Students ’ attitudes to lecturers ’ English in English-medium higher education in Denmark

This study examines the evaluative reactions of uni vers ty students to their nonnative lecturers’ English skills in English-medium instruction, i.e. when English is used as a lingua franca in an academic context. In particular, we examine the relationship between perceptions of English languag e proficiency and perceptions of general lecturing competence (defined here as kn owledge of subject and teaching skills). Statistical analyses of 1,700 stu dent responses to 31 non-native English-speaking lecturers at a major business scho ol in Denmark revealed that the students’ perceptions of the lecturers’ English language proficiency is a significant predictor of their perceptions of the l cturers’ general lecturing competence and vice versa. We interpret this as a t wo-way relationship caused by speech stereotypes similar to those which have been d monstrated in socialpsychological experiments. This effect should be ad dressed when universities use student ratings to evaluate teaching in English-med ium content courses.


Introduction
English is used increasingly as the medium of instruction at universities and business schools around Europe (Van der Wende 1996; Wächter & Maiworm 2008: 10).In Denmark, many institutions of higher education are offering a steadily growing number of English-medium courses, especially at postgraduate level in the natural sciences and business programmes, as evidenced by the curricula at the Danish Technical University, the Faculty of Life Sciences at the University of Copenhagen, and at the Copenhagen Business School (CBS).Many so-called "prestige programmes", such as the Copenhagen Masters of Excellence programmes at the University of Copenhagen, are conducted entirely in English, partly in order to be able to attract the very best international students and partly to prepare graduates for a globalised job market (see e.g.Coleman 2006: 7ff;Wächter & Maiworm 2008: 67).At CBS there is a surging demand from students for English-medium programmes, and a proportionate decrease in interest in Danish-language programmes, even among Danish students (Sven Bislev, vice dean of education at CBS, personal communication).
This shift towards English-medium instruction at universities has been attracting attention because of its political and educational perspectives.The political interest mostly revolves around the fear of "domain loss", the risk that Danish can no longer be used to communicate scientific knowledge (Haberland et al. 1991;Jarvad 2001;Danish Ministry of Culture 2008;Gregersen et al. in press).With regard to the educational perspective, on the other hand, the major concern seems to be what might be termed "content loss", i.e. that learning is impaired.
To this can be added the related issues of attitude and image, which have not received much attention in the literature on English as an academic lingua franca.Interviews with directors of study and deans of education reveal that there is no shortage of anecdotes about the poor English skills of some teachers (and students).Recent studies report similar comments or responses from students in surveys conducted in Sweden (Bolton & Kuteeva 2012), Austria (Tatzl 2011) and Norway and Germany (Hellekjaer 2010).There is thus good reason to attempt to shed further light on this issue through more systematic and controlled investigations of students' evaluational reactions to their teachers' English skills and of the potential effects this may have on the image of both individual lecturers and the institution as a whole.
The issues of content loss and image, or what Lavelle (2008) refers to as "credibility", are clearly interdependent.He notes that "age, gender, appearance and nationality each can affect student perceptions of teacher credibility, and so too can language proficiency when English is the instructional lingua franca", and he comments that "it is difficult, perhaps impossible, to know definitively how students assess credibility" (Lavelle 2008: 143).According to Lavelle, repeated errors, such as consistently mispronouncing terms or expressions for key concepts in a lecture, or stigmatised L1 features can "erode teacher credibility" and lead to students paying more attention to linguistic errors than to the message of the lecture (Lavelle 2008: 144).This, of course, must be expected to lead to content loss and reduced learning.
This notion of credibility, or image, is consistent with a general finding in the literature on the social psychology of language.Social psychological experiments have demonstrated that listeners may judge speakers negatively both on indicators of social attractiveness and on indicators of competence, based merely on variation in accent-not only with respect to regional accent (e.g.Giles 1970), but also to native vs. non-native accents (Coupland & Bishop 2007;McKenzie 2008).Such evaluations are based, at least to some extent, on stereotyping, i.e. on "over generalizations that are applied to any ethnic group member regardless of his or her individual characteristics" (Grant & Holmes 1981: 107).In the area of linguistic stereotyping, it was shown as early as the 1960s that listeners make judgements about speakers' social attractiveness and competence from hearing even fairly short samples of speech in experimental settings, and that these judgements reflect general attitudes towards the group of which the speaker is judged to be a member (Lambert et al. 1960;Lambert 1967).The same social psychological mechanisms can be expected to have an influence on how teachers are perceived in the classroom.In other words, it is possible that variation in teachers' linguistic abilities may invoke stereotyped impressions of their overall competences and thereby have an impact on whether students perceive them to be competent, not only linguistically but also academically and/or pedagogically.
Universities in Denmark are beginning to address the issue of the lecturers' English skills in different ways.One of these is to simply ask the students whether the teachers' English is adequate for the purposetypically as part of the course evaluation.This inevitably raises the question of whether we can actually trust students' assessments of their lecturers' English.Do their ratings accurately reflect the lecturers' English skills in that specific context, namely teaching graduate and undergraduate courses in higher education?Our expectation is that they do not.Some studies have shown that judgements about language can be influenced by the listeners' knowledge (whether false or accurate) about the speaker's "status", in terms of the speakers' accomplishments in a number of different tasks and educational or vocational background (Thakerar and Giles 1981;Ball et al. 1982;Rubin 1992).And Orth (1982) found a very low correlation between student ratings and the ratings of a group of 12 experienced EFL teachers of the speaking proficiency of 10 foreign teaching assistants.However, student evaluations, including evaluations of their lecturers' English language skills, are currently used as a measure of success (or failure) of courses and lecturers at universities.It is therefore important to learn more about the accuracy and potential biases of these evaluations.
At least one study, Rubin & Smith (1990), has demonstrated that when students perceive lecturers' speech to be highly accented, they also judge them to be poor teachers.This study was concerned with International Teaching Assistants at North American universities-a subject which has received a lot of attention, not least in the 1980s and 1990s (Orth 1982;Brown 1988;Gill 1994;Rubin 1992; Plakans 1997).In most of these studies the listeners (students) are either predominantly or exclusively native speakers of (North American) English.It therefore still has to be determined to what extent their findings are valid for the English as a lingua franca (ELF) context which we find in European universities offering English-medium content courses.
Based on the above discussion we have formulated two hypotheses which are in fact mirror images of each other: 1) Students' perceptions of their lecturers' English language skills influence their perceptions of the lecturers' general lecturing competence (knowledge of their field and teaching skills); 2) Students' perceptions of their lecturers' general lecturing competence influence their perceptions of the lecturers' English language proficiency.
We suggest that the relationship between perceptions of English language proficiency and perceptions of general lecturing competence is one of mutual influence.The issue of directionality cannot be determined by the statistical tests, however, but is treated in some detail in the discussion.
In addition to the main hypotheses, we explore the extent to which the hypothesised relationships between the two variables are influenced by a range of background variables which are linked to either the "object" (the lecturers) or the "subjects" (the students): • Student variables: gender, year of study, academic results, selfassessed English skills, L1 • Lecturer variables: gender, age, L1, teaching experience, job category

Method
Evaluative reactions to language, most typically accent, are traditionally examined using the matched-guise technique (Lambert et al. 1960;Lambert 1967) or a variant of this design, which allows for direct control of the background variables.However, we wanted to see if such attitudinal effects, specifically the ones listed as hypotheses 1 and 2 above, could be detected in a more ecologically valid, and consequently less controlled, design that is by and large identical with the method we normally use to collect student evaluations of courses.The aim was to show that the use of appropriate statistical techniques allows us to test our hypotheses even under these less controlled conditions, thus paving the way for developing and applying a valid procedure for future research of this type.
A combination of questionnaires and audio recordings, all collected at CBS, was therefore used to answer the issues outlined above.Audio recordings were made of 31 45-minute lectures, in which the teacher gave a 20-30 minute presentation, usually followed by a brief discussion or questions from the students.At the end of each lecture, separate questionnaires were distributed to students and teachers, who filled them in on the spot.
The student questionnaire contained 38 items on attitudes to the lecture, the teacher, and the teacher's command of English.The first three items served to gauge the students' global and immediate responses to each of these three aspects: they were asked to rate, on a scale from 1 to 5, (1) this lecture, (2) the teacher and (3) the teacher's English.Of the remaining 35 items, six were excluded from further analysis (four because they covered "interaction", of which there turned out to be little, and two for technical reasons).The final 29 items were phrased as statements to which the students were required to respond on a four-point Likert scale ranging from "strongly agree" to "strongly disagree" (with an additional option of "don't know").The responses were subjected to Rasch analysis (Rasch 1960), which confirmed that a single scale could be constructed for each section.These scales and the statements on which they are based are shown in Figure 1.

Rasch scale label
Items Lecture Content.
Statements about the level, usefulness and general appeal of the lecture/topic I found the academic level appropriate I found the content of this lecture difficult I found the lecture useful in advancing my knowledge I learned a lot from this lecture I found that this lecture improved my knowledge of the area I found the lecture interesting Lecture Structure.
Statements about the structure and general presentation of the lecture I found the lecture well-structured I found the lecture well-presented I found the lecture well-organised I found the lecture easy to follow

General Lecturing
Competence.
Statements about the lecturer's knowledge of the field and ability to communicate in an effective and engaging manner.*I found the teacher very knowledgeable about the subject I found the teacher to be a real expert in this field I found that the teacher was good at explaining the subject I found the teacher engaging I found that the teacher kept my interest I found the teacher enthusiastic about the subject I found the teacher pleasant English Language Proficiency.
Statements about various aspects of the lecturer's English, such as grammar, fluency, vocabulary and pronunciation.
I found the teacher's English fluent I found the teacher's English easy to understand I found that the teacher often struggled to find the appropriate words I found that the teacher was good at re-phrasing the meaning of key concepts and terms I found that the teacher had adequate vocabulary to describe the subject matter well I found that the teacher had too many long hesitations I found that the teacher had good English grammar I found that the teacher made basic grammatical errors I found that there were too many unfinished sentences I found that the teacher has good English pronunciation I found that the teacher sounds like a native speaker of English I like the teacher's accent Figure 1.Rasch scale labels and questionnaire items * Originally two separate sections, but they were combined into one Rasch scale.
Most of the statements express positive attitudes, with negative statements only in the scales English Language Proficiency and (in one case) Lecture Content.In hindsight it would have been better to have positive and negative statements distributed more evenly across the four scales.
We stress that the scale English Language Proficiency was not designed to measure the lecturers' communicative competence, in terms of "getting the message across" and interacting effectively with students while lecturing in an ELF context.Both the anecdotal evidence of students' complaints about their lecturers' English (including student evaluations) and comments in surveys on English-medium instruction (Klaassen 2003;Bolton & Kuteeva 2012) suggest that issues of language proficiency can trigger negative attitudes towards the lecturers.The scale was therefore intended to be a measure of the students' perceptions of the lecturers' proficiency in English.
The attitude statements were followed by a section with questions on the students' biodata, including age, gender, academic results in upper secondary school, nationality, native language, previous exposure to English, and self-assessment of English skills (both general proficiency and proficiency in connection with specific academic activities).
The teacher questionnaire contained questions on the teachers' own presentations and their perceptions of the students' motivation and interest in the specific class.They were also asked to provide information about their preparation for giving the lecture in English, e.g.checking terminology, pronunciation and grammar.In addition, questions were included on whether they thought they would have been able to perform better (on a number of parameters) in their native language.Finally, they were asked to provide the following background data: job category, age, gender, nationality, native language(s), experience using English in English-speaking countries, teaching experience and self-assessment of English skills.Only the background data from the teacher questionnaire are included in the analyses in this paper; details of the teachers' selfassessment can be found in Jensen et al. (2011).
The sample was drawn from 12 English-medium degree programmes at CBS-six BA/BSc programmes and six MA/MSc programmes within the fields of economics, politics, management and business administration.In total, 31 lectures were included in the study, 21 of which were at undergraduate level and 10 were at postgraduate level.
The 31 lecturers comprised seven women and 24 men.In terms of nationality, 24 were Danes and seven were non-Danes.
Altogether, 1,707 student questionnaires were collected, but the actual number of individual respondents is smaller than this, since some students attended two sessions.All student responses were anonymous, and the response rate was close to one hundred per cent.This high rate was achieved because we opted for handing out the questionnaires in class rather than using online questionnaires.The number of responses per session varied between 20 and 183, with a mean of 55.Approximately 60% of the respondents were Danes, whilst the remaining students came from a variety of other nationalities.
The spoken English proficiency of the non-native speaking (NNS) lecturers was assessed by three experienced EFL examiners, referred to below as the "EFL examiners".All three examiners were trained EFL teachers with extensive experience of assessing English in an ELF context, as teachers and testers of diplomats at the School of Languages at the Ministry of Foreign Affairs and as examiners in the CBS Project in Language Assessment for Teaching in English (PLATE, Kling & Hjulmand 2008).They were given access to the audio recordings of the lectures, but no other information about the lecturers was revealed to them.All three examiners assessed all 31 lecturers; after they had given their initial ratings independently of each other, they met and discussed their assessments before arriving at one joint (or "communal") rating for each lecturer.The examiners were asked to relate their assessment to the Common European Framework of Reference global scale (CEFR: 24) with the added instruction that they should indicate whether a performance was, for instance, a "high C1" (C1+), a "low C2" (C2-), etc.These ratings were then coded numerically for subsequent statistical analysis.The lowest rating was a B1+ and the highest a C2-, corresponding to the values 9 and 16, respectively, on the numerical scale.

Effect of perceived English skills on perceived general lecturing competence and vice versa
The effect of perceived English skills on perceived general lecturing competence cannot be tested directly with our data, since we did not control for variation in actual competence (as determined by some external measuring instrument).Obtaining such a measure would be difficult, partly because our variable General Lecturing Competence also includes statements on the lecturer's "knowledge about the subject".However, by examining the effect that certain control variables, or independent variables, have on the two dependent measures (the students' ratings of the lecturer's English and general lecturing competence, respectively), it is possible to gain indirect evidence of the connection between those variables.We constructed two mixed effects regression models, one with the students' ratings of the lecturers' general lecturing competence as the dependent variable and one with the students' ratings of the lecturers' English language proficiency as the dependent variable.Mixed effects models allow for the control of random variation between the levels of certain sampled variables-in our case the lecturer and the students' nationality-through the inclusion of random effects.This means that we can assess the effects of our variables of interest over and above such variation between the sampled variables (Baayen 2008: 241ff).Both models were fitted using the lme4 package in the statistical computing environment R. We applied a forwards stepwise approach to fitting the models, building the models by adding one variable at a time-starting with the most control-oriented variables and finishing with the most interesting variables in terms of the tested hypotheses.At each step in this process the contribution of the included variables was evaluated.Variables which contributed significantly to the model were kept while those which were nonsignificant were excluded.

Predicting perceived general lecturing competence
The following variables were tested in the order in which they are presented below: Most of the above-mentioned student variables have in previous studies been found to have a significant effect on evaluations of lecturers' competence.Ling and Braine (2007) found an effect of year of study on undergraduate students' attitudes to NNS English teachers in Hong Kong, and Plakans (1997) found an effect of both gender and "year of enrolment" on students' attitudes towards International Teaching Assistants (ITAs).McKenzie (2008) found an effect of both gender and self-assessed proficiency in English on perceptions of speaker "competence" in a verbal-guise experiment, such that the female Japanese informants rated the speakers more positively than did the male informants, and informants who assessed their own English higher gave more favourable ratings to (some of) the speakers in the experiment.Finally, Carrier et al. (1990, cited in Plakans 1997) found that NNS undergraduates gave higher ratings to ITAs than did native Englishspeaking students.However, the findings obtained in earlier investigations were not confirmed in the present study, where none of these variables were found to have any effect on the perception of general lecturing competence.
With regard to the lecturer variables, there was a significant effect of gender before job category was included.It should be pointed out that only seven of the 33 lecturers are women, including the only two PhD students in the survey.These two female PhD students both received fairly low ratings, which may be an effect of both the gender and job category variables or some combination of the two.However, in our data, the variance was explained better by the job category variable.Adding the job category variable also meant that the effect of the variable teaching experience in English was no longer significant.
The final model revealed significant effects of four explanatory variables, namely the factor job category and the co-variates Lecture Content, Lecture Structure, and English Language Proficiency.In addition to these four explanatory variables, random intercepts were included for the variables nationality and lecturer, and random slopes were included for the variables Lecture Content and English Language Proficiency.The residuals were inspected for the initial model, outliers were removed from the data set using a cut-off point of 2 standard deviations (6.6% of the responses), after which the model was refitted.The final model explains 75% of the variance in the trimmed data set.Baayen et al. 2008).The second column shows the mean estimate for 10,000 MCMC samples, while the third and fourth columns show the credible intervals within which 95% of these MCMC estimates lie (corresponding to 95% confidence intervals).For the co-variates, if the number in the Estimate column is positive, it means that a higher score on this variable is associated with a higher score on the dependent variable, here General Lecturing Competence.A negative estimate indicates that a higher score on this variable is associated with a lower score on the dependent variable.For the factor job category, each estimate indicates the difference in ranking between a reference level, in this case full professor, and the level of the factor specified in the relevant line.
It appears from Table 1 that there was a significant effect of job category on General Lecturing Competence.This factor has five levels, four of which can be ranked.The reference level here is "full professor" (not shown in the output), and there was an increasing negative effect of job category with each level lower in the hierarchy: associate professor, assistant professor, PhD student.The difference between full professors and associate professors was not significant, but assistant professors and PhD students were rated significantly less favourably than full professors.Part-time lecturers fall outside this hierarchy, since they are generally recruited both among recent graduates and among high-level executives from the business community; this is reflected in the fact that the difference between full professors and part-time lecturers is nonsignificant.
The measures Lecture Content and Lecture Structure were also both significant predictors.In other words, lecturers whose lectures were evaluated more positively in terms of content and structure also received more positive evaluations in terms of their general lecturing competence.
However, the result that we are most interested in here is the last row, which shows that there was a significant effect of English Language Proficiency on General Lecturing Competence.The effect was positive, as expected, which means that lecturers whose English was perceived as better were also perceived as having higher general lecturing competence, even after the other explanatory variables had been taken into account.

Predicting evaluations of lecturers' English
Having established that students' perceptions of the lecturers' English have predictive value for their perception of the lecturers' general lecturing competence, we turn to the analysis of which individual factors have an effect on the students' ratings of the lecturers' English.Most of the variables which were entered into this model are the same as for the previous model, but a few additional factors were tested.Most importantly, we included the ratings from the EFL examiners as an expression of the lecturers' actual, rather than perceived, proficiency in spoken English.Obviously, the EFL examiners' ratings are also subjective, but we believe that the communal ratings of a panel of highly experienced EFL teachers and examiners provide the best possible approximation to an "objective" measure of actual proficiency.To the extent that this assumption is valid, the remaining variables in the model can be expected to capture the variance in the students' ratings which is not directly related to language skills.
The following variables were tested in the model, again in the order in which they are presented below: There was a significant effect of age before job category was introduced into the model.The contribution of job category as well as that of Lecture Content became non-significant once General Lecturing Competence was included.It should be noted that there is a fairly complex relationship between teaching experience in L1, teaching experience in English and the amount of time spent in an Englishspeaking country, which will not be examined further in this paper.
The model was trimmed in the same way as the model above, but using a cut-off of 2.5 standard deviations (removing 1.8% of the data points).The estimates and associated values of the fixed effects that turned out to be significant in the final model are presented in Table 2. Table 2 shows that the EFL examiner ratings were found to be a significant predictor of the students' ratings of the lecturers' English: lecturers who received a higher rating by the EFL examiners were also evaluated more positively by the students.In addition, there was a significant positive effect of Stay Abroad (log) (the log-transformed number of months a lecturer had spent abroad, using English as the working language).This is perhaps surprising in one respect-if a prolonged stay in an English-speaking country results in improved proficiency in English, then this improvement should have been captured and explained in the model by the EFL examiner ratings.One explanation may be that such an improvement includes areas which are not covered well by the CEFR scale or which require visual contact, for example greater confidence as reflected in body language or facial expressions, or the use of visual aids such as slides (recall that the examiners only had access to audio recordings of the lectures).The students' perceptions of the Structure of the lecture were also found to be a predictor of ratings of English proficiency, so the lecturers' English was evaluated more positively in lectures which received higher scores in terms of their structure.A significant effect was found for the variable Lecturer/Student Shared L1, which is a two-level (yes/no) factor indicating whether the lecturer and the student have the same L1.In most cases where this is the case, both have Danish as their L1.The effect was negative (estimate = -0.3737),which shows that students who shared a lecturer's L1 rated his or her English lower than when this was not the case.Given that Danish is the shared L1 in almost all such cases, the result essentially shows that Danish students gave lower ratings to Danish lecturers.
Finally, and importantly, the analysis also showed a positive effect of General Lecturing Competence, the measure which was used as the dependent variable in the first model.In other words, there is an effect in both directions between the variables English Language Proficiency and General Lecturing Competence.Lecturers who are evaluated positively on one of these variables are also evaluated positively on the other variable.The possible interpretations of these findings are discussed below.

Discussion
The mixed-effects analyses reveal an effect of perceived English skills on perceived general lecturing competence and vice versa.However, owing to the design of our study, this is essentially a correlational analysis which cannot explain the causality of these effects.There seem to be at least two plausible explanations.The first possibility could be that the two underlying skills are indeed correlated, in the sense that there is a tendency for lecturers with better skills in spoken English to also have higher general lecturing competence, and that the student ratings simply reflect this.The second possibility is that all or part of the effect may be caused by attitudes, or stereotyping.This attitudinal effect could be monodirectional, in the sense that either perceptions of language skills affect perceptions of general lecturing competence or perceptions of general lecturing competence affect perceptions of English skills, or it could be bidirectional so that there is a reciprocal influence between the two types of competence.In the following each of these possibilities will be discussed.
The first of these possible explanations, that lecturers with better English skills generally have higher general lecturing competence, can be examined, albeit indirectly and in part, by comparing the relationship between the students' perceptions of the lecturers' general lecturing competence and a) their perceptions of the lecturers' English skills and b) the EFL examiners' ratings of the lecturers' English skills.The students' ratings of English skills and general lecturing competence are very highly correlated (rho = 0.791, p < 0.001).If the students' ratings reflect a genuine tendency for correlated skill level, then we would expect to find a similar correlation if we replace the students' ratings of English skills by those of the EFL examiners.However, a Spearman's rank order correlation analysis between the examiner ratings and the mean values of the students' ratings of general lecturing competence reveals only a low to moderate, though just significant, correlation (rho = 0.363, p = 0.045).The difference between the students and the EFL examiners strongly suggests that the effect cannot merely be a reflection of a genuine relationship between lecturing skills and English language skills.Rather, it is likely that the low, albeit significant, correlation between the EFL examiners' ratings of English skills and the students' ratings of general lecturing competence is caused by the impression that the students' perceptions of the lecturers' English skills has left on their perceptions of general lecturing competence.Unfortunately, as stated earlier, we do not have an assessment of the lecturers' actual lecturing competence with which the EFL examiners' ratings of English skills can be compared, and it is not obvious how such an assessment could be obtained.
On the basis of the above, we find it reasonable to reject the first of the suggested explanations, that the effects that emerged from the statistical analyses reflect a correlation of actual skill level within the two areas, and turn to the second possibility, that the results are caused by stereotypical attitudes.Here the main problem is to establish the direction of the effect.By nature of the experimental design used in this study, it is not possible to draw any conclusions based on direct evidence.However, it seems reasonable to expect the effect to be bidirectional.The effect of accent variation on perceptions of competence and social attractiveness is well-documented in controlled (often matched-guise) experiments (Giles 1970;Rubin 1992;McKenzie 2008) and has also been documented in a previous study on university students' evaluations of the speech of foreign teaching assistants (Orth 1982).Gill (1994) also found that standard accented (American) students gave more favourable ratings to standard accented teachers than to nonstandard accented teachers (British and Malaysian).The reverse effect seems to have attracted much less attention, but Thakerar and Giles (1981) found that evaluations of British English speakers in a matchedguise experiment varied with the information they provided about the speakers (after the informants had listened to the recording).For example, pronunciation was deemed to be more "standard" when informants were provided with "high status" information about the speaker and less standard when they were given "low status" information about the speaker (compared with a control group).In a related study, Rubin (1992) found that expectations of speaker nationality can affect not only evaluations of the speaker's language but also listener comprehensibility.This effect of perception of group identity on evaluations of language is sometimes referred to as reverse linguistic stereotyping (Kang & Rubin 2009).It is perhaps not surprising then that the relationship between these factors seems to be two-way rather than one-way.If it is based on stereotyping, as we assume it is, then it seems natural that speakers who have been assessed, rightly or wrongly, as particularly competent, based on other evidence, should be perceived as having better English language proficiency.
As we stated in the Introduction, most previous research on students' attitudes to their lecturers' English has focused on the North American situation as a response to native speaker students' complaint about their International Teaching Assistants.But the central findings of those studies would appear also to be valid for the English as a lingua franca situation that we find in European universities, including the one examined in this study.Students' attitudes towards their lecturers' general lecturing competence are affected by their perceptions of the lecturers' proficiency in English.In the light of recent research on ELF, a different result might have been expected.Some findings on ELF are summarised by Jenkins in a recent article entitled "Accommodating (to) ELF in the international university", contrasting ELF with English as a Foreign Language (EFL).Jenkins states that "ELF takes a difference perspective as contrasted with the deficit perspective of EFL" and writes that ELF speakers "innovate in English […] code-switch […] make skilled use of the accommodation strategy of convergence [..] [a]nd in all of this, they prioritise communicative effectiveness over narrow predetermined notions of 'correctness'" (Jenkins 2011: 928).Furthermore, a study of 22 undergraduate physics students by Airey (2009) reports that, when asked directly, students say that "there were very few differences between being taught in English or in Swedishthey believed that language played an unimportant role in their learning"; the students "suggest that the limiting factor for their learning is the lecturer's ability to mediate physics knowledge in the chosen language" (Airey 2009: 108, 78).(The students' actual behaviour did not fully support their claims, though, and varied with the language of instruction.)Finally, Björkman (2010) highlights the importance of frequent use of pragmatic strategies in ELF communication and refers to a finding in Hellekjaer (2010) that the "lectures that were rated higher and reported as 'most comprehensible' were those which had made use of a number of interactive features, e.g.questions" (Björkman 2010: 86).She adds that "high proficiency does not ensure communicative effectiveness" (Björkman 2010: 87).
The above statements about ELF in higher education all point to communicative effectiveness as the primary consideration in ELF interactions in higher education, downplaying the importance of good language proficiency in the traditional sense of correctness according to standard native norms.Yet, our study shows that-even after the other explanatory factors, such as the lecturer's teaching experience, age, gender and the students' perceptions of lecture content and structure have been taken into account-students still seem to be influenced by their perceptions of language proficiency as regards the use of grammar, vocabulary and pronunciation.And this is in fact in line with another finding in Hellekjaer (2010) than the one mentioned above, namely that "[t]he perhaps most important source of lecture comprehension difficulties found in the present study was due to unclear pronunciation" (Hellekjaer 2010: 24).
In this study, we have been concerned only with students' perceptions of their lecturers' competences; we have not examined whether the lecturers were in fact effective communicators and lecturers, so this remains to be investigated.Two scenarios seem to present themselves depending on the outcome of such an investigation: either English language proficiency turns out to be highly correlated with communicative effectiveness, which would justify the connection that has been established here between perceived language skills and perceived general lecturing competence.Or language proficiency turns out not to be correlated with communicative effectiveness, which would indicate that students' explicit evaluations of their lecturers' English do not provide useful information about the lecturers' ability to teach in English.In ordinary course evaluation forms, questions about the lecturer's English are usually quite similar to one of the first items in our questionnaire, namely "on a scale from 1 to 5 […], how would you rate the teacher's English?", which in our study correlates extremely highly with our measure English Language Proficiency (r = 0.98, Pearson's product-moment correlation).If perceptions of language proficiency turn out to be poor approximations of actual communicative effectiveness while lecturing, new methods of evaluating lecturers' English will have to be developed and implemented.
To what extent can we expect these results to generalise to other institutions in other countries?At least two issues need to be addressed to answer this question: the role of the setting itself (graduate programmes at a business school in Denmark) and the composition of the sample with regard to cultural and linguistic background.The first of these issues would have to be examined empirically by repeating the study at universities in other countries.The second would require a few changes to the design of the study, so that variables relating to cultural and linguistic background are collected in a more controlled and systematic manner.However, we did collect two variables which are relevant in this context: nationality and first language(s) of both students and lecturers.Student nationality emerged as a significant random effect in both our statistical models, which indicates that we arrive at a better estimation of the observed (fixed) effects when the students' nationality is taken into account.However, although about 40% of the respondents were non-Danish nationals, we cannot compare the responses of groups with different nationalities because of the way these were sampled, namely randomly and with very varying group sizes.We found an effect of language background, though, in the sense that students who shared the lecturers' L1 rated the lecturer's English lower than did the other students.Since the shared L1 was almost always Danish, it is a matter for future research to determine whether similar effects can be observed for students from other cultural and/or linguistic backgrounds.

Conclusion
Our study has illustrated that the statistical techniques adopted above enable us to test the hypotheses formulated in this paper under less controlled conditions.The students' perceptions of the lecturers' English were found to be influenced by their perceptions of the lecturers' general lecturing competence.And perhaps more importantly, we found that the reverse was true as well, that the students' ratings of the lecturers' general lecturing competence were influenced by their perceptions of the lecturers' English skills.This has potential consequences both for the individual lecturers and for the academic institutions.Lecturers whose English skills are perceived as problematic by their students risk being downgraded on their general lecturing competence, i.e. they are perceived as less competent teachers, which may have serious consequences, for example for lecturers seeking tenure.
Because of the mutual effect the two factors can have on each other, it may be difficult for both the lecturers and for the institution to determine whether any problems noted by students with either language skills or general lecturing competence can really be attributed entirely to one of these competences only.Crucially, our results indicate that universities should be aware that the English skills of their teaching staff will be reflected not just in the students' perceptions of language skills but also in their perceptions of the lecturers' overall lecturing competence, which may have a negative impact on the impression the students have of the academic level of the institution as a whole.In addition, depending on the method used to obtain evaluations of lecturers' English skills, it is likely that those evaluations will not be a reliable measure of the actual communicative competence which is required to be a successful lecturer in an English as a lingua franca setting (cf. Björkman 2010(cf. Björkman , 2011)).
There is evidence that at least some lecturers are aware of the consequences that their English skills may have on students' perceptions of their qualifications in general.Tange (2010: 143) reports how difficult it can be for lecturers to be 'subjected to student criticisms' and how some defend themselves by questioning the students' ability to judge their fluency, while others describe how it has caused them to drop several points in student assessments as compared with evaluations on the basis of classes conducted in their first language.Individual lecturers respond to student attitudes and expectations in different ways, as can be illustrated by the following two cases.In May 2010, the University Post, the English-language version of the University of Copenhagen newspaper, published an open letter written by a Polish lecturer to some of his students after he had received a negative evaluation; he stated that their criticism was unacceptable and ill-informed since they believed that when a lecturer's English is not good according to their standards, then the whole quality of the lecturer's teaching [and] the overall educational dimensions of the course are insufficient and bad (University Post, 1 May, 2010, http://universitypost.dk/article/documentation-letter-students-sociology-lecturer).This lecturer's perception of the situation obviously echoes the general finding we have presented here.A different approach, from CBS2 , may provide a more constructive way forward in dealing with situations of this type.Here, a lecturer who was aware of his own weaknesses with English encouraged his students to assist him in finding the right words and to ask him to clarify matters whenever something was unclear.Combined with the lecturer's considerable pedagogical skills, this helped defuse a potentially problematic situation and let the students contribute actively in creating successful communication and effective learning.Other researchers have recommended different approaches, not only for the individual, but at the institutional level.For example, Vinke et al. (1998) recommend screening of lecturers' English competence, offering courses that focus on the use of English for teaching content courses, assigning lecturers with previous experience of teaching in English to EMI courses and easing the workload of lecturers who start teaching in English.This should improve the quality of the EMI which the students receive.However, while such strategies would push the general level of EMI upwards and thereby reduce the potential threat to the overall image of the institution (and hopefully improve student learning), they do not address the issue of stereotyping.Even if the general level of the lecturers' English is raised, there will still be cases where students are taught by lecturers whose English is in some way "substandard" or non-standard.And although research in English as a lingua franca has shown that native speaker standards, or norms, are not particularly relevant in ELF interaction, such norms continue to exert influence on students' perceptions of the interactions.It is therefore advisable that universities-when interpreting the results of a course evaluation-carefully consider the interplay between students' perceptions of the lecturer's language skills and their perceptions of course content and structure and the lecturer's teaching skills: lecturers who receive low ratings on language and teaching skills are not necessarily seen as both poor teachers and as having poor language skills but are perhaps seen as poor teachers because they are perceived to have poor language skills (or vice versa).

Table 1 .
Summary of the mixed-effects analysis of variables predicting General Lecturing Competence.The model also includes random intercepts for lecturer (SD estimated at 0.4125) and students' nationality (SD 0.0638), and by-lecturer random slopes for Lecture Content (SD 0.1447) and English Language Proficiency (SD 0.0830).The estimated coefficients and related values for the fixed effects in the final regression model are presented in Table1, with an indication of significance level as determined by Markov chain Monte Carlo (MCMC) sampling (see

Table 2 .
Summary of the mixed-effects analysis of variables predicting lecturers' English competence.The model also includes random intercepts for lecturer (SD estimated at 0.5136) and students' nationality (SD 0.2051).