Excessive Multilingualism in EU Trade Agreements

Written by , , , , and

The legal protection of multilingualism is an important principle and an indispensable guarantee for the functioning of the institutions of the European Union (EU) as well as for their relationships with EU citizens. This is not only evidenced by Article 22 of the Charter of Fundamental Rights, which obligates the Union to respect linguistic diversity. Beyond that, legally protecting multilingualism is, as the European Parliament stated, “not a matter of communication only, but also a question of democratic legitimacy towards citizens and respect for the cultural diversity of the Member States. It affects the way in which EU legislation is drafted and interpreted”.

Multilingualism is also well established in the EU Treaties themselves, concluded between the Member States in 24 equally authentic languages (Article 55 TEU), which can be interpreted authoritatively by the Court of Justice of the European Union (CJEU) whenever necessary.

The practice of the European Union is quite similar with regard to treaties concluded with non-Member States. In particular, several free trade agreements (FTAs) concluded or negotiated with such states have been drawn up in no less than 23 or 24 equally authentic languages.

The Canada-European Union (EU) Comprehensive Economic and Trade Agreement (CETA), for instance, has been drawn up “in duplicate in the Bulgarian, Croatian, Czech, Danish, Dutch, English, Estonian, Finnish, French, German, Greek, Hungarian, Italian, Latvian, Lithuanian, Maltese, Polish, Portuguese, Romanian, Slovak, Slovenian, Spanish and Swedish languages, each version being equally authentic” (see Article 30.11. Irish is not included here in accordance with Council Regulation 2015/2264). In addition to that, the FTA with Vietnam also includes Vietnamese as an equally authentic language.

However, even though such practice promises public accessibility of legal documents and thus more transparency, it is – especially from the perspective of international law – highly questionable and should be carefully reconsidered. Article 33 of the 1969 Vienna Convention of the Law of the Treaties (VCLT) – a treaty which is regularly referred to in FTAs concluded by the EU, e.g. in Articles 8.31.1 and 29.17.1 of CETA – first posits that where authenticated in two or more languages a treaty has the same legal value, and is presumed to bear the same meaning, in each such language. With 23 or 24 equally authentic languages, one can only hope that everything goes smoothly. Serious problems arise if the comparison of the authentic texts discloses a difference of meaning, an event that is not so unlikely considering the complexity of trade and investment provisions. Unlike the CJEU, investment tribunals or arbitral panels exercising their jurisdiction under FTAs are likely to be ill equipped to interpret treaties concluded in such large numbers of equally authentic languages. Yet, even when the CJEU itself had to deal with multilingualism and diverging interpretations in Case C-36/98 Spain v Council [2001] ECR I-779, para 49. It cursorily swept away any concerns and declared that “[i]n the case of divergence between the language versions of a [Union] measure, the provision in question must be interpreted by reference to the purpose and general scheme of the rules of which it forms part […].”

Article 33(4) VCLT then directs the interpreter to apply Articles 31 and 32 of the VCLT. The first of these provisions contains a general rule of interpretation. The element of this rule directing attention to the ‘ordinary meaning’ to be given to the terms of a treaty (in their context and in the light of its object and purpose) immediately presents obvious potential for diverging starting points for interpretation, if extensive multilingualism is used. The second of these provisions governs the use of supplementary means and in particular the legislative history of the treaty. Applying these provisions is rarely a simple exercise and it may be expected to become rather arduous with 23 or 24 authentic versions of a text. Working with such treaties would provide the perfect ground to unleash lawyers with a taste for linguistic nuances and for particular niceties flowing from the legal systems most associated with each of the various languages. Lastly, if application of the general rule and supplementary means fails to yield the meaning, the interpreter returns to Article 33 VCLT which requires a (probably desperate) attempt to determine the meaning that best reconciles the 23 or 24 texts, having regard to the object and purpose of the treaty. Even assuming that it is possible to resolve the issues of interpretation by this means, the process is likely to be complex, time consuming and fraught with difficulties.

It is evident that Articles 31 to 33 VCLT are unlikely to function well (if at all) with such large numbers of equally authentic languages. Indeed, our argument is that the conclusion of treaties in 23 or 24 authentic languages

  • is a call for discrepant and inaccurate outcomes;
  • makes the treaty vulnerable to a multitude of potential interpretations which would be extremely difficult to manage under Articles 31 to 33 VCLT;
  • may be expected to complicate intolerably the settlement of disputes and inflate the related costs; and
  • ultimately undermines legal certainty (not to mention the risks of significantly slowing down the negotiations and increasing the costs generated by the treaty throughout its existence).

The number of authentic languages of FTAs concluded between the EU and other treaty partners should therefore be greatly reduced. One option could be to limit the authentic languages to the relevant UN official languages (English, Spanish and French) plus the language(s) of the counter-part, if appropriate. Incidentally, English, Spanish and French are the official languages of the World Trade Organization (WTO), whose jurisprudence the arbitration panels have to take into account in interpreting FTAs (i.e. Article 29.17 CETA). Furthermore, the UN Treaty Section recommends State to conclude treaties only in the UN official languages in order to facilitate their registration under Article 102 of the UN Charter.

Alternatively, the FTAs could be concluded in three or four authentic languages different from those used in the WTO and presumably including German and Italian. This option would have the advantage of better reflecting the main languages spoken in the EU, but the selection may require endless negotiations and be rather unpalatable to some Member States. As the lingua franca of international trade, English cannot be excluded, regardless of Brexit.  

Another option is to agree that in case of divergence a particular text prevails. This option may be even more problematic than the previous two, or even impracticable when the counter-part’s language is not an official language of the EU (as in the case of the FTA with Vietnam).

Thus, from the standpoint of the law of the treaties, but also of common sense, concluding a treaty in 23 or 24 equally authentic languages seems most unwise. While legal multilateralism is of great importance, and must be defended within the EU (with regard to both EU treaties and legislation), it is not indispensable in the EU’s external relations and is likely to prove counter-productive from the standpoint of the FTAs concluded by the EU with third States. 

The EU should seriously reconsider its practice and strictly limit the number of authentic languages. As happens with treaties in other fields, for practical purposes, ‘official versions’ (that is translations that have not been agreed as authentic, but prepared by competent international or national authorities) may reduce day-to-day difficulties in the application of treaties, it being understood that in case of controversy over interpretation authentic texts must be used. This may be unappealing to certain Member States, but in this case, practicality should prevail over national pride and parochialism.

Print Friendly, PDF & Email


Leave a Comment

Comments for this post are closed


Lorand Bartels says

February 15, 2018

Not irrelevantly, there is a significant number of WTO reports in which the three official languages have been scrutinised. This has as a rule been to confirm an interpretation. But the point is, as you say, one cannot assume the same approach as the CJEU: Article 33 does mean something in trade disputes. Good argument!

Theodor Schilling says

February 15, 2018

Thank you for an interesting, and overdue, contribution. There can be no doubt about the correctness of your analysis concerning EU trade agreements. However, as I have argued repeatedly (eg https://static1.squarespace.com/static/56330ad3e4b0733dcc0c8495/t/56b7125959827e73e35b3c36/1454838361490/GLJ_Vol_12_No_07_Schilling.pdf), the same applies to internal EU law. To quote from that article's conclusion, the question must be "whether the opportunities offered by strong multilingualism as prescribed by EU law effectively offset the problems it causes. Two such problems have been identified: the possibility of intractable contradictions between language versions which ... is real enough, and the very indeterminacy of all laws which is the necessary counterpart of the possibility — purely theoretical in the EU context — of linguistic triangulation. This ... question must be answered in the negative. The main aspect of this evaluation is that the one additional opportunity strong multilingualism is claimed to offer over the weak variety to wit the possibility of linguistic triangulation has historically never been grasped by the ECJ and that today a 23-angulation is not a practical possibility. In short, strong multilingualism offers nothing to offset the problems it causes, leaving aside that it may address certain political preoccupations of Member States. ...
[T]here is therefore much to be said for replacing the 23 equally authentic official language versions of EU law by one such version and 22 official translations, thereby opting for weak multilingualism. The above study of linguistic triangulation can
be read as an argument in favor of the adoption of weak multilingualism which, compared to strong multilingualism, (i) avoids the problem of intractable contradictions between language versions, (ii) allows to reduce the indeterminacy of EU law to what is inevitable in any legal system,(iii) facilitates a meaningful access to EU law and (iv) offers citizens the same practical opportunities as strong multilingualism. Weak
multilingualism therefore should be the system of choice for the EU also under the aspects discussed in this contribution."

Martin Holterman says

February 15, 2018

I would think that, realistically, the language or languages that were actually used to negotiate the treaty will typically get priority in case of an inconsistency between the versions. The court/arbiter may not want to admit it, but I'm confident that that's what they would normally do. (And that seems reasonable, from the point of view of the general law of contracts.)

Alessandra Asteriti says

February 16, 2018

Very interesting, I am very taken by these issues, and have had to tackle issues of interpretation and translation a lot in my non-legal past. But I was wondering if this is a problem likely to present itself in reality precisely on these terms, with 25+ languages' competing interpretations overwhelming the relevant tribunal or court? Most likely, the languages actually playing a role in the dispute will be looked at (ie, if it is a case involving a French company in Vietnam, French and Vietnamese, or whatever other combination thereof). I doubt that, in the mentioned example, the tribunal will decide to look at the Bulgarian translation of the FTA to see if it sheds some light on the dispute. Alternatively, as proposed, the negotiating languages will have precedence, also because, in the event the travaux préparatoires will be looked at as per Article 32 VCLT, those will be the relevant languages.

It seems to me the real issue is that even only two languages, equally authoritative, can engender intractable difficulties in interpretation, especially overtime, as languages evolve and might diverge in subtle ways not easily captured and understood by tribunal members (who are not linguists). Equally, treaties capture a moment in time and the relation between the parties at that moment in time, which means, among other things, that the choice of words might have been dictated by the stronger party, with the counterpart laboriously translating (I very much doubt Vietnamese was the starting point of the EU-Vietnam FTA). Whether or not Vietnamese is an equally authoritative version, the fact remains that the Vietnamese text is a translation in a way the English text probably is not. And this is of course because translations do not translate words, but worlds. There are points of contact between these worlds, and that is how translation is possible in the first place, but words are in contact with each other the way Venn diagrams are, where convergence and divergence are always present. The task of the tribunal or court is to find the space where the words meet, and disregard where they diverge, and this will always be a bilateral exercise, even if repeated 20 times.

I am aware these are not the stated conditions of a dispute, if nothing else, because a presumption of sovereign equality applies, hence 'the equally authoritative versions' as if all were translations/versions of some platonic idea of the perfect text.