Language policy and the law

Universities in the Netherlands are currently faced with finding a balance between the implementation of English-medium instruction and the protection and promotion of Dutch. In this article I analyse university language policy documents from a discursive and critical perspective. I explore the intertextual transformations involved in a multilevel process of policymaking; that is, as policy discourse shifts from the state legislation governing the language of instruction in higher education to the codes of conduct for language of the publicly funded universities. The institutions use various discursive strategies, including intertextuality and recontextualisation, to legitimate their reinterpretation of the basic legal principle ‘Dutch, unless’ as ‘English, unless’ (at master’s level, and increasingly at bachelor’s level too). Although the current law is set to be amended, it appears the proposed new law will simply require universities to do more paperwork while continuing on their current path.


Introduction
Globalised universities today have to walk the tightrope between being at once 'fundamentally international and essentially national' (Saarinen, 2014, p. 127). One the one hand, they are expected to operate and compete on an international level, increasingly in line with neoliberal economic values (Piller & Cho, 2013). On the other hand, they are popularly viewed as bastions of the national culture and language, which they are obliged to protect and develop (Cots, Lasagabaster, & Garrett, 2012). In the Netherlands, we find an interesting case of a university system seeking to balance these two, often contradictory pressures.
In the context of the Bologna Agreement and the creation of the European Higher Education Area, Dutch universities have in recent decades pursued an intensive internationalisation process. In the decade from 2007 the proportion of international students doubled to account for almost one quarter of all university students (Inspectie van het Onderwijs, 2018a, p. 175). Both a cause and an effect of this growth is the rise of English-medium instruction (EMI): study programmes in which the content is taught in English. Outside the UK, the Netherlands is Europe's leading provider of English-medium higher education (Maiworm & Wächter, 2014). In the academic year 2017/18, 23% of bachelor's degrees and 74% of master's degrees were English-taught (VSNU, 2018).
This development towards EMI has triggered a backlash in some sections of the media and public opinion. In May 2018 Beter Onderwijs Nederland, an organisation that campaigns for Dutch-language education, sued Maastricht and Twente universities for offering English-language degrees in contravention of the law. The judge ultimately decided in favour of the universities (Beter Onderwijs Nederland vs. Maastricht University/University of Twente, 2018). In March 2019, 183 representatives of the cultural sector and academia -including 93 full professors -signed a petition calling on the government to halt the verengelsing (' Anglicisation') of universities (De Groot, Jurgens, & Verbrugge, 2019); just one of several such calls in recent years. There is a perception that the rise in international students has left universities at breaking point and reduces the accessibility of higher education for Dutch students (Van Engelshoven, 2018, p. 2). Opponents of EMI often express concern that it impedes knowledge transfer, although this has not been conclusively demonstrated. 1 In addition, there are concerns that EMI may be detrimental to the development of high-level literacy in Dutch, when the majority of graduates will work in the domestic labour market (Inspectie van het Onderwijs, 2018a, p. 196). Although the evidence for this is anecdotal, the public concerns have not gone unnoticed by politicians and lawmakers. As noted by the Inspectorate of Education, '[w]here the advantages and the necessity of internationalisation were initially dominant, in the last few years there have been calls to 1. De Vos, Schriefers and Lemhöfer (fc.) found that Dutch first-year psychology students studying in Dutch received slightly higher grades than Dutch and German students studying the same content in English. In contrast, De Jong (2018) found no negative effects of EMI on the acquisition of knowledge and skills of Dutch students of an International Business and Management Studies programme. Similarly, figures from the National Student Survey 2017 show that students studying in Dutch and English rate the quality of their study programmes equally positively (4.1 on a 5 point scale) . More research is needed using comparable methods and a longitudinal study design. contain and safeguard internationalisation' (Inspectie van het Onderwijs, 2018b, p. 10; my translation). 2 The political importance of championing the Dutch language is evidenced by the coalition agreement of 2017, in which the four ruling political parties pledged to ensure that there are sufficient Dutch-language study programmes on offer and that EMI is used only in cases of clear added value (VVD, CDA, D66, & ChristenUnie, 2017, p. 12).
Nonetheless, a key grievance of those opposed to the rapid rise of EMI is that the relevant legislation, Article 7.2 of the Higher Education and Research Act (Wet op het hoger onderwijs en wetenschappelijk onderzoek, WHW) of 1992, is not properly enforced. The basic principle of that legislation is referred to as 'Dutch, unless': that is, the language of instruction in higher education is to be Dutch except in certain special cases. The wording surrounding the permitted exceptions, however, is formulated in such a way that the legal protection for Dutch seems rather tenuous (see also Van Oostendorp, 2012, p. 157): Article 7.2 WHW Teaching shall be provided and examinations conducted in the Dutch language. Notwithstanding the first sentence, another language may be used: a. if the study programme relates to that language, b. if teaching is provided within the context of a guest lecture given by a non-Dutch-speaking lecturer, or c. if the specific nature, organisation or quality of the teaching or origin of the students so requires, in accordance with a code of conduct adopted by the board of the institution.
As we shall see, universities tend to interpret subparagraph c in particular as a kind of 'free pass' for EMI, providing they set out their rationale in a code of conduct. The present article focuses on these institutional codes of conduct in relation to Article 7.2 WHW. In what way do the universities interpret and enact the legislation? Specifically, what discursive strategies do the universities use in their codes of conduct on language of instruction to legitimate their use of EMI? By addressing this question, the article will explore the transformations involved as policy shifts from state to institutional level, as institutions manoeuvre to balance the competing pressures of internationalising while also catering to domestic concerns. Barakos and Unger (2016, p. 7) describe 'a new field of inquiry at the intersection of discourse analysis and language policy' , which they term DALP: discursive approaches to language policy. DALP encompasses a growing body of work bringing discursive and critical perspectives to bear on language policy situations and texts. In this approach, discourse is defined as 'essentially a text in its social context, or language treated as a form of social action' , and critical refers to the adoption of a 'problem-oriented approach: questioning what is taken for granted, indicating problematic discursive practices by policymakers and other elites, and challenging dominant ideologies and normative assumptions' (Barakos & Unger, 2016, p. 3). Although 'critical language policy' -in the sense of rejecting traditional apolitical analyses and emphasising the role of ideologies, power and agency in language policy -dates back to at least the early 1990s (Tollefson, 1991), the contributions in Barakos and Unger (2016) bring specific techniques from critical discourse analysis to bear on language policy texts, an approach I take up here. Further, I take a multilayered view of language policy (e.g. Hornberger & Johnson, 2007;Ricento & Hornberger, 1996) which recognises that policy is made at different levels. This makes policy texts 'sites of struggle' (Wodak, 2009, p. 35), whereby policymakers at the different levels are influenced by divergent goals, interests and audiences. Consequently, as policy texts move from one level to the next, they are recontextualised. The notion of recontextualisation, originally a sociological concept (Bernstein, 1990), has become an important category in critical discourse analysis (Wodak & Fairclough, 2010). Illustrating the process of recontextualisation, Johnson (2015, pp. 168-169) explains how the authors of a federal educational language policy will interpret it in their own (multiple, potentially conflicting) ways, but when the policy is interpreted and appropriated by educators, new meanings may emerge. Wodak and Fairclough (2010) further developed this concept in their study of the recontextualisation of European higher education policies, exploring how the Bologna Agreement was appropriated into the domestic policy contexts of Austria and Romania. As they explain, recontextualisation is concretely manifested in linguistic transformations which can be explored through the analysis of intertextuality (Wodak & Fairclough, 2010, p. 24).

Theoretical framework
Texts are always 'constituted by elements of other texts' , and intertextuality refers to how, in a 'chain of texts' , elements of previous documents are incorporated into a new text (Fairclough, 1992, pp. 270-71). As discourse elements move along this textual trajectory they undergo specific transformations, which can be investigated by considering types of intertextuality such as 'discourse represen-tation' (Fairclough, 1992, p. 281). As Fairclough (1992, p. 281) reminds us, 'when one "report[s]" discourse one necessarily chooses to represent it in one way rather than another' , and it is fruitful to analyse the motives of the writers of a text in representing previous discourses in the way that they do. In the present study, I explore the intertextual transformations that take place as policy is recontextualised from state to institution (i.e. university) level in a specific set of higher education policy documents.

Data and method
Pursuant to Article 7.2 sub c WHW (see Section 1), higher education institutions that offer programmes taught in a language other than Dutch are legally required to justify this in an institutional code of conduct for language. The present analysis focuses on these linguistic codes of conduct from the publicly funded universities in the Netherlands. Although a number of universities also have additional language policy documentation, those documents (unlike the codes of conduct) vary greatly in scope and are not readily comparable across institutions. As can be seen in Table 1, 12 codes of conduct for language were included in the investigation. All texts are publically available and were downloaded in February 2019 from the respective university websites. Two publicly funded institutions, Tilburg University and the Open University, had to be excluded as I was unable to locate such a code of conduct for them. The included documents date from between 2000 and 2018 3 and range in length from 210 to 1,750 words (mean 846) ( Table 1).
Eight of the documents were available in both Dutch and English, three in Dutch only (Delft, Erasmus and the Vrije Universiteit) and one in English only (Wageningen). The analysis consisted in a close reading of the Dutch versions of the documents (with the exception of Wageningen, which is only available in English). Specifically, I charted (i) what language(s) are identified as the language of instruction at each university, and (ii) how this choice is motivated in light of the permitted reasons for deviating from Dutch as the language of instruction (cf. Article 7.2 WHW, Section 1). Recalling the discussion in Section 2 on the transformation of policy discourse as it shifts along an intertextual chain, I paid particular attention to the specific discursive tactics or strategies deployed in the texts that result in an institution-level recontextualisation (or reinterpretation) of the statelevel policy. Table 2 shows the language(s) of instruction at the universities as stipulated in their codes of conduct. At bachelor's level, 7 of the 12 universities reflect the 'Dutch unless' principle of Article 7.2 WHW, stating that their language of instruction is Dutch, with certain exceptions. At master's level only three universities do so. The main other option given is 'Dutch or English' , indicated by four universities at both bachelor's and master's level. Two universities specify only English as their language of instruction for the master's programmes; for a further two it is English 'or another language' . This euphemistic use of the term 'another language' is noteworthy; in the case of Radboud's phrasing Dutch or 'a language other than the Dutch language' , where the other language is exclusively English, the formulation may have political motivations. Euphemistic references to English have frequently been found in higher education language policy documents, notably in the Nordic countries (e.g. Saarinen, 2012;Saarinen & Nikula, 2013;Soler-Carbonell & Gallego-Balsà, 2016;Soler, Björkman, & Kuteeva, 2018). Universities may wish to leave the door open for another foreign language (e.g. French or German). Alternatively, it may be face-saving tactic by university administrators, so as not to appear too focused on English only (Soler-Carbonell, 2015).  (Nuffic, n.d.; Taalunieversum, n.d.) (doctoraal under the old system). I therefore take UvA's code to cover both the bachelor's and master's phases.

Context: Stated language of instruction vs statistics on language of instruction
I now compare the stated language of instruction in Table 2 with the data in Table 3, which shows the number of study programmes actually available in Dutch as a proportion of each university's programme offer (adapted from VSNU, 2019a, 2019b). At bachelor's level, Dutch-taught programmes predominate. Of the seven universities that indicate following the 'Dutch unless' principle (Table 2), five appear to follow this policy in their actual programme offer (Table 3): more than 80% of bachelor's programmes are available in Dutch at Erasmus (18 out of 22 programmes), Leiden (41/46), Utrecht (41/45), the University of Amsterdam (UvA; 53/58) and the Vrije Universiteit (VU; 40/45). The figures for the other two universities that claim to follow the 'Dutch unless' principle, Groningen and Twente, are 50% (24/48) and 30% (6/20) respectively. The stated policy, in other words, does not always match up with practice: universities that claim their language of instruction to be 'Dutch unless' do not, on closer inspection, appear to be any more committed to this legal principle than other institutions. Of the universities with 'either Dutch or English' as their stated language of instruction, the proportion of programmes available in Dutch ranges from 95% (18/19) at Wageningen to only 25% at Eindhoven (3/13). At master's level, only Radboud offers a (slim) majority of programmes in Dutch (40/69, 58%). For the other 11 universities, most (or in the case of Wageningen, all) master's degrees are taught in English. Of the three universities that claim to follow the 'Dutch unless' principle also at this level -Groningen, the UvA and the VU -the proportion of Dutch-taught programmes is around 40%. At the remaining universities, it is equally low or indeed (much) lower. Zooming in on exactly which master's programmes are still taught in Dutch is revealing. Taking Groningen as an illustration, we see that 41 out of 48 programmes offered in Dutch are either teaching-training courses, programmes with inherently Dutch subject matter (e.g. Dutch Law, Dutch Studies), or programmes whose graduates are being trained to provide (health) services to the Dutch public (medicine, dentistry, pharmacy, psychology). At master's level, therefore, the state policy 'Dutch unless' seems to have been turned into 'English unless' , except for those study programmes undeniably oriented towards the domestic labour market.

Results of the textual analysis
All 12 documents employ a similar basic structure, rhetorical moves and semilegal language. Ten of them quote the text of Article 7.2 WHW verbatim; an example of overt or manifest intertextuality (Fairclough, 1992, p. 271). The other texts -Twente and the Vrije Universiteit -reproduce almost the exact wording of Art. 7.2, but do not reference it explicitly (covert intertextuality). Similarly, although the codes of conduct do not refer to each other, a number of them have lengthy passages which are exactly identical (Leiden and Utrecht) or almost identical (six other universities). As Björkman (2015) surmised in her study of language policy documents at Swedish universities, it seems likely that one or more of the earlier texts served as a model for the others.
Initially striking from a close reading of the codes of conduct is just how little policy discourse 'work' the universities do to justify deviations from the 'Dutch unless' principle. Rather than providing detailed reasons for why the nature, organisation or quality of the teaching or the origin of the students necessitates the use of EMI, in accordance with Article 7.2 sub c WHW (Section 1), they tend to simply state that it does, with no or little further elaboration: (1) The master's education at the UT is conducted in English because the specific nature, the design, and the quality of the education make this necessary. 4 (Twente) 4. Text taken from the English version of the document. All extracts in this analysis are taken from English-language documents insofar as they exist; where extracts have been translated, the original Dutch versions are given in footnotes.
A sort of circular rationale can be seen in (2), which essentially states that the EMI programmes at Erasmus exist because it is the university's policy to offer EMI programmes: (2) The introduction of the bachelor/master structure has led to the offer of a number of English-language master's programmes. This fits with our university policy, which is ultimately aimed at offering a broad range of English-(Erasmus) language programmes. 5 The quote above draws a direct link between the bachelor/master structure and the introduction of English-language programmes. Although the rise of EMI is undoubtedly a side effect of the Europe-wide efforts to make study programmes more comparable across countries (including the switch to the BaMa structure), the BaMa system itself does not dictate any particular language of instruction. The effect of this formulation is to present the introduction of EMI as something outside the university's control, imposed externally rather than initiated from within.
The subsections below illustrate how the universities invoke the reasons for deviating from Dutch as the language of instruction as permitted by Article 7.2 sub c WHW. The examples are structured according to the categories identified in Art. 7.2(c), i.e. the nature, organisation or quality of the teaching or the origin of the students, plus an additional rationale (preparing students for the master's phase) as noted by a number of universities. Virtually all 12 universities give justifications falling under several or all of these categories. It should be noted that the categories are not mutually exclusive. For example, the rationale of fostering contact between Dutch and international students by means of an 'international classroom' relates not only to the origin of the students but also to the organisation of the course. What follows is therefore primarily intended as a qualitative illustration of each category, although where possible I indicate how many universities invoked each type of rationale.

Nature of teaching
As pointed out in Section 1, the permitted reasons for deviating from the 'Dutch, unless' principle are somewhat vague. The explanatory notes to the legislation provide some further detail, indicating that the 'nature' and 'organisation' of a study programme can relate to matters such as internationalisation and international exchange programmes. Eight universities (Eindhoven, Erasmus, Groningen, Leiden, Radboud, Twente, Utrecht and the VU) use either the text below

De introductie van de bachelor-masterstructuur heeft geleid tot een aanbod van een aantal Engelstalige masteropleidingen. Dit past in het beleid van onze universiteit dat erop is gericht op termijn een breed scala van Engelstalige opleidingen aan te bieden.
or an almost identically worded clause to indicate that the use of a language of instruction other than Dutch (read: English) is (3) determined by the international orientation of a programme in terms of its academic field of study, or by the fact that it prepares students for a specific (Utrecht) field of activity or professional career.
Going beyond this, several codes of conduct justify the use of EMI by highlighting the generally international profile of the university as a whole and/or its global reputation/aspirations -a significant discursive shift from the legislative text, which permits deviations from the 'Dutch, unless' rule on the basis of the international content of a specific programme. This can be seen in (4), together with, again, a circular logic that seems to imply that Groningen is justified in using EMI because it is 'internationally oriented' , as evidenced (among other things) by its English-language programmes: (4) [T]he University of Groningen is internationally oriented, which in the area of education is demonstrated by its: a. international student population b. English-taught degree programmes (Bachelor's, Master's and PhD degree programmes) c. international cooperation agreements with universities on: -joint and double degree programmes -international mobility of students and staff (both outgoing and incoming exchange programmes) -participation in European Commission projects (such as Erasmus+) -strategic partnerships (e.g. U4 Network, Coimbra Group) […] (Groningen) In (5), intertextuality is used as a tactic of legitimation: Delft's code of conduct makes detailed reference to another formal document, laid down by the governing body of the university, which justifies why some master's programmes can be offered English. The fact that all programmes are now taught in English is then simply presented as a fait accompli.

Organisation of teaching
In the English versions of the codes the term inrichting is translated in various ways: organisation is used, as in the official translation of Article 7.2, but so too is structure, design and setup. The plethora of terms is telling, pointing to the difficulty of pinning down exactly what the Dutch text really means. The codes of conduct identify several aspects that could be considered 'organisational' justifications for EMI. Groningen refers to teaching that forms 'an optional part of the programme, i.e. [which] does not form part of the mandatory course units of the curriculum' . The UvA mentions cases in which 'it is deemed essential to provide components in a language other than Dutch as a skill component in the specific discipline of the programme' , or in which the programme 'includes specialisations for which knowledge of a language other than Dutch is indispensable' (although how terms like 'essential' and 'indispensable' are operationalised remains unclear). The UvA code also indicates that 'financial aspects' may play a role in the decision to deviate from Dutch. Again, these are not further elaborated upon, but presumably include situations in which, for example, there are too few student enrolments for a Dutch variant of a programme to be viable.

Quality of teaching
According to the explanatory notes to the legislation, teaching quality may be a permissible reason to use a language of instruction other than Dutch if the appropriate expertise is unavailable in the Netherlands and thus a foreign lecturer needs to be used. In this vein, Delft's code of conduct mentions 'the provision of courses in English because the nationality of the lecturer […] does not permit otherwise' (my emphasis). Beneath this reference to nationality lies the ideological assumption that a non-Dutch lecturer will, self-evidently, not be able or expected to teach in Dutch. In addition, here again we see an important discursive shift: where the state policy seems to concern situations in which non-Dutch staff need to be brought in temporarily to teach a specific class or course, the institutional codes of conduct interpret this as referring to their non-Dutch employees in general. Another discursive transformation can be seen in the following rationale by the VU, which states that a language of instruction other than Dutch can be used (6) if teaching is provided within the context of a guest lecture by a non-Dutchspeaking lecturer. "Teaching" is understood to include giving assessments and (VU) exams. 7 While the first clause in (6), which reproduces Article 7.2 sub b WHW verbatim, refers explicitly to a 'guest lecture' -a typically one-off, isolated event -the second sentence refers to teaching involving 'assessments and exams' , implying one or more entire courses.

Origin of students
Neither Article 7.2 WHW nor the codes of conduct specify exactly how many non-Dutch-speaking students need to be present in a classroom or enrolled in a course to justify the switch to English. What can be seen in the codes, however, is yet another discursive shift from the specific to the general: while the state legislation refers to specific subjects or programmes including international students (a less common situation back when the law was adopted in 1992), the codes of conduct reference the general presence of international students across the university as a whole. For example, Utrecht seeks to (7) […] create an international learning environment with a student population of Dutch and international students, which leads to necessary choices related to (Utrecht) the language of instruction.
Here a chicken-and-egg argument can be discerned: the presence of international students is said to 'lead to' the need for a different language of instruction (i.e. English), yet these students were almost certainly drawn to the university because of the English-language programmes on offer (see also Inspectie van het Onderwijs, 2018b, p. 22).

5.5
Additional rationale: Preparing students for master's phase As we have seen, Article 7.2 WHW provides several permissible reasons for deviating from the 'Dutch, unless' principle. Four universities (Delft, Erasmus, Utrecht and the UvA) identify an additional reason: because the majority of master's programmes are in English, it is necessary to teach in English at bachelor's level too in order to prepare students for the master's phase. An example of this circular logic can be seen in (8): (8) […] the BaMa law stipulates that every student after the bachelor's must be able to continue on with a master's programme within the institution. Against this background it is of great importance that the student, if the continuing master's programme is taught in English, is optimally prepared for this transi-(Erasmus) tion. 8 As in (5), intertextuality -the reference to the BaMa law -functions here as a strategy of legitimation: by emphasising the required continuity between the bachelor's and master's phases, the university discursively positions itself as explicitly following the law (i.e. the BaMa law), while stretching the intended meaning of the law on language of instruction.

Discussion and latest developments
At bachelor's level, 8 of the 12 universities investigated here provide the majority of their bachelor's programmes in Dutch, thereby complying with the 'Dutch, unless' principle of Article 7.2 WHW. At three universities (Eindhoven, Maastricht and Twente), the majority of undergraduate programmes are in English, as are half of the bachelor's at Groningen. At master's level, the exceptions have become the rule: English-language programmes predominate at 11 of 12 universities, and at Radboud the ratio is 50/50. To justify this policy transformation from 'more Dutch' in the state legislation to 'more English' at institution level, the university codes of conduct use various discursive tactics. They deploy intertextuality as a strategy of legitimation, referencing other laws and regulations to discursively construct the institution as acting lawfully and responsibly. In addition, they effect shifts in interpretation from the specific to the general to provide blanket justification for the use of EMI. Where Article 7.2 WHW permits a language of instruction other than Dutch if a specific programme or course is internationally oriented, the codes reference the international orientation of the university as a whole. Likewise, where the law identifies the participation of international students or the need for a foreign lecturer as permissible reasons to teach a particular course in English, the codes highlight the presence of many international students and staff at the university in general.
These results echo those of Hultgren (2014), who found that state-authored policies in Danish higher education promote the national language, while institution-authored policies turn this on its head in favour of English. Other research in the closely related Nordic context similarly illustrates the conflicting interests at the state and institutional levels (Saarinen & Taalas, 2017). Where national policymakers are concerned with language policy in connection to identity politics in the public debate, university administrators embrace EMI in the 8. […] de BaMawet bepaalt wel dat iedere student na zijn bachelor naar in ieder geval een masteropleiding binnen de eigen instelling moet kunnen doorstromen. Het is tegen die achtergrond van groot belang dat de student, indien de doorstroommaster in het Engels wordt verzorgd, optimaal in staat wordt gesteld daarin in te stromen.
context of internationalisation (Soler-Carbonell, Saarinen, & Kibbermann, 2017). In the codes of conduct studied here, internationalisation was both an implicit and explicit motivation underlying language-policy decisions, although the term was never defined, nor was the rationale for pursuing it made explicit. The largescale switch to EMI by Dutch universities is a predictable consequence of (and perhaps even intended by) the open formulation of Article 7.2 WHW: the government may need to be seen as safeguarding Dutch, but it wants to facilitate internationalisation in order to increase student numbers as well as the global competitiveness of the country's universities. As Fairclough (1992, p. 271) reminds us, 'a text may be designed to be interpreted in different ways by different readerships or audiences, [an] anticipatory, intertextual source of ambivalence' . Thus, the state legislation offers loopholes of which the universities in turn make liberal use, highlighting '[t]he gap between the rhetoric of medium-of-instruction policy and the reality of its implementation' (Tsui & Tollefsen, 2004, p. 5): i.e. the state discourse on protecting Dutch while tacitly promoting English.
This rhetorical gap is also evident in the latest developments surrounding higher education language policy in the Netherlands. A recent report by the Inspectorate of Education found that many universities' linguistic codes of conduct, with their flimsy treatment of the considerations involved, are not in compliance with Article 7.2 WHW: 'The idea appears to be that if an institution has a code of conduct, it ultimately complies with the law' 9 (Inspectie van het Onderwijs, 2018b, p. 32). Prompted in part by the findings of this report, the legislation is currently being updated and clarified, amid tough political rhetoric: the new rules are intended to be easier to enforce, and the education minister has announced that study programmes may lose their accreditation if they fail to adequately justify the use of EMI (Rijksoverheid, 2018). Yet the proposed new law (Van Engelshoven, 2018), intended to come into effect 1 March 2020, appears, if anything, to be even less strict than the current law. While it retains the 'Dutch, unless' principle, the requirement to demonstrate the 'necessity' of EMI (which is in any case often ignored) will be replaced by the seemingly looser principle of 'added value' . Recall from Section 1 that sub c permits deviations from Dutch as the language of instruction if the specific nature, organisation or quality of the teaching or origin of the students so requires, in accordance with a code of conduct adopted by the board of (Art. 7.2c WHW) the institution.
The proposed new law will permit a language other than Dutch if 9. Als een instelling een gedragscode heeft, voldoet de instelling tenslotte aan de wet, zo lijkt de gedachte te zijn. this is, more than the use of Dutch, and in view of the specific nature, organisation or quality of the teaching, in the interests of the acquisition of the knowledge, insight or skills that a student should have acquired by the end of the pro-(Van Engelshoven, 2018, p. 2) gramme. 10 It is unclear how this will alter universities' current practice, other than the fact that, instead of a code of conduct for language, they will be obliged to draw up more comprehensive language policies. Specifically, the new policies will be required to describe (a) not only the rationale but also the official procedure by which the decision is made to teach a programme in a language other than Dutch, and (b) the measures the institution will take to ensure the quality of the teaching and the accessibility of the programme for Dutch students (Van Engelshoven, 2018, p. 2). Additionally, rather than a static code of conduct, these policies will need to be evaluated cyclically (Inspectie van het Onderwijs, 2018b, p. 37). It therefore seems the new law will simply require universities to jump through more administrative hoops to continue along the same road they are already travelling.

Conclusion
This article has provided a snapshot of current language policy at the main Dutch universities. Although limited to the institutions' codes of conduct for language, the analysis has revealed several salient aspects and similarities in the way in which the universities seek to implement the state legislation on language of instruction in higher education. Future research would do well to take into account other language policy documentation, which will be required by law once the amended legislation comes into effect in 2020 (Section 5). Moreover, while this study considered the state and institution levels of policymaking, a microanalysis at the programme level would provide additional insight. Different disciplines have different motives and needs in implementing EMI (Kuteeva & Airey, 2014;Saarinen & Taalas, 2017). Moreover, the study of language policy documents and research on actual linguistic behaviour are complementary; as Spolsky (2004, p. 222) put it, 'real' language policies lie in language practices. Recent research on university EMI programmes in different countries, for example, has shown that classroom interaction is much more complex than the 'official' language policy: students and teachers alternate between languages and negotiate language policy 10. dit gelet op de specifieke aard, de inrichting of de kwaliteit van het onderwijs, meer dan het voeren van het Nederlands, in het belang is van het verwerven van de kennis, het inzicht of de vaardigheden die een student bij beëindiging van de opleiding moet hebben verworven.
at the local level (e.g. Haberland, 2014;Mortensen, 2014;Söderlundh, 2012). The present study of language policy documents in Dutch higher education can provide a jumping-off point for more ethnographically oriented investigations of how the official language policy is implemented on the ground. Although the focus was on the Dutch context and no claims to greater generalisability can be made, it is hoped that this discussion will connect and resonate with other countries where EMI has become an important issue in higher education.