3.1Overview of text production

The first full textual mapping that gave rise to the LINST corpus shows that, with 379 980 texts, the EU setting is by far the most productive compared to the 80 806 texts compiled from the UN (and the ICJ) and 52 854 texts from the WTO. The prominence of the EU’s legislative activity and the significant number of draft and preparatory documents used to this end certainly contribute to the high volume of the EU subcorpus, although the work of the CJEU also accounts for a large proportion of the volume (see Section 3.2). Unlike the EU institutions, the UN and the WTO lack the supranational authority to adopt directly applicable legal acts. As IGOs, their work is more focused on promoting and monitoring Member States’ compliance with international agreements adopted under their auspices or other non-binding initiatives and instruments (see Section 3.4).

Since the LINST dataset includes non-translated texts, it also offers an indication of the standing of the three selected languages in terms of their frequency of use for text production, and more particularly, for procedural purposes that do not require the translation of documents into all the official languages. In other words, the first textual mapping provides a snapshot of language regimes as applied in the three settings of the study, and confirms the predominant use of English as a lingua franca in international organizations. Overall, texts compiled in English (either as monolingual documents or as versions of multilingual documents) account for 43.40% of the total number of texts, whereas French accounts for 29.66% and Spanish for 26.94%. These figures also imply that French and Spanish are predominantly target languages in translation (except in the CJEU and the ICJ). Approximately 5% of UN documents in French and 3% in Spanish in the LINST subcorpora are source texts. In the WTO subcorpus, 13.32% of documents were originally drafted in Spanish, whereas less than 2% have French as the original language.

However, the most marked difference between languages is found for the EU institutions,33.For the sake of brevity, unless otherwise specified, ‘the EU’ will hereinafter refer to the EU institutions selected for the project. particularly at the EUCO and the EP. The upward trend in the use of English is most pronounced for texts compiled from the EP Public Register, where the number of documents in English grew from 18 836 in 2005 to 43 215 in 2015, compared to 15 257 and 12 169, respectively, in French. This shows a trend in terms of language preference for committee work and internal communication at the EP after the 2004 and 2007 EU enlargements. In the case of the EUCO, according to the Language Service of its General Secretariat, up to 70% of the total pages produced for the EUCO are not translated “since, for practical purposes, most of the Working Parties work on the basis of a text drafted in a single language” (General Secretariat of the Council of the EU 2012General Secretariat of the Council of the EU 2012 The Language Service of the General Secretariat of the Council of the European Union – Making Multilingualism Work. Brussels: European Union., 8). This language tends to be English, as also reflected in the LINST data from the EUCO Public Register, where 10 576 of the documents were produced in English in 2005 and 12 151 in 2015, while the number of documents in French decreased from 6379 to 4911 in the same period.

As shown in Figure 1, the selection of translated texts for the LETRINT 0 corpus entailed discarding a much higher proportion of texts available only in English in the EU institutions (more than 50% compared to roughly 40% at the UN and 25% at the WTO). It is also in this context that the most remarkable difference is found between English and the second most used language, French, in terms of the number of documents discarded. In the case of the UN, the far greater volume of French texts discarded compared to Spanish reflects the distinctive status of French, together with English, as a working language of the organization and an official language of the ICJ. In contrast, the difference in the number of documents in these two languages in the WTO is negligible, and very limited with regard to English, which bears out the more systematic application of the WTO trilingual regime. This also points to a clear pattern in which the coexistence of more official languages leads to more instances of restricted multilingualism in practice and the more frequent use of English as a lingua franca (for a more detailed analysis of language regimes and hierarchies at these institutions, see Prieto Ramos [2020b]Prieto Ramos, Fernando 2020b “Translation at International Organizations: The Legal and Linguistic Hierarchies of Multilingualism.” In The Oxford Handbook of Translation and Social Practices, edited by Meng Ji and Sara Laviosa, 455–477. New York: Oxford University Press. ).

3.2Distribution of text production per legal function

The results of the categorization of translated texts corroborate the legal dimension of institutional translation at international organizations, with significant variation between settings. Overall, 90.37% of the LETRINT 0 corpus (based on total number of words) falls within the three main legal functional categories presented in Section 2, while 9.63% falls within the fourth category, administrative functions. Of the three main categories, documents translated as part of law- and policy-making procedures account for 50.16% of the dataset, followed by monitoring (37.37%) and adjudication (12.48%).

Interestingly, the distribution of key and secondary texts varies according to the legal function. The highest proportion of secondary texts is found in hard law-making, with an almost four times greater text volume (mostly input documents) than that of the legal instruments themselves. This is in line with the nature of the preparatory work required and made publicly available in law-making procedures. The distribution of key and secondary texts is similar in the case of soft law and other policy formulation at the UN (where input is also significant in the preparation of resolutions in particular) and the WTO (which is characterized by time-consuming consensus-building processes in the negotiation of multilateral agreements). The reverse trend is found in monitoring procedures, where there is a higher volume of performative key documents than preparatory or derived texts. This trend is more marked in the case of adjudicative procedures, where the group of accessible translated genres fulfilling secondary functions, mainly composed of instrumental or derived texts such as minutes, annual reports or press releases, represents a very small proportion compared to the volume of key genres (see Section 3.4).

A closer comparison of the subcorpora (volume of translated texts based on the word count of final versions of source texts)44.Drafts, corrigenda, and revisions are excluded from this comparison in order to focus on translation volumes of new texts (see Section 3.3). highlights the distinctive features of each institutional legal setting (see Figure 2). The main difference between them is the prominence of law-making at the EU and of monitoring procedures at the IGOs. As a supranational legal order in its own right, the EU regulates a wide array of areas that affect the lives of its more than 500 million citizens. A total of 57.67% of the translation volume, including a significant proportion of secondary texts (see Table 2), is produced in preparation for binding legal acts by the EC, the EUCO, and the EP in the so-called ordinary legislative procedure (co-decision before the entry into force of the Lisbon Treaty on 1 December 2009). Non-binding law- and policy-making accounts for a more modest 12.91% of the translation volume. This figure is similar to that obtained for this subcategory at the UN (11.20%) as the main one within law-making in that setting, compared to only 0.63% for binding instruments. This reflects the UN’s more common reliance on instruments of limited enforceability, especially resolutions (see Section 3.4). At the WTO, the translation of binding instruments, including schedules of trade commitments, accounts for 5.68% of the subcorpus, as opposed to 1.54% for soft law and policy formulation. In both subcategories, the bulk of translation volumes is composed of preparatory texts issued in the framework of trade negotiations.

At the UN and the WTO, translation volumes for monitoring mechanisms stand out with 61.34% and 65.28% of the corresponding subcorpora, respectively, of which 27.34% and 44.07% originate in the surveillance of Member States’ mandatory compliance with their international commitments. At the UN, a great proportion of these texts are produced by or for human rights treaty bodies, such as the Human Rights Committee, while at the WTO they are most often part of periodic Trade Policy Review (TPR) procedures. These figures are much higher compared to the 12.95% in the case of EU monitoring functions, including 3.49% in connection with mandatory compliance. This is certainly due to the fact that the role of the EC in monitoring the implementation of EU law, as the ‘Guardian of the Treaties’, is conducted through mechanisms that do not always involve public access to documents in all official languages, but rather bilateral contacts with Member States that may lead to infringement procedures55.See https://​ec​.europa​.eu​/info​/law​/law​-making​-process​/applying​-eu​-law​/infringement​-procedure​_en. and, eventually, referrals to the CJEU. Other monitoring and implementation matters account for 9.35% of translation volume at the EU, compared to 14.20% at the WTO and 34.01% at the UN, where this share goes hand-in-hand with the important role of soft law in the same organization.

As regards pre-accession monitoring (Subcategory 2.2), its absence from translation data in the UN subcorpus has a twofold explanation. On the one hand, the two latest accessions to this organization (Montenegro and South Sudan) took place in years that are not covered by the LETRINT corpus (2006 and 2011, respectively). On the other hand, following the recommendation of the Security Council, the General Assembly grants UN membership to States without the need for extensive negotiations, as the UN “is open to all peace-loving States that accept the obligations contained in the United Nations Charter and, in the judgment of the Organization, are able to carry out these obligations” (United Nations 1945United Nations 1945 Charter of the United Nations and Statute of the International Court of Justice. San Francisco: United Nations., art. 4). This means that accession procedures do not generate significant translation volumes as in the case of the EU and the WTO, where these procedures often take several years. At the EU, where pre-accession monitoring accounted for 0.11% of the relevant subcorpus, candidate States must comply with EU standards and rules, and secure the consent of their citizens, EU institutions, and other Member States.66.See https://​ec​.europa​.eu​/neighbourhood​-enlargement​/policy​/conditions​-membership​_en. Several countries were negotiating their accession during the 2005–2015 period. Among them, Bulgaria and Romania joined the EU in 2007, and Croatia in 2013. Similarly, accession to the WTO involves the establishment of a working party through which negotiations lead “to the drafting of an accession package (‘terms of accession/entry’)”.77.See https://​www​.wto​.org​/english​/thewto​_e​/acc​_e​/acc​_status​_e​.htm. These negotiations generate a considerable amount of translation in this organization, which accounts for 7.02% of the WTO subcorpus.

For their part, the volume of adjudication texts varies considerably between institutions, and reflects the prominence of this function and the dynamism of the relevant bodies in each setting. The most significant proportion is found at the WTO (24.09%), as a result of its successful dispute settlement system (e.g., McRae 2010McRae, Donald 2010 “The WTO Appellate Body: A Model for an ICSID Appeals Facility?” Journal of International Dispute Settlement 1 (2): 371–387. ); although it is the CJEU that deals with the highest number of cases and produces the largest volume of translation in absolute terms. While the overall proportion of adjudicative texts at the EU is 12.88%, key texts of this functional category (totalling 20.78 million words) actually outnumber the translation volume of key texts in all other subcategories in the LETRINT 0 corpus, and in the case of the second largest subcategory of key texts, hard law (totalling 19.59 million words), this is by more than one million words. At the other extreme, the ICJ production, with a low number of cases, accounts for just 2.66% of translation volume in the UN setting. In this organization, it is administrative functions rather than adjudication that accounts for the second largest proportion of translated texts (24.16% of the subcorpus), much more than in the EU institutions (3.60%) and the WTO (3.41%). This may be due not only to the attention devoted to budgets and other internal matters through bodies such as the Office of Internal Oversight Services and the Advisory Committee on Administrative and Budgetary Questions, but also to the high number of miscellaneous follow-up questions derived from monitoring over the years (such as logistics, appointments, and other coordination matters).

3.3Document status

In order to complete the picture sketched in Sections 3.1 and 3.2, which focuses on the translation volume of new texts (i.e., the final versions of texts only), it is worth examining the volume of translated texts with other statuses in the production workflow. This may help to better understand multilingual text production practices in each setting. Four main statuses were identified as part of the metatada: ‘draft’, ‘first final version’, ‘corrigendum’, and ‘revision’. Drafts are used to produce the first final versions of documents, whereas corrigenda and revisions draw on the first final versions of documents to introduce corrections or to update contents, respectively. Overall, first final versions account for 78.23% of the LETRINT 0 corpus (based on source text word count), followed by drafts (18.2%), revisions (3.41%), and corrigenda (0.16%).

The most notable differences between settings (see Figure 3 and Table 3) are the high volume of publicly accessible drafts at the EU (25.93% versus 6.65% at the WTO), the large proportion of revisions at the WTO (13.23% versus 2.32% at the EU) and the comparatively low figures for these statuses at the UN, where drafts (3.88%), revisions (1.31%), and corrigenda (0.07%) represent less than 6% of the subcorpus. In contrast, first final versions account for just less than 80% at both the EU (71.55%) and the WTO (79.98%), compared to 94.74% at the UN. Although the share of corrigenda is very limited in the subcorpora, the proportion for the UN (64 722 words or 0.07%) is particularly low when compared with the corresponding proportion at the WTO (0.14%) and the higher figures at the EU (457 472 words or 0.2%). This points to differences in the frequency of corrigenda at the observed institutions (see Bobek 2009Bobek, Michal 2009 “Corrigenda in the Official Journal of the European Union: Community Law as Quicksand.” European Law Review 34: 950–962.; Prieto Ramos 2020aPrieto Ramos, Fernando 2020a “Facing Translation Errors at International Organizations: What Corrigenda Reveal about Correction Processes and their Implications for Translation Quality.” Comparative Legilinguistics 41 (1): 97–133. ).

An examination of document status per legal category provides more nuance on institutional singularities (see Figures 4 to 6). In absolute terms, the amount of publicly accessible drafts in the EU’s deliberative procedures for law-making (with a combined total of 47 million words and 28.32% of hard and soft law-making texts) is, by far, the most significant of all the subcorpora (see Figure 4), although, in relative terms, drafts are even more prominent in EU accession negotiations (62.86% of Subcategory 2.2 in Figure 4). On the contrary, as in the other settings, drafts are rarely accessible in the case of adjudication procedures (Subcategory 3), and documents other than final versions are the exception for this functional category. These trends align with the needs and dynamics of each function, and with the use and visibility of secondary texts in the same procedures (see Section 3.1).

In the UN subcorpus (see Figure 5), drafts are also more accessible in law-making procedures (23.42% and 10.52% of the word counts of Subcategories 1.1 and 1.2, respectively) than in other legal categories. Interestingly, draft papers in the framework of negotiations on binding instruments (65.48% of Subcategory 1.1 drafts) and resolutions (73.18% of Subcategory 1.2 drafts) are the most frequent among law-making textual genres (see also Section 3.4).

In the WTO subcorpus (Figure 6), similar proportions are found in law-making (with a combined percentage of 8.53%), but it is pre-accession monitoring (Subcategory 2.2) that registers the largest proportion of drafts (43.16%). As mentioned above, nonetheless, the main peculiarity of the WTO’s text production is the high amount of revisions. Unlike the EU and the UN, where this status rarely exceeds 3% of word counts in any single legal category, revisions reach an overall proportion of 8.13% at the WTO, and 22.54% in the case of mandatory compliance monitoring (Subcategory 2.1), essentially because of the frequent recourse to revisions of TPR Reports by the WTO Secretariat.

3.4Textual genres

As mentioned in Section 2, all texts were examined to establish their genres and situate them with regard to the main functional categories, which involved checking not only document titles, series, and symbols, but also (especially in cases of doubt) conventions such as macrostructure and phraseology associated with specific functions. Key performative genres such as legal acts or judgments were generally easy to identify, while other more hybrid or multifunctional texts, particularly certain types of reports and communications, required further scrutiny for their categorization. From a legal perspective, law-making genres, despite representing a more limited volume of translation at the two IGOs, are systematically found at the top of each legal order and condition other key genres related to the application of legal instruments (e.g., in national reporting, surveillance by institutional bodies, or judicial decision-making). Overall, the interconnection of all key and secondary texts is reflected in their intertextuality as a central feature of “systems of genres” or “interrelated genres that interact with each other in specific settings” (Bazerman 1994Bazerman, Charles 1994 “Systems of Genres and the Enactment of Social Intentions.” In Genre and the New Rhetoric, edited by Aviva Freedman and Peter Medway, 79–101. London: Routledge., 97). Within these systems, some procedures generate “bundles or chains of genres” that are very closely linked to each other (Prieto Ramos 2019aPrieto Ramos, Fernando 2019a “Implications of Text Categorisation for Corpus-Based Legal Translation Research: The Case of International Institutional Settings.” In Research Methods in Legal Translation and Interpreting: Crossing Methodological Boundaries, edited by Łucja Biel, Jan Engberg, M. Rosario Martín Ruano, and Vilelmini Sosoni, 29–47. Abingdon: Routledge. , 42). The main ones found in the LETRINT 0 corpus were those produced to articulate the EU’s legislative procedure and its CJEU’s proceedings, the UN’s human rights treaty body monitoring mechanisms, and the WTO’s TPRs and dispute settlement.

Reports were the most common macrogenre across institutions and also presented the widest variety of functions and subtypes, both as a key genre in monitoring procedures (and also dispute settlement in the case of the WTO)88.In this institution, the terminology coined for the dispute settlement system avoids the connotations of judicial terminology commonly used by adjudicative bodies in national and international jurisdictions (e.g., ‘panel’ is used instead of ‘court’, and ‘report’ instead of ‘judgment’). and as an essential preparatory genre in most legal categories and for internal management. When a particular type of report adhered to conventions determined by specific purposes and authorship (e.g., International Law Commission reports, national reports to UN human rights treaty bodies, WTO Secretariat reports in TPRs, and Appelate Body reports), the subtype (or subgenre) was isolated as such, whereas other miscellaneous groups of reports were classified under the general genre denomination in the relevant legal subcategory. Minutes (with the UN variant of summary records) are also functionally heterogenous in terms of key and secondary functions across institutions, albeit homogeneous in their denomination. Likewise, press releases are found in all the settings but within secondary derived genres in several categories, depending on the main function reported as subject matter, especially at the EU institutions. Notes (and letters in the case of the UN) are also frequently used as instrumental texts within several functional categories.

Tables 1 to 3 in the Appendix list the five largest volumes of translation per setting, legal subcategory, and genre. Within the text production of EU institutions (Appendix Table 1), judgments (11.56 million words translated by the CJEU’s lawyer-linguists) are the most voluminous key genre, followed by regulations (10.16 million words, most often translated at the EC), EP resolutions (10.07 million words), opinions of the Advocate-General (6.57 million words), and decisions within binding law-making (5.66 million words). However, it is reports as input for law-making that constitute the largest volume (almost 40 million words in binding law-making and almost 5 million words in soft law-making and other policy formulation) if secondary functions are considered. Similarly, the amount of minutes translated as instrumental texts in law-making (7.55 million words) stands out in the list of secondary genres and is greater than the volume of minutes translated as a key genre in monitoring procedures (3.25 million words).

At the UN (Appendix Table 2), reports of several subtypes are also the predominant genre. These subtypes include follow-up reports of the Secretary-General (9 million words), reports of UN bodies (8.96 million words) on the implementation of resolutions and UN policies, and reports issued in the framework of human rights treaty compliance monitoring, with the largest volumes being registered by States parties’ reports (7.45 million words) and special rapporteur reports (2.05 million words). Certain features of the latter as well as treaty body concluding observations – and reports in the case of the Working Party on the Universal Periodic Review, albeit not listed among the largest text volumes – partly resemble adjudicative decisions, such as WTO dispute settlement reports. However, despite their quasi-judicial role with regard to specific countries or situations, the enforceability of such UN recommendations is very limited. This also applies to resolutions as the main genre within UN law-making, with more than 1.5 million words. Nonetheless, summary records within secondary functions more than double that amount and constitute the largest genre contributing to law-making. As regards adjudication, the translation volume generated by the ICJ is the most modest of the three adjudicative bodies considered in the study, with judges’ separate or dissenting opinions accounting for more text than judgments.99.In contrast to judges’ opinions, whose translations are systematically indicated as such, ICJ judgments are theoretically co-drafted. In contrast, the UN translation volumes of genres classified under administrative functions, particularly budgets (4.08 million words), are unrivalled within this category across all the settings.

Finally, the most significant volumes of text at the WTO (Appendix Table 3) also reflect the most salient functions in this organization. Notifications (predominantly of a technical nature) to several monitoring bodies have the highest volume (with more than 6 million words), closely followed by dispute settlement panel reports (5.92 million words), and (with comparably lower volumes) Secretariat’s reports on TPRs (3.35 million words) and minutes as a key genre in other monitoring work by several committees (3.12 million words). The same genre also fulfills a supporting function in practically all other legal subcategories, together with reports.