Index of subjects Page 1 of 25 Varieties of English in the Netherlands and Germany Alison Edwards Leiden University Robert Fuchs University of Hamburg Abstract It has been suggested that continental European varieties of English with their own endonormative standards may be emerging. We surveyed more than 4,000 Dutch and German respondents on their degree of acceptance of ‘Dutch English’ and ‘German English’, respectively. German respondents were found to be significantly more open than Dutch respondents to the notion of a local English variety. Multivariate regression analysis showed that lower acceptance was associated with higher proficiency levels, a more positive orientation towards English, a stronger belief in the importance of English and greater exposure to English in higher education. Interaction effects were found between nationality and both proficiency levels and positive attitudes, whereby the most proficient and English-oriented Dutch respondents were more strongly oriented towards native varieties than their German counterparts. The findings have implications for the development of targeted interventions to raise awareness of the diverse forms of English in Europe and around the globe. Keywords ELF, Expanding Circle, variety, English in Europe, English Language Teaching, attitudes 1 Introduction As English becomes ever more entrenched as an additional language in mainland Europe, a growing body of research suggests it is no longer used purely instrumentally, for the purposes of international communication, but instead is becoming an integral part of the sociocultural identity of many continental Europeans (e.g. Berns, De Bot, and Hasebrink 2007; Berns 1995, 2005; Edwards 2016; Leppänen et al. 2011; Preisler 1999; Proshina 2005). In this context, an increasing number of studies have focused on mainland Europeans’ attitudes to different non-native (NNS) varieties of English. Attitudes to the notion of a pancontinental ‘Euro-English’ have been found to be largely negative (Van den Doel and Quené 2013; Forche 2012; Gnutzmann, Jakisch, and Rabe 2015; Groom 2012; Mollin 2006; Murray 2004; Sing 2004). However, there are indications that more localised (regional or national) varieties of English may be viewed as more acceptable. In their survey of undergraduates at a German university, Gnutzmann, Jakisch and Rabe (2015: 82) found that students had a strong conception of the link between nation and language, and viewed the development of a Euro-English as undesirable because, as one of their respondents put it, ‘every nation wants to identify with its own language’. This perceived primacy of national identity would Index of subjects Page 2 of 25 seem to suggest that the notion of differentiated national varieties may be more palatable to speakers than that of a single pan-European variety. Furthermore, many scholars have expressed scepticism about the plausibility of linguistic convergence in the form of a Euro-English – yet while Görlach (2002: 151), for instance, described the term ‘European English’ as ‘little more than a catchphrase’, he also suggested that the English used by continental speakers who share the same L1 may display recurrent features that ‘if a tradition establishes itself, [may] lead to a national variety of English.’ Numerous other scholars, too, have suggested that different continental European varieties of English, with their own endonormative standards, may emerge or already be emerging (Berns 1995; Bruthiaux 2003; Hilgendorf 2001; Kirkpatrick 2007; McArthur 1998; Wilkinson 1990) (we therefore refer to them as ‘(emergent) national varieties’, although see further section 4). Focusing on the cases of Germany and the Netherlands, this paper sets out to explore whether this prediction is borne out in lay attitudes: given the high prestige and linguistic capital associated with traditional native (NS) varieties, do continental users of English 1 recognise and accept the notion of ‘legitimate’ local varieties of English? 2 Previous research The acceptance of local varieties of English would seem to be supported by social identity theory, whereby individuals align themselves with the language variety associated with their most salient in-group (Lambert 1967; Tajfel and Turner 1979). Studies of English accent evaluations typically show that native accents are preferred by continental Europeans, but also that a sense of solidarity plays a role in enhancing perceptions of continental varieties, with NNS groups tending to rank their own accents better than those of others, and better than they are ranked on average by other groups (e.g. Dalton-Puffer, Kaltenboeck, and Smit 1997; Risan 2014). Yet this effect of solidarity does not seem to be uniform: while it appears to hold for German raters (Davydova 2015, Jenkins 2007), ‘inverse solidarity’ has repeatedly been reported for the Netherlands (Van den Doel and Quené 2013; Hoorn, Smakman, and Foster 2014; Koet 2007), with Dutch participants in listening experiments typically giving more negative evaluations of compatriots’ accents than NS participants. In addition to L1 background, age, too, has been suggested to play a moderating role in attitudes to European varieties of English. Following Mollin’s (2006) widely cited survey of European academics, several replication studies have been conducted among younger populations of university students (Forche 2012; De Meerleer 2012). Their results point towards a possible generational shift, whereby young, mobile urbanites seem to be more open to the notion of NNS 1 We are aware of the arguments (e.g. Makoni and Pennycook 2007) surrounding the notion of languages and language varieties as discrete entities, a notion on which a varieties-based approach is inevitably premised. While we do not necessarily subscribe to such a strictly bounded view, we note that languages are nevertheless popularly perceived and variety labels popularly used as such, and it is these perceptions/ideologies that were are interested in exploring. Index of subjects Page 3 of 25 varieties and English as a Lingua Franca (ELF) and feel less need to associate ‘good’ English with its NSs (e.g. displaying higher rates of agreement than Mollin’s academics with statements such as ‘English doesn’t belong to the native speakers anymore, but to anybody who uses it’). A similar sense of ownership was reported in Kuteeva, Hynninen and Haslam’s (2015: 95) survey of Swedish business undergraduates, who appeared to view English as a language ‘that is almost theirs (“we know how to use English” type of attitude)’. Yet the effect of age does not seem to be entirely straightforward. In Leppänen et al. (2011), older Finns reported having a NS target model more frequently than did their younger compatriots, but they were also overrepresented among people who chose Finnish English as their target model. Furthermore, while relative youth may be associated with a weaker degree of exonormativity, this need not mean younger people are necessarily in thrall to their own national varieties; instead they often express favourable attitudes towards an ‘international’ or ‘neutral’ variety (Erling 2004; Ranta 2010; Rindal and Piercy 2013; Rindal 2015). Attitudes towards endonormative continental varieties are also influenced by the speaker’s personal relationship with and orientation towards English. Studies focusing on (future) teachers of English (e.g. Grau 2005; Murray 2004; Nykänen 2015; Ranta 2010) tend to show that their personal investment in the language moderates their selection of target model and that they hold themselves to stricter standards than their (future) students. Ranta (2010: 174), for instance, reported that while 85% of the Finnish teachers of English she surveyed personally aimed for British English, they ‘wanted to convey to their students the message that English was a “universal” language, and that the flow of communication in it was more important than normative accuracy’. 3 The present study In this paper, therefore, we aim to explore the relationship between the degree of lay acceptance of endonormative European varieties of English and a range of moderating sociodemographic and attitudinal variables. Our focus is on Germany and the Netherlands, neighbouring countries that both belong to the Expanding Circle of English (Kachru 1985) but which have quite different linguistic situations. Dutch, with approximately 22 million speakers (Lewis, Simons, and Fennig 2015b), is used in the Netherlands as well as Belgium, the former Dutch Antilles and Suriname. German has roughly 87 million L1 speakers, mainly in Germany, Austria, Switzerland, Luxembourg and Liechtenstein (Ammon 2015: 170), and a further 53 million L2 speakers (Lewis, Simons, and Fennig 2015a), thanks in part to its historical popularity as an L2 in the Central European and Nordic countries. As such, German is classed in De Swaan’s (2001) global language system as one of 13 ‘supercentral’ languages (see Mair, this volume) that serve as regional lingua francas. By contrast, Dutch is considered one of around 100 ‘central’ languages whose roles are largely limited to the nation-state level (De Swaan 2001), and indeed the Netherlands has long been known for its tradition of foreign language learning, being a small country located at the crossroads of three major language areas (German, French and English) whose ‘economic survival depends to a large Index of subjects Page 4 of 25 degree on cross border economic and cultural transactions within these areas’ (Ammon and McConnell 2002: 98). In the second half of the twentieth century, English became firmly rooted as the first foreign language in both Germany and the Netherlands. Although English has periodically been claimed to be transitioning from a foreign (EFL) to a second (ESL) language in both countries (Ammon and McConnell 2002: 99; Berns 1995b: 9–10; Booij 2001; Jenkins 2009: 16–17; Kirkpatrick 2007b: 165; McArthur 1998: 54; Mesthrie and Bhatt 2008: 211; Ridder 1995: 44), the Netherlands is clearly further along in this process. This is reflected in proficiency levels, with 90% of the Dutch population reportedly able to hold a conversation in English compared to 56% of Germans (European Commission 2012: 21). The practice of subtitling films and television programmes, rather than dubbing them, affords the Dutch population many more hours of exposure to English than their German peers (Bonnet 2002). Bilingual secondary education is offered by as many as one fifth of German grammar schools, although instruction in the second language (which is often but not always English) usually concerns only one or two subjects over a limited period of time (FMKS 2014; Annette Lommel, p.c.). By contrast, in bilingual curricula in the Netherlands, now followed by one quarter of grammar school students (Dronkers 2013), the majority of subjects are taught in English for the first three years, decreasing to under 50% in the final years before the national exams, which are in Dutch. At university level, the Netherlands offers the highest absolute number of English-medium programmes in continental Europe: some 30% of all degrees on offer at Dutch universities are taught in English (compared to 6% for Germany) (Maiworm and Wächter 2014). While only 1.0% of students enrolled at German universities in 2013 were following English-medium programmes, one third of them being domestic students, 7.2% of university students in the Netherlands were enrolled in English-medium programmes, just over half of them domestic students (Maiworm and Wächter 2014). Moreover, the majority of Dutch universities now bill themselves as officially bilingual; indeed, the internationalisation process undergone by Dutch universities over the last two decades is seen by some as more or less synonymous with Englishisation (Zegers and Wilkinson 2005). The different sociolinguistic contexts of the two countries thus make for a compelling comparative study. Hand in hand with the increasing ESL status and intranational uses of English in the Netherlands come anecdotal reports of proficient Dutch users of English employing nativised lexicogrammatical and pragmatic features even after being made aware that they do not conform to ‘native’ norms (Edwards 2016) – yet previous research has shown that Dutch attitudes to English in general continue to be highly exonormative (Edwards 2016; Van der Haagen 1998). Conversely, although Germany remains relatively closer to the EFL end of the spectrum, attitudes towards a local NNS variety of English appear to be enhanced by a sense of solidarity (Davydova 2015, Jenkins 2007: 165), in contrast to those in the Netherlands (Van den Doel and Quené 2013; Hoorn, Smakman, and Foster 2014; Koet 2007). In this context, we set out to explore the degree of acceptance of national, endonormative varieties of English in Germany and the Netherlands. The research questions are as follows: Index of subjects Page 5 of 25 1. 2. 3. 4 To what extent are users liable to recognise and accept varieties such as ‘German English’ and ‘Dutch English’? What are the sociodemographic and attitudinal factors that predict such acceptance (or otherwise)? What are the similarities and differences in this regard between Germany and the Netherlands? Methodological considerations Much previous research (section 2) has focused on a specific population segment (e.g. English students or ELT practitioners), been limited to accents only, and/or been exploratory in nature, drawing on small-scale survey and/or qualitative interview data to tease out factors involved in shaping Europeans’ attitudes to NNS English varieties. Building on such work, we aim to investigate the attitudes of a broad cross-section of the populations across the two L1 backgrounds by means of a dataset large enough to be subjected to quantitative analysis (specifically, multivariate regression analysis) in an effort to (i) disentangle the interplay between the factors involved and (ii) evaluate statistically which of these factors appear to be the driving forces behind acceptance rates. Our approach has similarities with studies in perceptual dialectology (Preston 1989), which investigate how people overtly categorise and evaluate language varieties. However, our approach also straddles the border between direct and indirect methods to accessing language attitudes (Garrett 2010) in that we use multiple survey items formulated in both a direct and an indirect fashion, and combine the responses so as to create composite measures that tap into latent (underlying) attitudinal constructs. Let us take the dependent variable in our regression analysis, ‘acceptance of national variety’, as an illustration. We seek to measure participants’ attitudes to the notion of locally constructed NNS varieties by posing multiple, related questions in different ways (i.e. using variety labels as well as more indirect formulations such as ‘As long as my English is good I don’t mind if it has a bit of [German/Dutch] “flavour”’), and subsequently combining the observed metrics into an overarching factor score per respondent for the outcome variable. Such an approach allows us to minimise the effect of stigma often associated with variety labels (see e.g. Rindal 2015 and Rindal and Piercy 2013 on ‘Norwegian English’), and to arrive at robust measures of ideological openness towards the notion of a ‘legitimate’ local variety even as ‘German English’ and ‘Dutch English’ cannot (yet) be considered established, focused varieties. 5 Data and methods 5.1 Questionnaire The questionnaire was originally developed as part of a broader project on the functions, forms and attitudes towards English in the Netherlands (Edwards 2016). The present analysis draws on (i) data originally gathered for the purposes of that study by means of an attitudinal questionnaire among Dutch informants, and (ii) Index of subjects Page 6 of 25 data gathered via a parallel version of the questionnaire subsequently developed and disseminated in Germany. The questionnaires were administered in Dutch and German, respectively, and are available on request. The covering information stressed that actual knowledge of English (as opposed to opinions about the language) was not required. The questions were derived and adapted from relevant surveys on attitudes to English in Europe (Buschfeld 2013; Erling 2004; Leppänen et al. 2011; Preisler 1999). The first section collected personal information about the respondents, while the main section primarily consisted of closed questions, with respondents required to mark their responses to attitudinal statements on a four-point Likert scale ranging from strongly agree to strongly disagree (or for the proficiency scales, from not at all to fluently). In a final, optional open-ended question, respondents were invited to comment generally on the use of English in their country. 5.2 Data collection Both questionnaires were disseminated online via Google Forms using a snowball sampling procedure. Particular use was made of social media (Facebook, Twitter, LinkedIn, Reddit), platforms for student associations and language organisations, and, to gain as balanced a sample as possible, networks targeting older populations such as senior citizens’ groups and mailing lists. The Dutch version was live for approximately six months in 2013; the German version for nine months from June 2015. A total of 4,372 responses were received. Invalid responses (e.g. blank forms and duplicate forms) were excluded (n=209), leaving a total of 4,162 responses for inclusion in the analysis. 5.3 Variables The responses were coded as shown in the overview of variables in Table 1. These can be divided, roughly, into sociodemographic variables (nationality, age, sex, education level, higher education language, place of residence and self-reported proficiency level) and attitudinal variables (acceptance of national variety, positive attitude towards English, and belief in the importance of English). Table 1. Overview of variables used in the analyses Variable code ACCEPT_NA TIONAL _VARIETY POSITIVE_AT TITUDE_ENG Description Acceptance of ‘German English’ or ‘Dutch English’, respectively Positive attitude towards English Levels Average of responses to 3 questions:  As long as my English is good, I don’t mind if it has a bit of [German/Dutch] ‘flavour’  When I speak English to outsiders, they should not be able to recognise where I’m from*  [Gerlish/Dunglish] is ‘bad’ English* Average of responses to 4 questions:  English is very important to me personally  I like using English  I always use English when I have an Index of subjects Page 7 of 25 BELIEF_IMP ORTANCE_E NG Belief in the importance of English PROFICIENC Y English proficiency level NATIONALIT Y AGE SEX EDUCATION Nationality Age Sex Highest attained education level HIGHER_EDU _LANG Main language of instruction in higher education RESIDENCE Current place of residence opportunity to do so  Sometimes I resent the fact that I have to use English* Average of responses to 4 questions:  English offers advantages in seeking good job opportunities  English has a higher status than [German/Dutch] in [Germany/Netherlands]  For [Germans/Dutch], [German/Dutch] is more important than English*  English skills are overrated* Average of self-reported speaking, listening, reading and writing scores Dutch, German – male, female low = primary school/lower secondary education† medium = vocational education‡ high = university education§ NationalLang = (mainly) [Dutch/German] English = English bilingual = [Dutch/German] + English other = e.g. French, Swedish none = no higher education (yet) city+ = population >250,000 city = population 50,000–250,000 town = population 5,000–50,000 country = population <5,000 abroad = resp. outside Germany or the Netherlands *negatively keyed items † Haupt/Realschule (Germany) or VMBO/MBO (Netherlands) ‡ Abitur + Berufsausbildung (Germany) or HAVO/HBO (Netherlands) § Abitur + university/Fachhochschule (Germany) or VWO/WO (Netherlands) 5.3.1 Sociodemographic variables Appendix 1 provides a detailed breakdown of respondents per sociodemographic variable. In brief, roughly half the respondents were German (‘DE’, 47.7%) and 2 half were Dutch (‘NL’, 47.9%). The remaining 4.1% had a different nationality (e.g. Belgian) and were excluded from the analyses. Respondents ranged in age from 11 to 92 years (DE M=40.7, SD=16.0; NL M=44.3, SD=16.8). For both national groups slightly more women than men filled in the survey (female DE 53.0%, NL 56.4%). 2 People with dual nationality were coded as follows: e.g. German + British = German, Dutch + Polish = Dutch. Index of subjects Page 8 of 25 Education was coded into three levels. In the Dutch education system, pupils are divided into three different secondary school levels training them for trades, vocational education or academic education. As each secondary stream has its own dedicated further education stream, the division tends to persist into higher education. Hence, based on this three-way division we coded the Dutch respondents 3 as low, medium and high (Table 1). German respondents were classed in roughly the same way. The main difference is that the highly educated German group includes students or graduates of Fachhochschulen (universities of applied sciences), whereas those of the conceptual equivalent in the Netherlands (HBO) are classed as having a medium education level. This is because while it is not unusual for Germans with an Abitur (the highest-level secondary school diploma) to go on to attend a Fachhochschule, this is rather more unusual in the Netherlands; the vast majority of HBO students attended one of the lower secondary school streams (Ministry for Education Culture and Science 2015). This difference in classification is reflected in the comparatively larger highly educated group for Germany (DE 84.1%, NL 61.9%) and the larger medium group for the Netherlands (DE 7.9%, NL 34.6%); but as we shall see, this variable does not turn out to be highly pertinent for our findings. Given the increasing role of English in the European higher education space, medium of instruction in higher education was also included as a variable (where applicable; 13.2% of German and 2.8% of Dutch respondents reported no higher education). The majority of respondents followed higher education either entirely or mainly in their national language (DE 64.9%, NL 68.7%). This group is intended to include respondents who merely had some literature in English (hence the additional designation ‘mainly’). Higher education in both the national language and English (reported by roughly 14% of respondents in each country) could be interpreted as including formally bilingual programmes, programmes involving some classes in English (either in their own country or as an exchange/Erasmus student) or several programmes in different languages (e.g. a Dutch/German bachelor and an English master). Nearly twice as many Dutch respondents reported studying entirely in English (DE 6.9%, NL 12.4%). Very small proportions reported studying entirely in a different language, and these were excluded from the statistical analyses below. Finally, around half of the German respondents lived in cities with a 4 population of more than 250,000, as compared to one fifth of Dutch. Dutch respondents were more likely to live in small cities with between 50,000 and 250,000 people (DE 21.4%, NL 37.4%). Roughly a quarter of the respondents from both countries lived either in towns or in the countryside. A minority were currently living abroad and they, too, were excluded from the analyses so as to keep our data manageable. 3 We considered this ‘stream-based’ means of classification preferable to one based on the highest attained level at the time of participation, which would have resulted in e.g. fourthyear university students being classified the same as someone with a diploma from the lowest secondary school stream. 4 We began with a more fine-grained distinction between cities with 250,000–500,000 inhabitants and > 500,000 inhabitants, but there was no change in results whether we kept these categories apart or collapsed them. Index of subjects Page 9 of 25 5.3.2 Composite variables The three attitudinal variables and the proficiency variable were created on the basis of the combined responses to various questionnaire items (first four rows of Table 1). ACCEPT_NATIONAL_VARIETY, which will serve as the dependent variable, refers to respondents’ degree of openness towards or acceptance of an emergent endonormative variety. The other attitudinal variables, POSITIVE_ATTITUDE_ENG and BELIEF_IMPORTANCE_ENG, refer to the degree to which respondents have a positive orientation towards English and view the language as important (i.e. economically prestigious) in their respective societies; we expected higher scores on these two variables to be associated with traditionally exonormative attitudes and thus with lower acceptance rates. The procedure for arriving at the composite variables was as follows (cf. Starkweather 2012): (i) selection of observed metrics (i.e. questionnaire items) to combine, (ii) recoding of Likert-style responses as numeric responses, and (iii) computation of factor (i.e. combined) scores. With respect to (i), the questionnaire was explicitly designed to include multiple items tapping the same latent construct . Taking a minimum correlation coefficient of roughly 0.3 as a guide (Field, Miles, and Field 2012: 770), the relevant items were considered to be sufficiently correlated based on the correlations reported in previous research (Edwards 2016: 5 86–87). Turning to (ii) and (iii), we illustrate the procedure using the variable ACCEPT_NATIONAL_VARIETY. As shown in Table 1, this composite variable is based on three statements: (i) (ii) (iii) As long as my English is good, I don’t mind if it has a bit of German/Dutch ‘flavour’ When I speak English to outsiders, they should not be able to recognise where I’m from 6 Gerlish/Dunglish is ‘bad’ English. The responses (strongly disagree, somewhat disagree, somewhat agree, strongly agree) were recoded numerically, whereby the responses to item (i), which is positively keyed, were recoded from 1–4, while those for the negatively keyed 5 The exception is the composite variable BELIEF_IMPORTANCE_ENG: with 4 questionnaire items there are 6 correlation coefficients, 3 of which did not reach the threshold of roughly 0.3; however, common sense suggests these questionnaire items (row 3, Table 1) are tapping related attitudes. In any event, this variable does not play a major role in our results. 6 As pointed out by a number of survey respondents, the term used in the German version of the questionnaire, Denglisch, can also be used to refer to German characterised by heavy use of Anglicisms, rather than English characterised by transfer from German. We hope that the remainder of the sentence – ist schlechtes Englisch (rather than schlechtes Deutsch) – will have served in favour of the intended reading among the vast majority of respondents. Running the regression analysis with this question excluded (i.e. with an outcome variable based on just two rather than three questions) did not substantially change the results. Index of subjects Page 10 of 25 items (ii) and (iii) were reverse scored from 4–1 (in Table 1, all negatively keyed questionnaire items are indicated with an asterisk). Each respondent’s coded scores for the three statements were then combined to create a factor (i.e. aggregate) score using the sum-score method (DiStefano, Zhu and Mîndrilă 2009: 2); that is, the scores were simply added and divided by the number (out of three) of questions answered in order to arrive at a final score per respondent for the variable 7 ACCEPT_NATIONAL_VARIETY. This has the advantage that respondents who answered only one or two of the three questions need not be excluded. More broadly, the combination of questionnaire items helps not only to create robust factor scores but also to arrive at the most parsimonious regression model (see further section 5.4) by reducing the number of variables and at the same time avoiding potential problems with multicollinearity. The same procedure was followed for the other two attitudinal variables as well as the proficiency variable. Figure 1 presents the outcome of this procedure for all four composite variables, with boxplots indicating the full range of responses, the interquartile range and the median per nationality. The top left-hand panel shows that the Dutch respondents rated their proficiency levels somewhat higher than did the Germans. The two national groups had the same median scores in terms of positive attitudes to and belief in the importance of English. Most striking is the lower right-hand panel, which reveals a higher degree of acceptance of an emergent national variety of English among the German as compared to the Dutch respondents. We will return to this in the results section. 7 Composite variables often include items with different scale metrics (units of measurement), in which case it is typical to standardise the variables by converting them into z-scores before averaging them. In our case this was unnecessary, as all items were measured on a four-point scale. Index of subjects Page 11 of 25 Figure 1. Scores on the four composite variables per nationality. 5.4 Statistical analyses All analyses were carried out in RStudio version 0.99.491 (RStudio Team 2015). We performed a multivariate linear regression analysis with ACCEPT_NATIONAL_VARIETY as the dependent variable. Because any such analysis first requires careful selection of variables to be included, we first computed the mean rates of acceptance of the respective national variety for each variable listed in Table 1 individually. This was done using the aov function for one-way analysis of variance (ANOVA), followed by Tukey’s test for posthoc pairwise multiple comparisons. In this way, we were able to identify variables (i.e. sex and place of residence) that had no significant effect, and were thus not included in the regression model (see further the discussion in section 7). Our aim in conducting the regression analysis was to identify (i) which variable(s) had the largest effect on acceptance of the respective national variety, and (ii) whether there were differences between the German and Dutch samples. The model was computed using the lm function, again using Tukey’s test for posthoc comparisons. Initially, all variables listed in Table 1 were included (except sex and residence, as noted above), plus their interactions with NATIONALITY. Subsequently, one non-significant variable was removed at a time, starting with the one with the largest p-value, until only significant predictors remained. For each successive model the Akaike Information Criterion (AIC, Akaike 1974), which penalises models with more predictors to prevent over-fitting, was inspected in order to identify the parsimonious model. This led us to drop the interaction effects between NATIONALITY and EDUCATION and between NATIONALITY and HIGHER_EDU_LANG, leaving us with the following model: (1) 6 ACCEPT_NATIONAL_VARIETY ~ NATIONALITY + PROFICIENCY + POSITIVE_ATTITUDE_ENG + BELIEF_IMPORTANCE_ENG + AGE + EDUCATION + HIGHER_EDU_LANG + NATIONALITY:PROFICIENCY + NATIONALITY:POSITIVE_ATTITUDE_ENG + NATIONALITY:BELIEF_IMPORTANCE_ENG + NATIONALITY:AGE Results The German respondents displayed a higher mean rate of acceptance of a national variety of English than did the Dutch respondents: DE M=2.63, NL M=2.36. While this difference may appear to be small, it is highly significant (F(1,3550)=153.4, p<.001). Moreover, recall that on the four-point scale, 2=somewhat disagree and 3=somewhat agree (section 5.3.2), suggesting that the Germans respondents fall narrowly on the side of acceptance and the Dutch on the side of non-acceptance. In addition, as seen in the lower right-hand panel of Figure 1, the German respondents show a somewhat narrower spread than the Dutch respondents. This may, speculatively, suggest they are collectively more open to the notion of a German variety of English, while the greater heterogeneity Index of subjects Page 12 of 25 in Dutch attitudes may be indicative of greater collective ambivalence, with incipient acceptance among some individuals in addition to more traditionally exonormative attitudes. Turning to the results of the regression analysis, Table 2 presents all significant predictors and interaction effects, with corresponding slope estimates, p values and significance levels (a full overview is provided in Appendix 2). The table can be read as follows: with every unit increase (decrease) in the value of the predictor variable, the predicted score on the 1–4 scale of the outcome variable ACCEPT_NATIONAL_VARIETY increases (decreases) by the amount shown in the estimate column. To illustrate, consider the variable PROFICIENCY, which is implicated both as a main effect and in interaction with NATIONALITY. The main effect indicates that a one-point increase in self-reported proficiency is associated with a decrease of 0.099 in acceptance rate, while the interaction effect entails a further decrease of 0.098 for the Dutch respondents. In practice this amounts, for compatriots at opposite ends of the proficiency scale, to a 0.3 difference in acceptance rates for Germans and 0.6 for Dutch – considerable differences on a 1–4 scale. As Table 2 shows, all significant predictors have a negative effect on the outcome variable: for both nationalities, the higher respondents scored on the predictor variables, the lower they scored on the acceptance measure. Table 2: Overview of significant predictors of ACCEPT_NATIONAL_VARIETY Predictor (Intercept) PROFICIENCY POSITIVE_ATTITUDE_ENG BELIEF_IMPORTANCE_ENG HIGHER_EDU_LANG_bilingual_v_NationalLang HIGHER_EDU_LANG_English_v_NationalLang NATIONALITY_NL:PROFICIENCY NATIONALITY_NL:POSITIVE_ATTITUDE_ENG Signif. codes: ***=p<0.001, **=p<0.01, *=p<0.05, .=p<0.1 † Tukey-adjusted p values Estimate 3.466 -0.099 -0.055 -0.080 -0.096 -0.277 -0.098 -0.084 p <0.001 <0.001 0.024 0.042 0.014† <0.001† 0.049 0.021 *** *** * * * *** * * The first three rows of Table 2 draw attention to the composite variables PROFICIENCY, POSITIVE_ATTITUDE_ENG and BELIEF_IMPORTANCE_ENG, whose effects in predicting acceptance rates are visualised in Figure (the shaded areas around the regression lines represent 95% confidence intervals). The upper panel reflects the higher degree of acceptance among the German respondents as already seen, but also a negative effect of belief in the importance of English across both nationalities (b=-0.08, t=-2.03, p=.04): as this belief increases, so acceptance of the emergent national variety decreases, and this holds equally for the Dutch and German respondents. The lower two panels of Figure show the effect of the variables POSITIVE_ATTITUDE_ENG and PROFICIENCY on acceptance rates, respectively, whereby both variables have an interaction effect with NATIONALITY. That is, although acceptance rates again decrease among respondents with more positive attitudes to English (b=-0.06, t=- Index of subjects Page 13 of 25 2.25, p=.02) and higher proficiency levels (b=-0.10, t=-3.34, p<0.001), this negative effect is significantly stronger for the Dutch than the German respondents (Δb=-0.08 and Δb=-0.10, respectively; Table 2). Put another way, Dutch respondents with the most positive attitudes towards English and the highest proficiency levels appear to be significantly less open to the notion of a national endonormative variety than their German counterparts. Index of subjects Page 14 of 25 Figure 2. Effects of proficiency, positive attitude towards English and belief in the importance of English on acceptance rates of an emergent national variety. Index of subjects Page 15 of 25 Of the remaining variables, only HIGHER_EDU_LANG was identified as a significant predictor of acceptance rates (Table 2). Respondents whose higher education was bilingual as opposed to predominantly in the national language showed significantly lower acceptance of an emergent national variety of English (b=-0.10, t=-2.98, p=.003), and this effect was, unsurprisingly, even stronger for participants whose higher education was in English only (b=-0.28, t=-7.04, p<.001) (the latter indeed having the largest b value in the model). This is visualised in Figure , whereby for both nationalities acceptance rates progressively decrease as the amount of English in respondents’ higher education increases (from no higher education to higher education in the national language, to bilingual, to entirely in English) (F(7,3544)=44.87, p<.001). Posthoc Tukey tests revealed multiple pairwise differences in mean acceptance rates. Among the German respondents, all groups differed from one another except the two groups with the least and the two groups with the most English exposure (none vs bilingual t=3.24, p=.02; none vs English t=5.24, p<.001; NationalLang vs bilingual t=4.53, p<.001, NationalLang vs English t=6.31, p<.001). Among the Dutch respondents, there were significant pairwise differences between the English group and all other groups (English vs none t=4.19, p<.001, English vs NationalLang t=9.41, p<.001, English vs bilingual t=3.90, p=.002), as well as between the NationalLang and the bilingual group (t=4.67, p<.001). Finally, acceptance was lower among all groups of Dutch respondents compared to their German counterparts (NationalLang NL vs DE t=9.16, p<.001, bilingual NL vs DE t=-4.46, p<.001, English NL vs DE t=-4.04, p=.001) except for the groups with no higher education, which did not differ significantly from one another. In sum, Dutch respondents with any higher education, even that predominantly in their national language, showed lower acceptance of their emergent national variety than their German peers, but there was no difference in acceptance rates among Dutch and German respondents with no higher education. Figure 3. Effect of higher education language on acceptance rates of an emergent national variety. Index of subjects Page 16 of 25 7 Discussion and conclusion This study set out to explore the degree of openness towards emergent endonormative varieties of English in continental Europe. More than 4,000 participants from Germany and the Netherlands filled in an online survey on their use of and attitudes to English, with the German respondents displaying greater acceptance of their emergent national variety than the Dutch respondents. Our analysis thus confirms previous results from accent-evaluation research showing that Dutch raters are less lenient towards ‘Dutch English’ accents than are NSs, while German raters’ evaluations reflect a sense of solidarity with their compatriots (Davydova 2015, Van den Doel and Quené 2013; Hoorn, Smakman, and Foster 2014; Koet 2007). Further, it extends such findings by showing that evaluations are not shared across the board, but are mediated by a range of sociodemographic and attitudinal factors. Using multivariate regression analysis, we attempted to disentangle the effects of such factors in explaining the divergent acceptance rates between the German and Dutch samples. For instance, while higher self-reported English proficiency levels and greater exposure to English in higher education were associated with lower acceptance rates in both countries, they played an outsize role in the Dutch case since the Dutch respondents reported higher average proficiency rates than did the Germans, and twice as many Dutch respondents reported having an English-medium higher education (a proportion set only to increase given the rapid ‘Englishisation’ of Dutch higher education). Under the prevailing sociolinguistic conditions, therefore, it appears that the notion of a ‘legitimate’, locally constructed variety of English is liable to find greater acceptance in Germany than in the Netherlands. To compare our findings with earlier work on continental Europeans’ target models of English, we can convert our acceptance metric into a crude indication of endonormative orientation by calculating the proportion of respondents whose score was greater than the mid-point of 2.5 on the outcome variable ACCEPT_NATIONAL_VARIETY. Using this method, we find 63% (n=1,248) of German and 45% (n=896) of Dutch respondents falling on the side of acceptance of a local endonormative variety. These figures are considerably higher than those reported in much previous research, where the proportions of participants opting for local target varieties are typically under 10% (e.g. Leppänen et al. (2011) on Finnish English 7%, Risan (2014) on Norwegian English 6%). By contrast, Grau (2005), in her survey of German university students, found for her more indirectly phrased ‘it does not matter if [other speakers] can tell by your accent that you are German’ a 65% agreement rate, similar to that for the German respondents in the present study. This would seem to confirm that the formulation of the question matters. To illustrate the point further, let us look at two of the questionnaire items included in our acceptance measure. The statement ‘[Dunglish/Gerlish] is bad English’ makes use of stigmatised, hybrid variety labels, and when reverse scored, the mean score is 2; that is, the respondents somewhat disagreed that [Dunglish/Gerlish] is good English. By comparison, the mean score for the statement ‘As long as my English is good I don’t mind if it has a bit of [Dutch/German] flavour’ is 3 (somewhat agree); thus, it appears that the use of Index of subjects Page 17 of 25 more or less liberal wording can make the difference between apparent acceptance and non-acceptance. In addition to those variables identified as significant predictors of acceptance rates, interesting too are those found to be unimportant. As indicated in section 5.4, the factors sex and place of residence appeared to have no impact on acceptance rates, and were consequently not included in the model. By contrast, age and education level were included, as they were revealed by our preliminary oneway ANOVAs to have effects. First, acceptance rates increased with age in both national groups (F(1,3537)=22.89, p<.001), seemingly confirming the results of previous research which reported similar age effects (e.g. Leppänen et al. 2011). In the regression analysis, however, age appears to have been overshadowed by other effects; that is, given an older respondent with a high proficiency level, positive attitude towards English, strong belief in the importance of English or a bilingualor English-medium higher education, one would do well to predict a lower, rather than a higher, acceptance score. Second, there was an effect of education level (F(7,3544)=24.53, p<.001), whereby the acceptance rates of the lowest educated respondents, unlike those of their better educated compatriots, did not differ across the national groups. However, in the regression model the effect of education level appears to have been overshadowed by that of higher education language, suggesting that a key factor in driving down acceptance rates is, rather than education per se, the greater exposure to English that tends to go hand in hand with higher education. That sociodemographic variables such as age, education, sex and region seem in our study to be subordinate to other factors echoes findings by Coupland and Bishop (2007) which highlighted the primacy of attitudinal variables. Analysing the attitudes of over 5,000 British informants to different NS and NNS accents of English, they showed that while age, sex and region played a role to some extent (with young people being more tolerant, women being more positive and regional groups such as Scots showing high in-group loyalty), their influence was largely overridden by people’s attitudes: those who were more positively oriented to accent diversity gave more positive evaluations of different accents (and vice versa), regardless of age, sex and region. In our study, too, attitudinal factors came to the fore and seemed to cut across the population. Our results are suggestive of a spectrum of attitudes towards continental NNS varieties similar to that found in previous studies. Decke-Cornill (2002), in her comparative study of English teachers in a German Gesamtschule (comprehensive school) and a Gymnasium (grammar school), found that the two sets of teachers had different views on notions such as ELF and their usefulness in the classroom. While the Gymnasium teachers (who themselves had university degrees in English) were more exonormatively oriented and focused on NS culture, the Gesamtschule teachers (who did not have English degrees) prioritised successful communication over the approximation of a NS standard. Similarly, Mollin (2006) found that people with lower self-reported proficiency levels valued communication over correctness. Our findings help to flesh out this spectrum, which we illustrate below using comments given in response to the open-ended question in our survey. At one end of the spectrum are people with a strong orientation towards English (including high proficiency levels and a great deal of exposure to English) and, at the same Index of subjects Page 18 of 25 time, negative attitudes towards the notion of local NNS varieties (exemplified in (2)). At the other end are those whose less wholehearted orientation towards English appears to be accompanied by greater openness towards local varieties (3): (2) (3) Lots of people THINK they speak good English […], but then go and make terrible grammatical mistakes. […] Also with weird pronunciation. All ‘u’ sounds (must/fun) then have to be pronounced in a Dutch way. (Dutch female, 8 64, English-medium master’s-level degree, retired English teacher) I think Nederengels is a beautiful language! It should be appreciated more! (Dutch female, 28, Dutch-medium vocational bachelor’s degree) It is the second group that seems likely to lead the way in terms of lay acceptance of endonormative European varieties of English, and consequently it is they who may benefit most from raised awareness of the diverse ways in which English is used around the globe (in the form of both ELF and regional NNS varieties; see further Decke-Cornill 2002). Yet for the first group, too, such enhanced awareness would appear to be desirable. From the comments, it became clear that negative attitudes towards endonormative NNS varieties were often associated with ideologies of linguistic purity (4). L1 transfer was seen as tainting ‘proper’ English, and in many cases this perspective seemed to be tied up with an element of classbased prejudice (5). (4) (5) Gerlish is embarrassing and bad English to boot. [People should choose] [e]ither German or English. (German female, 39, high education level) Lower educated people often speak ‘Dutch English’. I […] find that toecurling. (Dutch female, 44, medium education level) Thus, a revamped curriculum approach to ELT in ELF/EFL contexts emphasising the cultural and linguistic value of drawing on one’s L1 resources may help to combat the stigmatising perception that ELF and NNS varieties are only relevant for the less well educated and/or less proficient in English (see also Erling 2004; 9 Risan 2014; Sing 2004). It may also go some way to reconfiguring the traditional notion of ‘achievement’, which may play a stronger part in shaping exonormative attitudes than animosity towards NNS varieties themselves (Risan 2014). Recall that English-medium higher education was the strongest effect in our model (section 6). English-language degree programmes are often viewed as prestigious, certainly in the Netherlands, where the entirely English-language ‘university colleges’ attract many of the highest school achievers; people who are accustomed to ‘achieving’ as it is traditionally defined (i.e. in terms of the native-speaker ideal). This confluence in the Dutch public imagination between nativelike English and the notion of achievement would seem to be reinforced by, on the one hand, the national self-image as a land with historically high language-learning standards, and on the other, a culture of inverse solidarity and criticism of what is perceived as 8 All quotes translated from German/Dutch by the first author. 9 Naturally, this is not to say that learners should be kept from striving for nativelikeness, if this is what they wish to achieve (see further Deterding 2013: 177 and the contributions in Dziubalska-Kolaczyk and Przedlacka 2005). Index of subjects Page 19 of 25 Steenkolenengels (‘coal English’, read: working class English) (Van den Doel and Quené 2013; Edwards 2014; Hoorn, Smakman, and Foster 2014; Koet 2007). By contrast, the link between English and achievement may arguably be less strong in Germany, where (academic) success is less obviously tied up with English. Whether this will change in the coming years with rising (elite) proficiency levels and the continuing internationalisation/Englishisation of the German higher education sector (Earls 2016: 16–17) remains to be seen. To conclude, we can assume there is a reasonable link between our respondents’ expressed attitudes and actual linguistic behaviour. In a grammaticality judgement study with these same Dutch respondents, acceptance rates of nonstandard uses of the progressive aspect (e.g. We are working together since 2005) were negatively correlated with high proficiency levels and positively correlated with permissive attitudes towards ‘Dutch English’ (Edwards 2016: 143– 155). Similarly, in Rindal (2010) and Rindal and Piercy (2013), the English pronunciation of Norwegian school students appeared to match up with their reported accent aims. Nonetheless, it is important to keep contextual factors in mind. Young continental Europeans in particular have been reported to play a ‘game’ of accent selection (Erling 2004; Rindal 2015; Risan 2014); Norwegian students, for instance, report that they ‘put on the Oxford accent’ to impress teachers, but in an ELF situation ‘then you would have spoken sort of Norwegian English, sort of’ (Rindal, 2015: 115). Likewise, many of our respondents indicated consciously varying their performance variety: (6) [My target variety] depends on where I am, in America American, in England British, and in the Netherlands I try to be neutral. (Dutch male, 27) This serves as a reminder that attitudes to language varieties are not static, but dynamic across settings and interlocutors according to the image one wishes to project. Future research would thus do well to consider under what conditions continental Europeans appear to be more or less permissive towards national NNS varieties, and how the attitudes identified here develop over time. References Akaike, Hirotugu 1974. A new look at the statistical model identification. IEEE Transactions on Automatic Control 19(6): 716-723. Ammon, Ulrich 2015. Die Stellung der Deutschen Sprache in der Welt. Berlin: de Gruyter. Ammon, Ulrich and Grant D McConnell 2002. English as an Academic Language in Europe. Frankfurt: Peter Lang. Berns, Margie 1995. English in the European Union. English Today 11(3): 3-11. Berns, Margie 2005. Expanding on the Expanding Circle: Where do WE go from here? World Englishes 24(1): 85-93. Berns, Margie, Kees De Bot and Uwe Hasebrink 2007. In the Presence of English: Media and European Youth. New York: Springer. Bonnet, Gérard 2002. The Assessment of Pupils’ Skills in English in Eight European Countries. European project commissioned by the European Index of subjects Page 20 of 25 Network of Policy Makers for the Evaluation of Education Systems. www.eva.dk/projekter/2002/evaluering-af-faget-engelsk-igrundskolen/projektprodukter/assessmentofenglish.pdf. Booij, Geert 2001. English as the lingua franca of Europe: A Dutch perspective. Lingua e Stile 36(2): 347-357. Bruthiaux, Paul 2003. Squaring the circles: Issues in modeling English worldwide. International Journal of Applied Linguistics 13(2): 159-178. Buschfeld, Sarah 2013. English in Cyprus or Cyprus English: An Empirical Investigation of Variety Status. Amsterdam and Philadelphia: John Benjamins. Coupland, Nikolas and Hywel Bishop 2007. Ideologised values for British accents. Journal of Sociolinguistics 11(1): 74-93. Dalton-Puffer, Christiane, Gunther Kaltenboeck and Ute Smit 1997. Learner attitudes and L2 pronunciation in Austria. World Englishes 16(1): 115-128. Davydova, Julia 2015. A study in the perception of native and non-native Englishes by German learners. Journal of Linguistics and Language Teaching 6(1): n.p. Decke-Cornill, Helene 2002. ‘We would have to invent the language we are supposed to teach’: The issue of English as Lingua Franca in language education in Germany. Language, Culture and Curriculum 15(3): 1-16. Deterding, David 2013. Misunderstandings in English as a Lingua Franca. Berlin: de Gruyter. Dziubalska-Kolaczyk, Katarzyna and Joanna Przedlacka (eds.) 2005. English Pronunciation Models: A Changing Scene. Berlin: Peter Lang. De Meerleer, Marieke 2012. Beliefs and attitudes towards English as a Lingua Franca: Native and non-native pronunciation: A Flemish and Walloon perspective. Unpublished master’s thesis, University of Ghent. De Swaan, Abraam 2001. Words of the World: The Global Language System. Malden, MA: Polity Press. DiStefano, Christine, Min Zhu and Diana Mîndrilă 2009. Understanding and using factor scores: Considerations for the applied researcher. Practical Assessment, Research and Evaluation 14(20): 1-11. Dronkers, Jaap 2013. Sluipenderwijs Wordt Nederland Tweetalig. De Volkskrant. www.volkskrant.nl/vk/nl/3184/opinie/article/detail/3476812/2013/07/17/Sluip enderwijs-wordt-Nederland-tweetalig.dhtml. Earls, Clive W. 2016. Evolving Agendas in European English-Medium Higher Education. Houndmills: Palgrave Macmillan. Edwards, Alison 2014. Mark Rutte heeft een accent. Get over it. Trouw. www.trouw.nl/tr/nl/4500/Politiek/article/detail/3624770/2014/03/29/MarkRutte-heeft-een-accent-get-over-it.dhtml. Edwards, Alison 2016. English in the Netherlands: Functions, Forms and Attitudes. Varieties of English Around the World (VEAW), vol. G56. Amsterdam and Philadelphia: John Benjamins. Erling, Elizabeth J. 2004. Globalization, English and the German university classroom. Unpublished PhD dissertation, Freie Universität Berlin. European Commission 2012. Europeans and their languages: Special Eurobarometer 386. http://ec.europa.eu/public_opinion/archives/ebs/ebs_386_en.pdf. Field, Andy, Jeremy Miles and Zoë Field 2012. Discovering Statistics Using R. London: SAGE. Index of subjects Page 21 of 25 FMKS/Verein für frühe Mehrsprachigkeit an Kindertageseinrichtungen und Schulen 2014. Bilinguale Kitas in Deutschland. Kiel: FMKS, http://www.fmks-online.de/_wd_showdoc.php?pic=1043 (Retrieved 16 June 2016). Forche, Christian 2012. On the emergence of Euro-English as a potential European variety of English – Attitudes and interpretations. Jezikoslovje 13(2): 447-478. Garrett, Peter 2010. Attitudes to Language. Cambridge, UK: CUP. Gnutzmann, Claus, Jenny Jakisch and Frank Rabe. 2015. Communicating across Europe: What German students think about multilingualism, language norms and English as a Lingua Franca. In Andrew Linn, Neil Bermel and Gibson Ferguson (eds.) Attitudes towards English in Europe. Berlin and Boston: Walter de Gruyter. 78-88. Görlach, Manfred 2002. Still More Englishes. Amsterdam and Philadelphia: John Benjamins. Grau, Maike 2005. English as a global language – What do future teachers have to say? In Claus Gnutzmann and Frauke Intemann (eds.), The Globalisation of English and the English Language Classroom. Tübingen: Gunter Narr. 261274. Groom, Chloe 2012. Non-native attitudes towards teaching English as a Lingua Franca in Europe. English Today 28(01): 50-57. Hilgendorf, Suzanne K. 2001. Language contact, convergence, and attitudes: The case of English in Germany. Unpublished PhD dissertation, University of Illinois at Urbana-Champaign. Hoorn, Myrthe, Dick Smakman and Tony Foster. 2014. Are we at sixes and sevens? Pronunciation assessment by lecturers of English in the Netherlands. In Rias Van den Doel and Laura Rupp (eds.) Pronunciation Matters: Accents of English in the Netherlands and Elsewhere. Amsterdam: VU University Press. 95-108. Jenkins, Jennifer 2007. English as a Lingua Franca: Attitude and Identity. New York: Oxford University Press. Jenkins, Jennifer 2009. World Englishes: A Resource Book for Students. second ed. Oxon, UK: Routledge. Kachru, Braj B. 1985. Standards, codification and sociolinguistic realism: The English language in the Outer Circle. In Randolph Quirk and Henry Widdowson (eds.) English in the World: Teaching and Learning the Language and Literatures. Cambridge, UK: CUP. 11-30. Kirkpatrick, Andy 2007. World Englishes: Implications for International Communication and English Language Teaching. Cambridge, UK: CUP. Koet, Antonius 2007. Polder English in Dutch ears: Empirical studies on the evaluation of the pronunciation of English as a foreign language. Unpublished PhD dissertation, University of Amsterdam. Kuteeva, Maria, Niina Hynninen and Mara Haslam 2015. ‘It’s so natural to mix Languages’: Attitudes towards English-medium instruction in Sweden. In Andrew Linn, Neil Bermel, and Gibson Ferguson (eds.) Attitudes towards English in Europe. Berlin and Boston: Walter de Gruyter. 89-96. Lambert, Wallace E. 1967. A social psychology of bilingualism. Journal of Social Issues 23(2): 91-109. Leppänen, Sirpa et al. 2011. National Survey on the English Language in Finland: Index of subjects Page 22 of 25 Uses, Meanings and Attitudes. Studies in Variation, Contacts and Change in English 5. www.helsinki.fi/varieng/series/volumes/05/ (September 10, 2012). Lewis, M. Paul, Gary F. Simons and Charles D. Fennig. 2015a. German, Standard. In Ethnologue: Languages of the World, Dallas, Texas: SIL International. http://www.ethnologue.com/language/deu (June 25, 2016). Lewis, M. Paul, Gary F. Simons and Charles D. Fennig. 2015b. Dutch. In Ethnologue: Languages of the World, Dallas, Texas: SIL International. http://www.ethnologue.com/language/nld (June 25, 2016). Maiworm, Friedhelm and Bernd Wächter 2014. The big picture. Wächter, Bernd & Friedhelm Maiworm (eds.). English-Taught Programmes in European Higher Education. The State of Play in 2014. Bonn: Lemmens, 25-62. Makoni, Sinfree and Alastair Pennycook (eds.) 2007. Disinventing and Reconstituting Languages. Clevedon, Buffalo and Toronto: Multilingual Matters. McArthur, Tom 1998. The English Languages. Cambridge, UK: CUP. Mesthrie, Rajend and RM Bhatt 2008. World Englishes: The Study of New Linguistic Varieties. Cambridge, UK: CUP. Ministry for Education Culture and Science 2015. Directe en indirecte instroom Hoger Beroepsonderwijs. Onderwijs in cijfers. www.onderwijsincijfers.nl/kengetallen/hoger-beroepsonderwijs/deelnemershbo/directe-en-indirecte-instroom-hbo (June 21, 2016). Mollin, Sandra 2006. Euro-English: Assessing Variety Status. Tübingen: Gunter Narr. Murray, Heather 2004. Teaching Euro-English: The results of a survey of Swiss teachers. ETAS Journal Spring: 1-5. Nykänen, Elisa 2015. English teachers’ attitudes towards non-standard English expressions: A Finnish context. Unpublished master's thesis, University of Eastern Finland. Preisler, Bent 1999. Functions and forms of English in a European EFL country. In Tony Bex and Richard J Watts (eds.) Standard English: The Widening Debate. London and New York: Routledge. 239-67. Preston, Dennis R. 1989. Perceptual Dialectology: Nonlinguists’ Views of Areal Linguistics. Dordrecht: De Gruyter/Foris. Proshina, Zoya 2005. Russian Engishes: Introduction. World Englishes 24(4): 437438. Ranta, Elina 2010. English in the real world vs. English at school: Finnish English teachers’ and students' views. International Journal of Applied Linguistics 20(2): 156-177. Ridder, Susan 1995. English in Dutch. English Today 11(4): 44-50. Rindal, Ulrikke 2010. Constructing identity with L2: Pronunciation and attitudes among Norwegian learners of English. Journal of Sociolinguistics 14(2): 240261. Rindal, Ulrikke 2015. Who owns English in Norway? L2 attitudes and choices among learners. In Andrew Linn, Neil Bermel and Gibson Ferguson (eds.) Attitudes towards English in Europe. Berlin and Boston: Walter de Gruyter. 109-121. Rindal, Ulrikke and Caroline Piercy 2013. Being ‘neutral’? English pronunciation among Norwegian learners. World Englishes 32(2): 211-229. Index of subjects Page 23 of 25 Risan, Maiken 2014. ‘Fake it till you make it’: Attitudes towards L2 Accents among prospective English teachers in Norway. Unpublished master’s thesis, Norwegian University of Science and Technology. RStudio Team 2015. RStudio: Integrated Development for R. Sing, Christine 2004. Curriculum ideals and classroom realities in the process of empowerment through Euro-English: A case study. LAUD Linguistic Agency Series A (General and Theoretical Papers, Paper No. 625). Starkweather, Jon 2012. How to calculate empirically derived composite or indicator scores. Benchmarks Online 2: 1-9. http://it.unt.edu/benchmarks/issues/2012/02/rss-matters (June 21, 2016). Tajfel, Henri and John Turner 1979. An integrative theory of intergroup conflict. In William Austin and Stephen Worchel (eds.) The Social Psychology of Intergroup Relations. Monterey, CA: Brooks-Cole. Van den Doel, Rias and Hugo Quené 2013. The endonormative standards of European English: Emerging or elusive? English World-Wide 34(1): 77-98. Van der Haagen, Monique 1998. Caught between Norms: The English Pronunciation of Dutch Learners. The Hague: Holland Academic Graphics. Wilkinson, Robert 1990. Translating into European English: Evolution and acceptability. In Marcel Thelen and Barbara Lewandowska-Tomaszczyk (eds.) Translation and Meaning. Maastricht: Euroterm. 323-335. Zegers, Vera and Robert Wilkinson 2005. Squaring the pyramid: Internationalization, plurilingualism, and the university. Paper presented to Biand Multilingual Universities – Challenges and Future Prospects, University of Helsinki. Index of subjects Page 24 of 25 APPENDIX 1: Full distribution of demographic variables Table 3: Distribution of respondents by nationality Nationality no. % German 1987 47.7 Dutch 1995 47.9 other 171 4.1 NA 9 0.2 Total 4162 100.0 Table 4: Distribution of respondents by sex male female NA Total German no. % 926 46.6 1053 53.0 8 0.4 1987 100.0 Dutch no. % 861 43.2 1126 56.4 8 0.4 1995 100.0 Table 5: Distribution of respondents by education level low medium high NA Total German no. % 158 8.0 156 7.9 1672 84.1 1 0.1 1987 100.0 Dutch no. % 59 3.0 690 34.6 1235 61.9 11 0.6 1995 100.0 Table 6: Distribution of respondents by medium of instruction in higher education national language English bilingual* other none† NA Total German no. % 1289 64.9 138 6.9 287 14.4 11 0.6 262 13.2 0 0.0 1987 100.0 Dutch no. % 1370 68.7 247 12.4 288 14.4 19 1.0 55 2.8 16 0.8 1995 100.0 *national language + English † no higher education Table 7: Distribution of respondents by place of residence city+ city German no. % 982 49.4 425 21.4 Dutch no. % 428 21.5 746 37.4 Index of subjects Page 25 of 25 town country abroad NA Total 388 19.5 59 3.0 69 3.5 64 3.2 1987 100.0 452 114 198 57 1995 22.7 5.7 9.9 2.9 100.0 APPENDIX 2: Complete output of regression model Predictor Estimate Std. Err. (Intercept) 3.466 0.153 NATIONALITY_NL 0.178 0.221 PROFICIENCY -0.099 0.030 POSITIVE_ATTITUDE_EN -0.055 0.025 G BELIEF_IMPORTANCE_E -0.080 0.040 NG AGE -0.001 0.001 EDUCATION_medium_v_l -0.028 0.060 ow EDUCATION_high_v_low -0.040 0.060 HIGHER_EDU_LANG_non -0.072 0.051 e_v_NationalLang HIGHER_EDU_LANG_bili -0.096 0.032 ngual_v_NationalLang HIGHER_EDU_LANG_Eng -0.027 0.039 lish_v_NationalLang NATIONALITYNL:PROFI -0.098 0.050 CIENCY NATIONALITYNL:POSITI -0.084 0.037 VE_ATTITUDE_ENG NATIONALITYNL:BELIE 0.053 0.054 F_IMPORTANCE_ENG NATIONALITYNL:AGE 0.001 0.001 Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ‘ 1 Residual standard error: 0.6244 on 3506 degrees of freedom (178 observations deleted due to missingness) Multiple R-squared: 0.1199, Adjusted R-squared: 0.1164 F-statistic: 34.12 on 14 and 3506 DF, p-value: < 2.2e-16 AIC: 6692.148 † Tukey-adjusted p values t 22.61 0.806 -3.340 -2.252 -2.030 p <0.001 0.420 <0.001 0.024 0.042 -1.373 -0.465 0.170 -0.672 -1.413 0.502 -2.977 -7.036 -1.970 -2.307 0.983 0.615 *** *** * * 0.642 0.473† 0.014† <0.001 0.049 0.021 0.326 0.538 ** *** * *