From national to supranational institutionalisation: a microdiachronic study of the post-accession evolution of the Polish Eurolect

ABSTRACT

The objective of this paper is to investigate the evolution of the Polish Eurolect, a supranational translator-mediated variety of legal Polish by: (1) analysing circumstances in which it emerged and took shape; and (2) tracking its development from the formative pre-accession to mature post-accession stage. The first part documents challenges faced at various stages of Eurolect formation, showing its growing institutionalisation as it shifts from the national to supranational level. Using corpus methods, the second part compares the formative stage of the Eurolect (1999–2003) to its mature stage (2011–2015) against the benchmark of the national legal variety. It demonstrates that Eurolects are dynamic and to some extent subject to controlled changes. The post-accession Eurolect is in principle better aligned with the national legal variety, especially for key genre markers (genre fidelity) and thanks to reduced translationese. It is also more formulaic and standardised. Nevertheless, the distance to national usage continues to be large for most patterns. A competing trend of the fossilisation of some over/underrepresented patterns was observed, contributing to the hybridity of the Eurolect. This is accompanied by the levelling out of translations: post-accession regulations and directives are more uniform in style and closer aligned with each other.

KEYWORDS:

• Eurolect
• EU translation
• institutionalisation
• Polish
• legal translation

Eurolects are ‘Europeanized’ constrained varieties of national languages which serve the needs of the European Union (EU) as a supranational organisation and are multilingually mediated through translators and other stakeholders (Biel, 2021, pp. 479). Although as a meta-construct, the Eurolect has in principle 24 parallel ‘mirror’ realisations in the EU official languages, there are important differences in the status and maturity of Eurolects. English and French Eurolects, which represent procedural languages and, hence, main drafting (source) languages for legislative and juridical texts, respectively, exert more influence over passive ‘recipient’ Eurolects. Eurolects range from 10 to 60 years old, depending on a country’s date of accession. The oldest Eurolects are Dutch, French, German and Italian, which started to take shape in the 1950s.

The group of young Eurolects comprises Polish, which gained the status of an official EU language on 1 May 2004. This paper investigates an evolution of its legislative variant. Its objectives are to: (1) reconstruct the circumstances in which the Eurolect emerged; and (2) track its evolution by comparing its formative pre-accession stage (1999–2003) to mature post-accession stage (2011–2015) with corpus methods. This will be done against the benchmark of the national legal variety to measure changes in the textual fit (cf. Chesterman (2004); Biel (2014, p. 118)), that is the distance between the Polish Eurolect and the domestic legislative language.

1. European integration, enlargements and the translation of acquis

Before it became a supranational organisation of 27 member states, the European Union (EU) has undergone a few metamorphoses and expansions since the 1950s, intensifying political, economic and social integration. The process of integration started with the European Coal and Steel Community in 1951, which was transformed into the European Communities in 1957 with six founding countries: Belgium, France, Italy, Luxembourg, the Netherlands, and West Germany. It was replaced by the European Community in 1992 and the European Union in 2009. The integration began with the steel and coal market and has been subsequently broadened to many policy areas, including the internal market, single currency, customs, competition, transport, energy and agriculture (cf. Dedman, 2010).

These transformations were accompanied by 7 territorial enlargements and 1 withdrawal, increasing the number of member states from 6 to 27 and the number of official languages from 4 to 24. The six founding countries were joined by the UK, Ireland and Denmark in 1973, Greece in 1981, Spain and Portugal in 1986, and Austria, Finland and Sweden in 1995. The ‘big bang’ expansion in 2004 admitted Cyprus, the Czech Republic, Estonia, Latvia, Lithuania, Hungary, Malta, Poland, Slovenia, Slovakia, which were followed by Bulgaria and Romania in 2007, and Croatia in 2013. The UK withdrew in 2020.

To access the EU, an applicant country has to respect European democratic values and meet the so-called Copenhagen criteria fixed by the European Council in 1993, which include: (1) political conditions — stable democratic institutions which guarantee democracy, rule of law and respect for human rights; (2) economic conditions — a well-functioning market economy; and (3) legal requirements — the ability to fulfil obligations of EU membership, including adherence to the aims of political, economic and monetary union (cf. Kaczorowska, 2011, pp. 61–62). A criterion of administrative capacity was added in 1995 (cf. Ban, 2013, p. 58). The third criterion entails an obligation to adopt the EU acquis (known earlier as the acquis communautaire), the entire body of EU law which has been passed so far, by harmonising the applicant’s domestic law with the acquis before it accesses the EU. This requirement was set at the very beginning of the expansion process — already with the UK’s accession (Vauchez, 2015, p. 199). It has far-reaching implications for translation as the applicant is obliged to translate the acquis into its official language before the accession and bears the cost of the translation task. The EU institutions are in charge of finalising and publishing translations in the special editions of the Official Journal (OJ). Thus, the translation of the acquis takes place outside the EU institutions and Eurolects start to form at the pre-accession stage.

The acquis is ‘nonnegotiable’ to candidate countries (Ban, 2013, p. 60) and forms the cognitive dimension of the European Union. It is ‘formulated, stretched, criticized, revised, finally naturalized as the most rigorous and objective measure of “Europe”’ (Vauchez, 2015, p. 196). It is not only ‘the instrument that officially defines and authenticates (…) “Europe”’ (Vauchez, 2015, p. 194) but also ‘Europe’s “constitutional operating system […], axiomatic, beyond discussion, above the debate”’ (Weiler, qtd. in Vauchez (2015, p. 194)). It consists of the normative acquis, including primary law (treaties), secondary law (regulations, directives, decisions, etc.), international agreements; the political acquis, comprised of declarations, resolutions and guidelines; and the judicial acquis, i.e., case law (Kaczorowska, 2011, p. 64). It is dynamic in size. At the time when the negotiations for the 2004 accessions started, the size of the acquis to be translated was estimated at 80,000–100,000 OJ pages (Ban, 2013, p. 60). By the time of Croatia’s accession, it had grown to over 160,000 OJ pages¹ (Bratanić & Lončar, 2015, p. 211).

The accession is a complex, lengthy and demanding process. First, the need to translate the acquis is ‘a huge challenge’ (Ban, 2013, p. 59) and ‘no mean feat in itself for small states like Estonia’ (Dedman, 2010, p. 191) nor for large countries like Poland. Secondly and more importantly, candidates often have to carry out numerous reforms to meet the Copenhagen criteria. Poland accessed the EU in 2004, as a culmination of political, economic and social transformations after the fall of Communism in 1989. It filed the membership application in 1994, following the European Agreement, and started the accession negotiations in 1998, ending with the Accession Treaty in 2003 and its ratification by the member states.

2. Circumstances in which the Polish Eurolect has evolved: from national to supranational institutionalisation

While all translations could be regarded as institutional to some extent (Koskinen, 2014, p. 481) and positioned on a cline from low to high institutionality (Koskinen, 2008, p. 22), EU translation is a prototypical example of institutional translation with a high degree of institutionalisation. High institutionalisation happens, argues Koskinen, when we deal with ‘autotranslation’ — ‘self-translation’ — by an institution which ‘uses translation as a means of “speaking” to a particular audience. […] [T]he voice that is to be heard is that of the translating institution. As a result, in a constructivist sense, the institution itself gets translated’ (2008, p. 22, 24). Institutional translation takes place in ‘a controlled and regulated environment’ (Koskinen, 2014, p. 481) — it is governed by institutional norms and standardisation (Wagner et al., 2002, p. 65), with translators being socialised to the organisation and to the profession (Koskinen, 2008, p. 96, 98). These practices enable an institution to ‘govern by translation’ and this function is typically performed through legislative, juridical and administrative instruments (cf. Koskinen, 2014, pp. 481–482), which form the core of the acquis.

This section approaches the evolution of the Polish Eurolect through the lens of institutionalisation. It analyses institutional documents² and other available materials, including the Commission’s regular reports on progress towards accession, memos, Polish government reports, proceedings of relevant parliamentary commissions, press articles and academic literature, to reconstruct circumstances in which the Polish Eurolect evolved. Its evolution is split into three stages: the pre-accession formative phase (mid-1990s–2004), transition phase (2004–2007) and mature phase (from 2008) (cf. Biel, 2018, pp. 296–297).

2.1. Pre-accession translation of the acquis: the formative phase (mid-1990s–2004)

The 2004 candidate countries translated the acquis by themselves. They were provided with some assistance under PHARE and participated in regular joint meetings and trainings organised by TAIEX (the European Commission). All of the candidates established a centralised translation coordination unit. Yet the organisation of translation and its institutionalisation differed. For example, Estonia opted for a fully in-house model. Most of the countries³ used a mixed model with varied degrees of in-house and external translation and revision.

Poland was among those countries which underestimated the challenges of acquis translation and had to adjust its approach due to time pressures. The translation task was coordinated by the Office of the Committee for European Integration (UKIE), a governmental body established in 1996, which created a Translation Coordination Team in 1997. The National Programme of Preparations for Membership set a target of completing over 90% of translations in 2001 and revisions up to 6 months before the accession. In early 1998, Poland was reported to be a leader among the candidate countries in terms of the number of translated documents. Ministries and state agencies were in charge of translating and revising legal acts which fell in their scope of competence. Translations were outsourced to individual translators and revised by in-house ministerial experts (Rzewuska, 2001, p. 72) and next sent to UKIE for a linguistic and legal revision and terminological standardisation. Significant delays were recognised by various bodies in 1999 and 2000, attributed mainly to two recurrent problems — limited funding and understaffing. To speed up the task, the Translation Department was created by UKIE in 2000 and coordinators were appointed by ministries to liaise with UKIE. UKIE formalised a translation procedure, set a schedule and compiled a list of over 90 acquis translators. The initial organisation of translations involved in-house institutional control over the individualised and externalised translation process.

UKIE early recognised the need to standardise translations. The first style guide for English-Polish translators of Community legal acts (KIE, 1997) was prepared in 1997 by UKIE’s law harmonisation team. It had 37 pages. It contained mainly templates of legal acts standardising their macrostructure, recurrent phrases, names of European institutions and editing units. Its 1-page introduction strongly (four times) recommended translators to maintain fluency, clarity and domestication to Polish language conventions and emphasised terminological consistency (KIE, 1997, p. 3). UKIE launched an online database of completed translations in 2000. In respect of terminological resources, UKIE compiled paper glossaries and transferred its terminological resources to an online term base in 2000. The European Commission provided Multiterm and Translator’s Workbench to standardise terminology management and enable term import into its internal resources after the accession (Rzewuska, 2002a, pp. 191–192). The pre-accession stage was critical for EU Polish term formation. Šarčević refers to this process as ‘a Herculean undertaking’ due to the need to create the entire terminology of EU law in the national language over a relatively short period of time (2015, p. 183).

After the new organisational measures were put in place, the number of translated and verified pages increased by ca. 50% within a year, but was still below the target. It became clear that the translation task will not be completed by 2001. UKIE shifted from individual translators and in-house revisions to public tenders for translation agencies, externalising and anonymising the translation and revision process and hence losing some control over it. It organised a few small tenders in 2001 and awarded contracts to four translation agencies. Yet as at September 2002, Poland had only 8,200 OJ pages of revised text registered in the Commission database and 68,400 pages of unrevised text. Among the candidate countries, this low number was comparable to Slovenia, Lithuania and Malta with 7,300, 6,000 and 5,300 revised pages, respectively. At the other extreme were the Czech Republic, Estonia, Hungary, Latvia and Slovakia, which made significant progress and had 30,000–40,000 revised pages. To eliminate the backlog, UKIE organised a huge tender in 2002 to translate and revise 27,000 OJ pages (1/3 of the acquis) and to revise 37,000 OJ pages, dubbed by the press as ‘the tender of the century’ (Niklewicz, 2002). Reputable translation agencies lodged formal complaints against unrealistic and unprofessional tender specifications prepared by UKIE, criticising excessively tight deadlines (170 days for translation and 140 for revision), missing information as to exact volumes of source texts and breakdown into source languages (Niklewicz, 2002). Polish newspapers reported irregularities in public procurements, overreliance on least price selection, award of contract to an agency without a domestic track record, engagement of inexperienced translators, and a low quality of translations. On the one hand, the tender reduced the backlog since the Commission’s report for 2003 singles only Slovenia, Lithuania and Malta as countries with substantial delays. Yet, as a result of such mismanaged tenders, 80% of translations were reported to have been returned due to low quality and only 80 regulations were ready in May 2003 to be printed in the OJ (Rosińska, 2003). Half a year after the accession, UKIE sent 4,500-page corrections of translations (ca. 5% of the acquis) to the EU institutions (Uhlig, 2005).

Another shortcoming was revision on both the national and EU side. The two-step in-house revision procedure introduced early by UKIE did not always work due to the understaffing. UKIE hired more revisers but admitted that it allowed for only random checks of outsourced translations and revisions. Secondly, translations were initially planned to be ‘revised’ in-house by EU institutions, a term which was later replaced by ‘finalisation’, used interchangeably with ‘document management’ (Rzewuska, 2002b, p. 145). The European Commission and the Council employed only 5 lawyer linguists per new language to legally revise translations, which means that a lawyer linguist had to check about 50 OJ pages (ca. 85 standard manuscript pages) per day (Rzewuska, 2002b, p. 145). Translations were split into three groups according to their importance and the acquis of lowest importance was most probably published without any checks (Rzewuska, 2002b, p. 146).

In January 2005 — 8 months after the accession, 97% of the priority acquis documents were finalised in the new languages. This contributed to significant delays in the publication of translations in the OJ,⁴ which was completed in 2006 (Bobek, 2007, p. 44).

2.2 The transition phase: the European institutions take over (2004–2007)

Upon the accession, Polish has become an official language and the European institutions took over full responsibility for translations. Following Council Regulation No 1 of 1958, EU institutions ensure that EU legislation is published in the OJ in the official languages. In contrast to the pre-accession period when translators worked on finalised source texts, post-accession translations are part of a complex multistage and multilingual drafting process with unstable, dynamic source texts (cf. Felici, 2010, p. 102).

The institutionalisation of translations at the EU level was gradual and required a running-in period. The institutions established Polish language units in their translation services. First in-house translators with Polish were already recruited in 2003. The recruitment was based on open competitions and translators were assigned to areas of specialisation. All institutions encountered significant delays and failed to meet their recruitment targets for the new EU-10 languages (CA, 2006). As at 1 January 2005, the Commission’s DGT (Directorate General for Translation) recruited from 22 to 38 translators per language. By April 2006, the DGT recruited 94% of the translators needed (473 full-time translators, of which 58 Polish translators); however, two-thirds of them were on temporary contracts (in the case of the Polish unit – about 50%). Only 27 out of 126 projected support staff were hired and the recruitment of management was still ongoing. The problems were caused by the ‘late availability of competition lists, reluctance of successful applicants to take up jobs, resignations and lack of a translation market in some languages’.⁵ Staffing problems were also reported by the Parliament. To cope with in-house translator shortages, EU-10 translations, except for legislative texts, were outsourced to external contractors — for example, 60% of translations in the Commission and 20% in the Parliament in 2004 and 37% and 43%, respectively, in 2005 (CA, 2006). During the transition period the quality of EU-10 translations was regarded as unsatisfactory and considerably lower than that of EU-15, mainly due to the use of external translators and insufficient internal capacity to revise translations (CA, 2006, p. 10, 19). In the case of Polish, the Commission imposed financial penalties for poor quality in 4.4% of assignments in 2005.⁶ Another solution was to establish limits on documents to be translated by the Commission’s DGT and introduce derogations and changes in the Rules of Procedure of the Parliament (Judge & Earnshaw, 2008, p. 156) and of the Court of Justice, with the latter also setting up the pivot translation system (McAuliffe, 2008, p. 810). The recruited translators were trained in the procedural languages, IT, translation technologies and field-specific knowledge. They were gradually socialised into their organisations and the profession.

EU institutions started to build their own standardisation resources, including style guides, term bases and IT tools. The DGT’s work on the Polish style guide and consultations with the Polish Language Council (RJP) commenced in 2005 and the first edition of Vademecum tłumacza was launched in 2007.⁷ The IATE term base was fed with acquis terms and since they were regarded as unreliable and inconsistent they were marked as Pre-IATE (Bhreathnach et al., 2013, p. 48). In 2005, Polish, as other new languages, had a low number of terminological entries in IATE (ca. 21,000), a sharp contrast to English and French with ca. 1.5 million entries (Bhreathnach et al., 2013, p. 22). Other knowledge-sharing resources included in-house glossaries and translation memories.

2.3 Post-accession translation of the acquis: the mature stage (2008- …)

The transition stage may be regarded to have ended by 2008, with fully operational Polish language units and reliable resources for translators. This year tends to be adopted internally as a landmark in terms of quality and reliability of translation memory suggestions within the DGT’s Polish Language Unit.

In the mature period efforts have been made to improve, update and integrate both general and language-specific technological, terminological and linguistic resources. The DGT regularly updates the style guide Vademecum tłumacza, which has its 19th edition in 2021 and has doubled in size since the first edition. Language units differ as to the degree of control through translation manuals, style guides and other resources of more or less pressing nature, that is, decrees, rules, requirements, recommendations, tips, advice (Svoboda, 2017, p. 96). For example, both Swedish and Lithuanian have as many as ca. 80 resources, while Hungarian, Dutch and Polish have fewer than 20 (Svoboda, 2017, pp. 83–84). In respect of terminology, IATE was regularly corrected, updated and populated with new terms; its Polish term entries increased to ca. 50,000 in 2012 (Bhreathnach et al., 2013, p. 50), growing to 103,345 in early 2021.⁸ Other developments included setting up the Polish Terminology Network for Commission translators, comprising 23 ministries and public institutions and coordinated by the Ministry of Foreign Affairs to consult terminology with national authorities (Bhreathnach et al., 2013, p. 46). A more active use of corrigenda to eliminate translation errors and standardise terminology has been found for the post-accession period (Biel & Pytel, 2021).

Recent years have observed a new trend of increased outsourcing: DGT outsourced 27% of translations in 2015, reaching 31% in 2020 against the planned target of 37% (DGT, 2016, p. 11; DGT, 2021, p. 13). It considerably externalises the translation process; however, unlike in the pre-accession period, it is commissioned, controlled and verified by the EU institutions.

Finally, the Polish Eurolect was affected not only by the growing institutionalisation but also general changes in the EU drafting style. There was a shift, as observed by Robinson, ‘from the “fuzzy” or general-principles style familiar to many countries of continental Europe towards the ‘fussy’ or more detailed drafting style familiar to common lawyers’ (2017, p. 247). This applies in particular to enacting terms, which have become more precise and detailed and contain more definitions (Robinson, 2017, p. 247). The growing length and complexity of EU legal acts (Robertson, 2015) has been accompanied by attempts at increasing the reader-friendliness and clarity of legislation (Robinson, 2017, pp. 247–248).

To sum up, the circumstances in which the Polish Eurolect evolved attest to the growing institutionalisation, shifting from the national to supranational level. In the pre-accession phase EU institutions use national institutions to ‘govern by translation’ with limited control over translations. The Polish authorities did not have experience in managing such a massive volume of translations and in consequence they underrated, underfunded and mismanaged this task to some extent. This has created adverse circumstances for the formation – especially at a later pre-accession phase — of the Polish Eurolect. A corpus study by Biel (2014) on the Eurolect up to 2006 shows a divergent textual fit to the national variety at many levels. The formative phase exerts impact on its later stages due to a frequent recycling of legal texts and the need to maintain continuity with previous translations (Stefaniak, 2017). Only in the post-accession mature period translation becomes fully institutionalised as self-translation by EU institutions. The next sections analyse quantitatively how this shift influenced the evolution of the Eurolect. Koskinen has proposed the law of growing institutionalisation whereby the institutional impact grows with the in-house processing of texts, which further diverge from ‘natural’ language, combined with the competing trend of added readability (2008, p. 141). It is hypothesised that despite the growing institutionalisation of the translation process and thanks to its professionalisation, the post-accession Polish Eurolect will be better aligned with the national variety of legal Polish.

3. Corpus design

The study uses micro-diachronic comparable corpora of Polish translations of EU legal acts, designed according to the 5-year frame: a pre-accession corpus dated 1999–2003 and a post-accession corpus of 2011–2015 (Table 1). The corpora comprise two types of EU legal acts: regulations, which are self-executing and directly binding, and directives, which need to be transposed into national law. The Polish pre-accession corpus was extracted from Version 3.0 of the JRC Acquis (Steinberger et al., 2006) while the post-accession corpus was compiled under the Polish Eurolect project.⁹ Following Piehl’s idea (2006), only enacting terms were extracted to exclude non-normative preambles and technical annexes and, hence, increase the comparability to Polish legal acts. The 2011–2015 Polish Law corpus compiled as part of the Polish Eurolect project served as a reference corpus and some trends were verified on the corresponding EU English corpora. The corpora were uploaded to Sketch Engine (Kilgarriff et al., 2014), where they were POS-tagged and lemmatised.

Table 1. Details of the corpora.

Enacting terms	# of texts	Tokens	Lexicon size: words	Lexicon size: lemma_lc
Focus corpus 1: pre-accession law 1993–2003
EU Regulations (R93-03)	908	1409869	46865	17424
EU Directives (D93-03)	185	607309	28356	9656
Focus corpus 2: post-accession law 2011–2015
EU Regulations (R11-15)	925	2193223	46723	14736
EU Directives (D11-15)	92	883499	27460	8521
Reference corpus: national law 2011–2015
Polish Law Corpus (PLC)	144	2199660	44237	14589

The analysis is based on keywords, wordlists and n-grams. The keyword extraction procedure was two-fold: first, each pre-accession corpus was compared to its post-accession corpus to identify changes in the Polish Eurolect. Second, both pre-accession and post-accession corpora were compared separately to the PLC reference corpus to identify their fit to Polish law. The full data set is available at REPOD, doi:10.18150/LCUVJ5.

4. Changes in the post-accession Polish Eurolect

Changes identified through corpus methods were grouped into: (1) formulaicity; (2) parts of speech; (3) key legislative genre markers; and (4) other features of translated Polish texts.

4.1 Growing formulaicity: lexicon size and n-grams

As already noted, the length of EU legal acts was reported to increase and this is confirmed by Table 1 below. The enacting parts increased on average from 1846 words per instrument in 1999–2003 to 3025 in 2011–2015, but are still 5 times shorter than the Polish legal acts. Despite an increase of tokens by +56% for post-accession regulations and +45% for directives, their lexicon size (lemmas) decreased by −15% and −12%, respectively. The corresponding English instruments (see the full data set) show a nearly identical growth of tokens but a noticeably lower change in lemmas: +1% for regulations and −6% for directives. The decrease of lemmas means that both English and Polish instruments use a smaller range of words and are significantly less lexically rich (varied). It may be due to a higher thematic uniformity of texts with their increasing length. Yet a markedly higher decrease of lemmas in Polish translations suggests their reduced variation.

The next evidence of the growing formulaicity of translations is offered by n-grams — repetitive sequences of words (Table 2). While both pre-accession regulations and directives had fewer n-grams than Polish law, they now have more n-grams in the majority of categories. Both the number of n-grams (2–6 n-grams, 3- and 4- grams separately; cut-off threshold 40 per million words (pmw)) and their total frequency increased markedly for post-accession Polish translations, significantly more for regulations (31–102%) than for directives (12–26%), which were better aligned with PLC in the pre-accession period. The change is less pronounced in the English corpora, where regulations observed a growth of 0–27% while directives mainly noted a decrease of n-gram frequencies (although their numbers slightly grew).

Table 2. N-grams in the Polish corpora (frequencies normalised pmw).

	R98-03	R11-15	R %DIFF	%DIFF to PLC	D98-03	D11-15	D %DIFF	%DIFF to PLC	PLC
2–6 grams: items	2000	2622	31	18	2708	3201	18	44	2222
2–6 grams: frequency	295522	395147	34	19	340176	379858	12	14	333094
3-grams: items	448	602	34	26	645	759	18	58	479
3-grams: frequency	59208	85114	44	12	72829	83596	15	10	76220
4-grams: items	147	297	102	36	259	326	26	49	219
4 grams: frequency	22966	44891	95	28	26466	32449	23	−8	35187

Note: For clarity of presentation, tables show only normalised frequencies pmw; raw frequencies are in the full data set.

Another interesting change is the growing similarity between Polish regulations and directives as regards the distribution of n-grams in the post-accession period compared to the pre-accession period, when differences were more pronounced (also compared to the English versions). Both the lower lexicon size and the higher number and frequency of n-grams are indicative of the reduced variation (idiosyncrasy), higher formulaicity and standardisation of post-accession translations.

4.2 Parts of speech: overview of changes

The most common parts of speech (PoS), with a frequency of over 100,000 pmw (that is nouns, adjectives, simple prepositions and verbs) observed a relatively small (up to 12%) change. More profound changes were seen for mid-frequency parts of speech and subcategories which have a distribution of over 10,000 pmw in at least one of the corpora – e.g., conjunctions, gerunds (deverbial nouns), adverbs, past participles, active adjective participles, and infinitives (see Table 3). These changes resulted in a closer similarity of directives and regulations, i.e., their levelling-out.

Table 3. Changes in the distribution of key parts of speech.

	R98-03	R11-15	R %DIFF	%DIFF to PLC	D98-03	D11-15	D %DIFF	%DIFF to PLC	PLC
nouns	334394	348706	4	−5	334869	342032	2	−7	366248
deverbial nouns	24771	30088	21	−11	30128	32386	7	−5	33938
adjectives	153583	164178	7	6	168583	166920	−1	7	155574
prepositions	117777	120029	2	−3	114716	114686	0	−7	123985
verbs	112878	99299	−12	27	116130	109066	−6	40	78092
past participles	42470	31180	−27	34	38638	30426	−21	31	23271
active adjective participles	12807	15587	22	20	13752	16995	24	31	13019
infinitives	9144	8053	−12	39	11518	10172	−12	75	5797
conjunctions	45914	52665	15	23	56332	59399	5	39	42675
adverbs	17368	15568	−10	97	11518	16215	41	105	7900

Although post-accession instruments have become better aligned with the Polish usage as regards the key noun and verb groups, they still use fewer nouns and significantly more verbs (i.e., the latter 27% for regulations and 40% for directives). Thus, the post-accession Eurolect continues to have a more verbal style compared to the Polish administrative usage. It is more typical of English and may have been caused by interference or EU clarity-oriented efforts into making texts more dynamic.

Overall, the fast rate of change is surprising, given that the grammatical layer of language changes slowly. These changes are not always correlated with the initial distance to national usage: for example, categories which are strongly overrepresented in the Eurolect have observed a decrease (e.g., infinitive, past participle) or an increase (conjunctions or adverbs in directives).

4.3 Key legislative genre features

This section focuses on selected keywords and salient genre markers identified in earlier studies on Eurolects (Biel (2014), Mori (2018)): deontic modality, hypothetical clauses and textual mapping, to investigate the stability of key genre features.

Deontic modality serves the fundamental governing function of legislation by imposing obligations on and granting rights to the governed. Deontic modality can be expressed through a range of means: the present tense (Polish), the coloured future tense (English), modal verbs and their phraseological substitutes. Each legal language has its distributional preferences for patterns recognisable to their speech communities and previous studies have shown that these preferences differ in the Eurolect (cf. Biel (2014), Mori (2018)).

The most frequent way of expressing deontic modality in legal Polish is the Present Tense (Table 4). This pattern is overrepresented in the Eurolect by ca. 30% and is stable over time. The post-accession Eurolect uses markedly fewer future tense verbs (ca. −50%), which aligns closer with Polish law. This form was excessively prompted by shall, recommended to be translated as będzie ‘will’/‘shall’ in the first style guide.

Table 4. Deontic modality.

	R98-03	R11-15	R %DIFF	%DIFF to PLC	D98-03	D11-15	D %DIFF	%DIFF to PLC	PLC
Verbs
Present tense	32485	32498	0	28	32036	34422	7	36	25393
Future tense	3201	1717	−46	15	5370	2029	−62	36	1494
Obligation modals
musi ‘must’	932	558	−40	1168	1139	892	−22	1927	44
należy ‘should’-imprs	481	328	−32	−1	594	326	−45	−2	332
powinien ‘should’	403	119	−70	−55	761	131	−83	−51	265
jest/są obowiązan* ‘is/are obliged’	9	2	−78	−100	16	0	−100	−100	680
Permission modals
można | może ‘may’	4056	3665	−10	41	4846	4197	−13	61	2606
jest/są uprawnion* ‘is/are empowered’	92	406	341	534	86	341	297	433	64
Prohibition modals
nie można | może ‘may not’	924	517	−44	−2	678	786	16	50	525

Interestingly, key modal verbs show a high rate of change (decrease), resulting in most cases in a better alignment with national usage and the levelling out between directives and regulations. In particular, it applies to impersonal należy and personal powinien ‘should’. Although the use of musi ‘must’ has decreased, the difference to Polish law — where it is barely used — is still vast. This modal is recommended in the Polish style guide as one of the equivalents of shall, which promotes this form contrary to the national usage. Instead, Polish law prefers a phraseological substitute jest/są obowiązan* ‘is/are obliged’, which is not prompted in either the pre-accession or post-accession Eurolect. Likewise, despite their decrease, permission modals continue to be strongly overrepresented in the Eurolect. A new salient phraseological permission pattern has emerged: jest/są uprawnion*, calqued from English is/are empowered. Tellingly, it is used predominantly with the Commission as the subject in the context of its right to adopt legal measures, which – unlike może ‘may’ – strengthens the entitlement. The stronger coding of authority corroborates Koźbiał’s (2020) findings on the overrepresentation of high-level markers of epistemic modality in the Polish judicial Eurolect to raise the authoritativeness of judgments and their rhetorical force.

Hypothetical patterns which form legal rules are mainly realised in Polish through the conjunction jeżeli ‘if’. Its usage grew in the post-accession Eurolect; however, the synonymous, more informal conjunction jeśli, despite a sharp decrease, is still 30–40 times more common in the Eurolect (Table 5).

Table 5. Hypothetical patterns.

	R98-03	R11-15	R %DIFF	%DIFF to PLC	D98-03	D11-15	D %DIFF	%DIFF to PLC	PLC
jeżeli ‘if’-form	1775	1811	2	−21	1796	2158	20	−6	2286
jeśli ‘if’-neut	784	451	−42	4000	1120	358	−68	3155	11

Another salient genre feature is a frequent use of intratextual and intertextual patterns which refer to specific points in legal acts. This function is realised in Polish law through the extremely repetitive pattern o któr* mowa w (art./ust. x), lit. ‘about which (it is) spoken in (Article/Paragraph x)’, with a frequency of 8840 pmw. This pattern is not directly primed by English and the pre-accession Eurolect had other preferences (patterns I.2-5 in Table 6) calqued from English and far less frequent in the domestic usage. These strongly overrepresented patterns were markedly reduced in the post-accession Eurolect as a result of the growing use of o któr* mowa w art./ust., which increased by 504% in regulations and 443% in directives. Despite this exponential growth, it is still twice less frequent than in PLC. Although all the patterns are now significantly better aligned with domestic legal Polish, they remain strongly underrepresented (I.1) or overrepresented (I.2-5) in the Eurolect.

Table 6. Textual navigation patterns.

	R98-03	R11-15	R %DIFF	%DIFF to PLC	D98-03	D11-15	D %DIFF	%DIFF to PLC	PLC
I. Intratextual and intertextual referencing patterns
1. o któr* mowa w ‘about which (it is) spoken in’	782	4726	504	−47	739	4014	443	−55	8840
2. określon* w ‘specified in’	5094	3317	−35	73	4469	2647	−41	38	1921
3. przewidzian* w ‘envisaged in’	1131	533	−53	461	1001	535	−47	463	95
4. ustanowion* w ‘established in’	930	259	−72	1339	1228	210	−83	1067	18
5. wymienion* w ‘mentioned in’	1124	741	−34	176	1093	346	−68	29	268
II. Authority and conflict markers
6. zgodnie z ‘according to’	4384	5354	22	541	4923	4684	−5	461	835
7. na podstawie ‘on the basis of’	1429	1349	−6	−25	990	1109	12	−38	1800
8. na mocy ‘∼pursuant to’	1056	1154	9	4516	700	1139	63	4456	25
9. bez uszczerbku dla ‘without prejudice to’	422	375	−11	18650	591	581	−2	28950	2
10. z zastrzeżeniem ‘∼subject to’	238	214	−10	33	331	229	−31	42	161
III. Adjectival deictic expressions
Navigating deixis	7143	7402	4	424	7439	6862	−8	385	1413
niniejszy ‘this’-form	5529	5599	1	717	6109	5209	−15	660	685
odnośny ‘respective’	153	259	69	64650	199	250	26	62400	0.4

The lower use of o którym mowa w is also compensated by the use of complex prepositions signalling legal authority or conflict (II.6-11). Authority markers are 2.5–3 times more frequent in the Eurolect than in PLC, which is also indicative of the EU drafting style, where an increased foregrounding of legitimisation is observed. The rate of change is in most cases rather small and not symmetrical between regulations and directives, although these complex prepositions are distributed relatively uniformly in both the Eurolect corpora, with marked differences as to preferences compared to PLC. The most strongly overrepresented zgodnie z (ca. 6 times more frequent in the Eurolect), triggered by in accordance with, has undergone fossilisation, increasing its use in regulations by 22%. Similarly, bez uszczerbku dla, a calque of without prejudice to and an expression that is not used in legal Polish, has slightly decreased but fossilised at this distribution level. The fossilisation trend has also been observed for navigating deictic expressions which are used for anaphroric/cataphoric reference, mainly pointing to ‘this instrument’, which remain to be strongly overrepresented in the Eurolect.

Key genre markers have mostly become more closely aligned with the domestic Polish usage, improving the ‘genre fidelity’ (cf. James, 1989) of the Eurolect. However, they still show substantial differences to PLC.

4.4 Other features of translated texts

The final section discusses strong areas of overrepresentation and underrepresentation identified through keywords and in the previous studies (Biel, 2014) as typical features of Polish translations.

Another example of fossilisation is offered by deictic patterns (Table 7), which are among most strongly overrepresented features of the Eurolect as a result of interference from English, which explicitly codes (in)definiteness. Demonstrative pronouns (e.g., taki ‘such’, ten ‘this’) and personal pronouns observed a 7–26% increase despite their strong overrepresentation in the pre-accession Eurolect. Indefinite pronouns decreased moderately, although some patterns grew strongly, e.g., wszelki ‘any’, contrary to their rare use in legal Polish. These areas of interference are not addressed in the style guide. In most cases, the changes result in the levelling out of the distribution of high-frequency patterns in regulations and directives.

Table 7. Deixis.

	R98-03	R11-15	R %DIFF	%DIFF to PLC	D98-03	D11-15	D %DIFF	%DIFF to PLC	PLC
Demonstrative pronouns	9645	12122	26	73	12907	13774	7	96	6417
Personal pronouns	6700	7686	15	43	8073	9289	15	73	5378
Indefinite pro-nouns/quantifiers	6817	5715	−16	425	6527	6083	−7	458	1089
każdy ‘each’	2937	1676	−43	239	2131	1675	−21	239	494
wszelki ‘all’	930	1198	29	1246	1393	1722	24	1835	89
wszystek ‘all’	1781	1698	−5	613	1395	1319	−5	454	239
jakikolwiek ‘any’	415	426	3	1274	598	502	−16	1519	31

A similar fossilisation was observed for some strongly overrepresented anchoring and particularisation patterns, in particular: complex prepositions w odniesieniu do ‘with reference to’ and w ramach ‘within the framework of’ and the adjective participle dotyczący ‘concerning’ (I, Table 8). Their distribution grew markedly in the post-accession Eurolect despite some discouragements in the style guide (cf. also Wasilewska (2021) on the limited impact of the style guide as regards complex prepositions). They increase the structural analyticity of translated language.

Table 8. Other areas of interference.

	R98-03	R11-15	R %DIFF	%DIFF to PLC	D98-03	D11-15	D %DIFF	%DIFF to PLC	PLC
I. Anchoring and particularisation patterns
complex prepositions	5086	6889	35	18	4679	6296	35	7	5859
w odniesieniu do ‘with reference to’	1804	2159	20	718	1422	1496	5	467	264
w ramach ‘within the framework of’	1010	1351	34	160	420	844	101	63	519
dotyczący ‘concerning’	2083	3328	60	179	2509	2935	17	146	1192
II. Purpose subordinators
aby ‘that’-rel	533	962	80	892	1307	2191	68	2159	97
by ‘that’-rel	81	317	291	10467	323	1274	294	42367	3
III. Other subordinators
że ‘that’-rel	2355	1923	−18	94	3463	2658	−23	169	989
który (excl. o którym mowa w) ‘which’	7677	7167	−7	37	8661	8548	−1	64	5222
gdzie ‘where’	406	72	−82	148	568	82	−86	183	29
gdy ‘when’	2097	1688	−20	69	1968	2723	38	173	996
IV. Depersonalisation
passive voice	11890	6622	−44	67	11454	7236	−37	82	3967
impersonal past-tense no/to pattern	1081	1094	1	111	920	956	4	84	519
V. Stigmatised patterns
i/lub ‘and/or’	359	5	−99	0	540	6	−99	0	0
*zapewni*, że ‘ensure that’	233	13	−94	44	955	16	−98	78	9

The frequent use of purpose patterns is a distinctive feature of EU drafting style that is connected with its focus on legitimisation and the main function of directives — specifying the purpose to be achieved in national transposing measures. The subordinator aby and its informal variant by (II, Table 8) are over 13 times more frequent in regulations and 24 times in directives. Both grew considerably in the post-accession Eurolect, mirroring source text patterns.

On the other hand, the growingly convergent textual fit was observed for overrepresented general subordinators (III, Table 8) – który, że, gdzie, gdy, which decreased (except for gdy in directives). Similarly, the passive voice (Table 8, IV), which is much less common in Polish and has been stigmatised in the style guide, decreased sharply by ca. 40% in the post-accession Eurolect. An impersonal (past tense) structure, which was strongly overrepresented in translations (-no/to) but not stigmatised in the style guide, remained at a similar level. Despite the growing fit of subordinators and the passive voice, they are still strongly overrepresented compared to the national variety.

Although the assessment of translation quality was beyond the ambitions of this study, the aforementioned changes which increase the fit of the Eurolect to the domestic legal variety attest to the improvement of quality in respect of the naturalness of translations. Reduced interference was observed at a number of levels, including key genre features. Since these changes mostly concern high-frequency patterns, they are likely to reduce the flavour of a variety perceived as translationese. Other linguistically-oriented improvements of quality include: (1) markedly fewer calques of punctuation (especially commas) from English, comprehensively covered in the style guide, e.g., after sentence-initial discourse markers; (2) reduction of excessive capitalisation, in particular the very common Państwa Członkowskie ‘Member States’; (3) elimination of incorrect variants, such as stigmatised variants of complex prepositions za wyjątkiem ‘except for’, odnośnie x ‘with reference to x’ and w porównaniu do ‘in comparision to’, with their marginal incidence being now similar to Polish instruments, as well as of a non-existent and tautological conjunction i/lub ‘and/or’ and pattern zapewnić, że ‘ensure that’ due to their stigmatisation in the style guide (cf. V, Table 8). Finally, during the extraction of enacting terms from corpus files, some major information transfer errors were identified in the pre-accession corpus: the enacting clause mentioned the wrong type of instrument in a few regulations: przyjmuje(-ą) niniejszą decyzję/dyrektywę ‘has/have adopted this decision/directive’. Such errors were not identified in post-accession instruments, probably due to the use of normative translation memory. Thus, without passing any judgments on the accuracy and adequacy of translations, it may be concluded that language-related aspects of quality have mostly improved in the post-accession Eurolect.

Conclusions

The first part of the paper has reconstructed the circumstances surrounding the emergence and formation of the Polish Eurolect, which transitioned from national to supranational institutionalisation. It contextualises the corpus study of Eurolect evolution and raises awareness of challenges faced at the pre-accession stage.

The second part evidences that Eurolects are dynamic constructs, which require time to stabilise and can be moulded to some extent through institutional style guides, with some changes being profound, even for high-frequency patterns. The study provides descriptive quantitative data on the formative and mature stage of the Polish Eurolect compared to its pre-accession stage. In principle, the post-accession Eurolect shows a growing alignment with national usage, which is particularly visible for key genre markers and some typical translationese features, and hence improves its ‘genre fidelity’. This trend is consistent with changes reported elsewhere for the terminological layer, where corrigenda often domesticated terminology in post-accession regulations (Biel & Pytel, 2021). The alignment seems to be closer for regulations, which is a desirable trend, considering that they automatically become part of national legal orders upon their authentication. Additionally, some improvements in the linguistic aspects of translation quality were observed and the post-accession Eurolect has been found to be more standardised, formulaic and less lexically rich (varied). It could be argued that the growing fit has also been affected by the Europeanisation of the national variety but, as shown by earlier studies (e.g., Biel (2014)) and as reconfirmed for this data set, except for the terminological layer, legislative Polish has been marginally affected by the Eurolect, especially as regards genre markers.

Nevertheless, despite these changes which increase the growing convergence of the textual fit with the Eurolect, the distance to the domestic variety is still large for most patterns. At the same time, there is a competing trend where some over/underrepresented patterns fossilise and/or further depart from national usage. This mainly happens due to interference when it is not addressed in the style guide and when these patterns serve EU-specific discursive practices, e.g., legitimisation or authority markers. These factors contribute to the development of a hybrid EU sub-genre of legal acts. It is combined with the levelling out of translations: regulations and directives have become more uniform in style and tend to be more closely aligned with each other in the post-accession Eurolect. This within-genre levelling-out is consistent with the across-genre levelling-out confirmed for four genres of the Polish Eurolect in an earlier study (Biel et al., 2019).

To the best of my knowledge, this is a first study into the evolution of a Eurolect. While the findings cannot be simply generalised to all new Eurolects due to different circumstances in which they might have evolved, it may be expected that at least some of these trends are shared. Further studies could cover other new Eurolects to verify whether the growing post-accession institutionalisation contributed to an increased convergence in terms of textual fit and to offer a comparative perspective on the Eurolects.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by the Narodowe Centrum Nauki (National Science Centre) under [grant number 2014/14/E/HS2/00782].

Notes on contributors

Łucja Biel

Łucja Biel is a translation scholar, a corpus linguist and a sworn translator. She is Associate Professor and Head of EUMultiLingua research group in the Institute of Applied Linguistics, University of Warsaw, Poland. She is an editor of the Journal of Specialised Translation and a board member of the European Society for Translation Studies. She holds an MA in Translation Studies (Jagiellonian University of Kraków), PhD in Linguistics (University of Gdańsk), and Diploma in English and EU Law (University of Cambridge). She has published extensively on legal/EU translation, legal terminology, translator training and corpus linguistics.

Notes

1 Since the 2004 accessions faced serious translation delays due to the massive size of the acquis, the Commission set an objective to reduce it to facilitate future enlargements, an objective which clearly was not achieved.

2 A full list of resources consulted is available at https://repod.icm.edu.pl/dataset.xhtml?persistentId=doi:10.18150/LCUVJ5.

3 See e.g. Šarčević (2015), who discusses the Czech and Hungarian experience.

4 This gave rise to disputes contesting whether provisions which were not translated and/or published in the candidates’ languages may be binding on them after the accession (Bobek, 2007).

5 Communication to the Commission. Translation in the Commission, 20.12.2006, SEC(2006) 1489 final.

6 38 out of 857 translated documents; e-mail communication with the Commission, 31.10.2006.

7 E-mail communication with Karolina Stefaniak from the DGT’s Polish Language Unit, 07.2017.

8 E-mail communication with the Terminology Coordination Unit, European Parliament, 8.02.2021.

9 https://eurolekt.ils.uw.edu.pl/pl-eurolect-corpus/. The EU corpora were manually sorted to exclude amending, repealing and implementing acts.