permanent Court of International Justice, International Court of Justice, https://perma.cc/2KCA-NLS7.
• 72Peter Germer, Interpretation of Plurilingual Treaties: A Study of Article 33 of the Vienna Convention on the Law of Treaties, 11 Harv. Int’l. L. J. 400, 401 (1970).
• 73Vienna Convention on the Law of Treaties art. 33, May 23, 1969, 1155 U.N.T.S. 331.
• 74Germer, supra note 72, at 413.
• 75Id.
• 76Id. at 414.
• 77Bastarache, supra note 11, at 166.
• 78Application of the Convention on the Prevention and Punishment of the Crime of Genocide (Gam. v. Myan.), Preliminary Objections, Judgment, 2022 I.C.J. Rep. 5, ¶ 80 (July 22); see also Application of the International Convention on the Elimination of All Forms of Racial Discrimination (Qatar v. U.A.E.), Preliminary Objections, Judgment, 2021 I.C.J. Rep. 5, ¶¶ 79, 80 (Feb. 4).
• 79Application of the Convention on the Prevention and Punishment of the Crime of Genocide (Croat. v. Serb.), Judgment, 2015 I.C.J. Rep. 10, (Feb. 3).
• 80Id. at ¶ 51.
• 81Id. at ¶ 156.
• 82Convention on the Prevention and Punishment of the Crime of Genocide, 9 December 1948, International Humanitarian Law Database, https://perma.cc/RDM4-2USA.
• 83Application of the Convention on the Prevention and Punishment of the Crime of Genocide (Bosn. & Herz. v. Serb.), Judgment, 2007 I.C.J. Rep. 8, ¶ 186 (Feb. 26); see also Prosecutor v. Vidoje Blagojević & Dragan Jokić, Case No. IT-02-60-T, Judgment, ¶ 642 (Int’l Crim. Trib. for the Former Yugoslavia Jan. 17, 2005).
• 84Prosecutor v. Vidoje Blagojević & Dragan Jokić, Case No. IT-02-60-T, Judgment, ¶ 642, n. 2057 (Int’l Crim. Trib. for the Former Yugoslavia Jan. 17, 2005).
• 85Id.
• 86Supra note 83.
• 87Convención para la Prevención y la Sanción del Delito de Genocidio, The Office of the High Commissioner For Human Rights, https://www.ohchr.org/es/instruments-mechanisms/instruments/convention-prevention-and-punishment-crime-genocide.
• 88Translation of matanza–Spanish–English dictionary, Cambridge Dictionary, https://perma.cc/G8ZL-DHE2.
• 89Конвенция о предупреждении преступления геноцида и наказании за него, United Nations, https://perma.cc/7XHJ-2XPQ.
• 90Translation of murder–English–Russian dictionary, Cambridge Dictionary, https://perma.cc/HDG8-35ZQ.
• 91防止及惩治灭绝种族罪公约, United Nations, https://perma.cc/G8QV-4QQ6.
• 92杀害 的翻譯 中文(简体)-英語詞典, Cambridge Dictionary, https://perma.cc/MB94-3NBW.
• 93Current judges are from: the U.S., Lebanon, Uganda, Slovakia, France, Somalia, China, India, Japan, Germany, Australia, Brazil, Mexico, Romania, and South Africa. See Current Members, International Court of Justice, https://perma.cc/8NND-WL2J.
• 94Application of the Convention on the Prevention and Punishment of the Crime of Genocide (Croat. v. Serb.), Judgment, 2015 I.C.J. Rep. 10 (Feb. 3).
• 95Vienna Convention on the Law of Treaties, art. 33, May 23, 1969, 1155 U.N.T.S. 331.
• 96Shabtai Rosenne, 16 Yearbook of the International Law Commission (1966), Vol. I, part 2, 874th meeting, at 209, ¶ 11.
• 97Christopher B. Kuner, The Interpretation of Multilingual Treaties: Comparison of Texts Versus the Presumption of Similar Meaning, 40 The Int’l and Compar. L. Q. 953, 962 (1991).
• 98Germer, supra note 72, at 401.
• 99Shelton, supra note 68, at 628.
• 100Jean Hardy, The Interpretation of Plurilingual Treaties by International Courts and Tribunals, 37 Brit. Y. B. Int’l. 72, 76 (1961).
• 101Mavrommatis Concessions, P.C.IJ. (ser. A) No. 2, at 19.
• 102Bastarache, supra note 11, at 165.
• 103Id.
• 104Shelton, supra note 68, at 628.
• 105Germer, supra note 72, at 415; see also [1922] P.C.I.J., ser. B, No. 2.
• 106Germer, supra note 72, at 415.
• 107Id.
• 108Vienna Convention on the Law of Treaties, art. 31, May 23, 1969, 1155 U.N.T.S. 331.
• 109Id. at 420.
• 110Vienna Convention on the Law of Treaties, art. 32, May 23, 1969, 1155 U.N.T.S. 331.
• 111Germer, supra note 72, at 425–26.
• 112LaGrand (Ger. v. U.S.), Judgment, 2001 I.C.J. Rep. 4, ¶ 11 (June 27).
• 113Id. at ¶ 32.
• 114Id. at ¶ 32.
• 115Id. at ¶ 44.
• 116Id. at ¶ 100.
• 117Id. at ¶ 101.
• 118Id. at ¶ 100.
• 119Id. at ¶ 102.
• 120Id. at ¶ 103.
• 121Id. at ¶ 101.
• 122LaGrand (Ger. v. U.S.), Memorial, 1999 I.C.J. Rep. 1, ¶ 4.149 (Sept. 16); see also LaGrand (Ger. v. U.S.), Counter-Memorial, 2000 I.C.J. Rep. 1, ¶ 152 (Mar. 27).
• 123Vienna Convention on the Law of Treaties, art. 33, May 23, 1969, 1155 U.N.T.S. 331.
• 124R. v. Daoust [2004] 1 S.C.R. 217 (Can.).
• 125Bastarache, supra note 11, at 165.
• 126Id.
• 127LaGrand (Ger. v. U.S.), Judgment, 2001 I.C.J. Rep. 4, ¶ 100 (June 27).
• 128Id.
• 129LaGrand (Ger. v. U.S.), Memorial, 1999 I.C.J. Rep. 1, ¶ 4.149 (Sept. 16).
• 130See LaGrand (Ger. v. U.S.), Counter-Memorial, 2000 I.C.J. Rep. 1, ¶ 152 (Mar. 27).
• 131Id.
• 132Baaij, supra note 9, at 220–21.
• 133Bastarache, supra note 11, at 167.