CLUE Research Paper Hannah Koetsier

You might also like

Download as docx, pdf, or txt
Download as docx, pdf, or txt
You are on page 1of 31

Rejected because of your accent; how can accent discrimination in a hiring context be

reduced?Hannah Elizabeth Koetsier

Radboud Universiteit

Docent: Frank van Meurs

24 juni, 2023
Abstract

Could prejudice toward non-native accents in a hiring setting be reduced by instructing

people on the dangers of discrimination on the basis of accent? This study attempts to find

answers to this question through a between-subjects design where Dutch participants were

put in one of three conditions; implicit instructions, explicit instructions, control group.

Accents also either listened to two speakers reading content in English in a moderate Dutch

accent or a slight accent. A questionnaire then assessed the attitudinal evaluations, perceived

hirability, perceived teaching quality, personal concerns and perceived comprehensibility in

194 participants (age 18-30). Results showed that both in the implicit and explicit instruction

conditions, discrimination for a moderate accent was eliminated for the variables status and

likeability. For the remaining variables, the prejudice intervention did not have any effect.

The findings and implications are discussed.

Key words: accent discrimination, prejudice control intervention, non-native accents, accent

prejudice, prejudice intervention


Introduction

As globalization increases and people are increasingly expected to be able to speak English,

English Medium Instruction (EMI) becomes more and more a part of education. EMI is

defined by Macaro et. al as “The use of the English language to teach academic subjects

(other than English itself) in countries or jurisdictions where the first language of the majority

of the population is not English.” (Macaro et. al., 2018, p. 37). This change in education has

as a consequence that many teachers now have to teach in the English language. However,

since it is often not their native language, they speak English with an accent. With this change

in education come questions about the consequences this may have for education itself, for

example: how does this affect the education of students?

Research into the effects of accents on perceptions and evaluation goes back to 1965,

when Lambert et. al. studied the evaluational reactions of people listening to different

dialects, using the matched-guise technique. This study suggested that this technique reveals

stereotyped attitudinal reactions toward accents. Later, in 1970, this matched-guise technique

was used to show that accents were rated differently depending on which social class they

were associated with (Giles, 1970).

Looking at more recent research, it seems that a native accent is preferred more than

non-native accents (van Meurs et. al., 2012; Inbar-Lourie, & Donitsa-Schmidt, 2020). A meta

analysis by Fuertes et. al., (2012) covered 20 studies that compared standard and non-

standard accents, assessing them on characteristics like status, solidarity and dynamism. The

results showed that speakers’ accents determine for a large part how people are perceived. To
be more specific, the results showed that speakers of non-standard accents are seen as less

intelligent, less attractive, and less successful, which affects even first impressions.

Judging someone on the basis of their accent (or: accent discrimination) has many

consequences for people who speak in non-native accents; it affects hiring possibilities

(European Commission, 2008) and subsequently affects non-native speakers’ lives (Lippi-

Green, 1994). Specifically in education, accentedness plays a large role in whether people

judge a speaker to be competent and suitable for a certain position (Boyd, 2003).

Prejudice toward accented people on the basis of their accent will therefore influence

whether students will be motivated to listen and subsequently influence the education of these

students, while also influencing the experience of teaching for the lecturer. It is therefore

important that research is done on how this prejudice arises and can be reduced.

Accent discrimination takes place in all kinds of contexts; in the workplace (Hideg et

al., 2022), but also in education. A recent study shows that accent discrimination also

specifically takes place in EMI contexts; Inbar-Lourie & Donitsa-Schmidt (2020) showed

that native English speakers were preferred more than non-native speakers by students in

universities. However, as these researchers looked at group differences, the results showed

that students who had a lower (self-assessed) level of English preferred non-native English

speakers. This shows that preference for (non-)native English speakers depends on the

characteristics and expectations of the listener as well, meaning that the effect of speakers’

accents may be mediated by listeners’ characteristics. Hendriks et. al (2021) conducted a

study investigating the effect of slight and moderate Dutch accents or native British English

accents on the evaluation by native and non-native English listeners. They concluded that the

speakers with a moderate Dutch accent were evaluated more negatively by Dutch and

international listeners, but not by native English listeners, suggesting that non-native listeners
evaluate non-native lecturers according to native speaker norms, confirming the idea that

listeners’ characteristics matter as well.

It has also been theorized that this prejudice exists because an accent serves as a cue

as to the speakers’ social status, where people are from and to what group one belongs to

(Freynet & Clément, 2019). Generally, non-standard accents are seen as minority accents

which are associated with lower socioeconomic status, less power and less use in the media

(Giles & Billings, 2004). A study by Santana-Williamson and Kelch (2002) found that the

attitudes of students toward teachers with English accents are not influenced by a speaker's

(non-)native accent, but by whether the student perceives the accent to be native or not. This

suggests that the ‘nativeness’ of the speaker plays a role in how speakers, and thus teachers,

are perceived.

This prejudice can also be explained using a concept called processing fluency. When

things are difficult to process, we automatically perceive them as less pleasant (Reber et. al.,

2004). Lev-Ari and Keysar (2010) studied the effects of non-native accent on the credibility

of the speaker. They found that because of the fact that non-native speakers are more difficult

to understand, the perceived credibility of the speaker will be lower.

Prejudice control studies

Several possible solutions have been developed to combat accent discrimination in an EMI

context. In the previously mentioned study by Lev-Ari and Keysar (2010), the researchers

attempted to remove prejudice toward non-native speakers by making the listeners aware of

what was causing their preference for speakers, namely the difficulty to understand the

speakers. This was successful; when listeners are aware of the difficulty, they rate the speaker

as more truthful. However, it was only successful for speakers with a light accent and not for

speakers with a strong accent. This generalizability of this study to different accents is
limited, however, as the native tongue of the speakers was different in the mild and the heavy

accent conditions. Another study by Roessel et al. (2019) shows that raising awareness to

reduce accent prejudice reduced the downgrading on affect, competence and hireability

evaluations. This therefore shows that raising awareness on prejudices is effective in reducing

the effects of prejudices themselves.

Another approach to reducing prejudice is influencing processing fluency. Dragojevic

(2020) attempted to reverse the prejudice effect by subtitling speakers, making the ability to

process the message easier. This resulted in more positive evaluations of the speaker,

showing that the negative evaluation of non-native speakers may be reduced by easing

processing fluency.

However, other prejudice control interventions have been suggested as well. A study

by Hansen et al. (2014) attempted to reduce accent discrimination by influencing people’s

attitudes towards non-standard languages by putting them in a similar position, where they

have to speak in their non-native language for a while before completing the experiment.

However, this intervention might not be as applicable to real life, as it will be hard to have

every member of a hiring committee speak in a foreign language before a hiring interview,

for example.

Even though research has been done on reducing accent discrimination, more research

is needed on this subject, as it is not yet known how the interventions that have been

developed thus far translate to real-world settings like EMI employment interviews and

whether they work for all different accents and languages. While Roessel et al. (2019)

introduced a prejudice control intervention for speakers speaking English with a German

accent in a hiring context, we want to extend this research by testing the effects of prejudice

control in a hiring context for EMI lecturers, instead of a job interview situation.
This study therefore aims to investigate how prejudice control affects the judgement

of non-native accents in English in students. We want to fill the gap in the research regarding

accent discrimination in EMI employment interviews, especially for Dutch-speaking listeners

and speakers. To investigate this, explicit and implicit prejudice control measures were used

and its effect on the listeners’ evaluations was measured. The speakers, who had either a

moderate or a slight accent strength, were evaluated on different scales like status,

competence, likeability, dynamism, hireability and perceived comprehensibility.

With this study, we want to shed more light on accent discrimination and give insight

into how to lower prejudice toward accented lecturers, thus filling the existing gap in research

and potentially laying the groundwork for the development of prejudice reducing measures in

employment interviews for EMI lectures.

To investigate the gap in research and lay the groundwork for the development of

prejudice control interventions, we want to investigate the following research question; Does

informing people about implicit prejudice towards non-native accents help prevent prejudice

against non-native accents in a hiring context?

We hypothesize that;

 There will be a difference in evaluations between the different accent groups;

 The attitudinal evaluations will be more positive in the slight accent group than in the

moderate accent group;

 In the prejudice control conditions there will be less difference in evaluations between

the accents than in the control condition.


Methods

Materials

The participants were given instructions that they should act as a person in a selection

committee in a Dutch university that was hiring a new lecturer in marketing, and that the

potential candidates had sent in an audio fragment of a lecture. The audio fragments used

were audio fragments used in Hendriks et al. (2016). The context of the questionnaire was a

EMI context, where the central question was: would you, as a student, recommend the person

as a lecturer? The participants were randomly assigned to one of three prejudice control

conditions; implicit instructions, where participants were only pointed to the fact that one

should not discriminate in general, explicit instructions, where participants were pointed to

the fact that accent often leads to prejudice and that people shouldn't discriminate on the basis

of accents, and a control group, where participants received no instructions The prejudice

control intervention was developed in a pretest. Furthermore, the participants listened to

either a speaker with a moderate Dutch accent or a speaker with a slight Dutch accent. The

degree of accentedness was based on a pretest conducted in Hendriks et al. (2016).

Subjects

A total of 194 native speakers of Dutch took part in the experiment (age: M = 24.60, SD =

3.44; range = 12). The distribution of the gender of the participants can be considered normal

(skewness = .030) The skewness of gender was found to be .030, indicating that the

distribution was normal. The completed education levels of the participants ranged from

HBO to WO, of which the larger part completed HBO (N = 125) (see table 1.)

Table 1. Percentages of highest completed education level

N %
HBO 125 64,4%

WO 69 35,6%

Of the participants, 84.5% had previous work experience, and 18.6% have previously

worked as a member of a hiring panel. The participants were not very familiar with the Dutch

language (Min = 2, Max = 7, M = 3.90, SD = 1.152). The self-assessed proficiency was

quite high (Min = 1, Max = 7, M = 5.74, SD = 0.79) , while the actual proficiency was

average (Min = 45, Max = 100, M = 73.15, SD = 14.81). The mean prejudice toward accents

was 3.90 with a standard deviation of 1.1524. The average listener’s accent was considered

average (M = 3.95, SD = 1.25). Most of the participants were both taught in English (N =

186, M = 53.12, SD = 29.24) and Dutch (N = 182, M = 65.22, SD = 30.22). Some

participants (N = 51) were taught in another language as well, but not to a large extent (M =

24.31, SD = 27.51). (See table below).

Table 2. Descriptive Statistics of Subject Background Variables

Maximu

N Minimum m M SD

How old are you? 194 18 30 24.60 3.438

What is your current or 194 3 4 3.36 .480

highest completed level of

education?

Actual proficiency as 194 45.00 100.00 73.1508 14.81055

measured on LEXTale test

cm_prejudicemeasure 194 1.00 7.00 3.8976 1.15241

cm_selfassessedproficienc 194 2.75 7.00 5.7371 .79414


y

Valid N (listwise) 194

To test the distribution of background characteristics, ANOVA tests were conducted.

Age (F(5,188) = 1.08, p = .375), familiarity with the Dutch language (F (5, 188) = 0.51, p

= .771), actual proficiency as measured on a Lextale test (F (5, 188) = 0.74, p = .598), self-

assessed general proficiency (F (5, 188) = 0.656), p = .657), prejudice measure (F (5, 199) =

0.655, p = .658), the listener’s accent (F (5, 199) = 1.692, p = .138) were all distributed

equally across the six different conditions. Also the programme languages English (F (5, 185)

= 0.93, p = .463), Dutch (F (5, 181) = 1.190, p = .316), and other languages (F (5, 50) =

1.331, p = .268 were distributed equally across all six conditions.

For the remaining background variables a Pearson Chi-Square test was conducted.

Gender (χ2(1) = 4.390, p = .928) education level (χ2(1) = 4.632, p =.462), work experience

((χ2(1) = 2.421, p .788), hiring experience (χ2(1) = 8.035, p = .154) and general proficiency

(χ2(1) = 6.022, p = .813) were distributed equally across the six different conditions.

Research Design

To investigate our hypotheses, we used a 3 x 2 experimental between-subject design. The

participants were randomly assigned to one of three prejudice control conditions; implicit

instructions, explicit instructions, and a control group, where participants received no

instructions. Furthermore, the participants listened to either a speaker with a moderate Dutch

accent or a speaker with a slight Dutch accent. The experiment was a verbal-guise

experiment, meaning that the same text was read but by different speakers. Each participant

evaluated one fragment in the online questionnaire.


Instruments

Listeners filled in a questionnaire in which they rated the fragments on perceived

comprehensibility of speaker, impressions of the speaker, attitudinal evaluations, teaching

quality and hirability. The participants rated the speaker on different scales on things like

perceived comprehensibility and teaching quality. Before conducting the reliability analyses,

we recoded several items of the comprehensibility, likeability, origin speaker, speaker accent

and listener accent questions.

Background characteristics of participants

Attitude prejudice was measured on a 7-point scale from completely disagree to

completely agree, based on Ura et al., 2015 with statements like: ‘Speakers with non-native

English accents should learn to speak English better’. Familiarity with the Dutch accent was

measured on a 7-point Likert scale from completely disagree to completely agree based on

Hendriks et al., 2021 with items like: ‘I am familiar with Dutch-accented English’. The

reliability of this scale was good (𝛂 = .77). Subjects were also measured on their self-

assessed level of English using a 7-point Likert scale from poor- excellent, based on Krishna

& Alhuwalia, 2008. The reliability of this scale was good (𝛂 = .78). Actual English

proficiency was measured using a Lextale test (Lemhöfer & Broersma, 2012). Prejudice

measure was assessed using a 7-point Likert scale with statements like: ‘It is irritating when a

speaker has a non-native accent in English.’, and the reliability was good (𝛂 = .83).

Listener’s accent was measured using two statements; ‘I sound like a native speaker of

English’ and ‘I have a strong foreign accent in my English.’ for which the answers were on a

7-point Likert scale from completely disagree to completely agree. The reliability of this

scale was low; (𝛂 = .36). Personal concerns were rated on a 7-point scale with items like: ‘If

I attend classes taught by this lecturer, I expect a high grade on my exam’ on a scale from
completely disagree – completely agree. The scale was constructed for this study; inspired by

‘personal concerns’ by Roessel et al. 2020. The reliability of the scale was good; 𝛂 = .67

Manipulation check

For the first manipulation check, we measured the degree of accentedness on a 7-point Likert

scale from completely disagree to completely agree, based on Jesney, 2004. For the second

manipulation check we measured the variable ‘speaker accent’, which tested whether

participants thought the speaker sounded like a native speaker of English. This was measured

on a 7-point Likert scale from completely disagree to completely agree. The reliability of this

scale was not very high; 𝛂 = .44.

Attitudinal evaluations

Attitudinal evaluations were measured using 7-point scales from completely disagree –

completely agree; based on Bayard et al., 2001; Grondelaers et al., 2019; Hendriks et al.,

2014, 2016; Nejjari et al., 2012. Statements started for example with: ‘In my opinion, the

lecturer sounds…’.

Perceived comprehensibility was measured on a 7-point Likert scale from completely

disagree - completely agree, based on Hendriks et al. 2016. The reliability of ‘perceived

comprehensibility’ comprising six statements (for example: ‘I have to listen very carefully to

be able to understand the lecturer’) anchored in a 7-point Likert scale was acceptable: 𝛂

= .85.

The reliability of ‘status’ comprising five items (authoritative, trustworthy, self-

confident, influential, a powerful voice) was acceptable: 𝛂 = .80.

The reliability of ‘competence’ comprising five items (reliable, intelligent competent,

hardworking, educated) was good: 𝛂 = .80.


The reliability of ‘likeability’ comprising eight items (credible, sympathetic, warm,

humorous, tactful, polite, irritating, unfriendly) was acceptable: 𝛂 = 82.

The reliability of ‘dynamism’ comprising four items (energetic, talkative, cheerful,

active) was good : 𝛂 = .87.

Teaching quality was measured with six items on a 7-point scale from completely

disagree – completely agree; based on Hellekjær, 2010. Perceived teaching quality was

measured using six statements anchored by a seven-point Likert scale from ‘completely

disagree’ to completely agree’ with statements like: ‘The lecturer's subject knowledge is

excellent’, ‘The lecturer can clearly communicate the content of the lecture’ and ‘The lecturer

is a good teacher’. The reliability was good : 𝛂 = .85.

Hirability was measured using a 7-point Likert scale from completely disagree –

completely agree and was adapted from/based on Deprez-Sims & Morris (2010). The

reliability of ‘hirability’ measured with five items analyzed on a 7-point Likert scale (with

statements like: ‘I would recommend employing this lecturer’) was good: 𝛂 = .89.

For all variables for which Cronbach’s alpha was .70 or higher, composite means

were calculated. Furthermore, for ‘likeability’, for which the alpha was between .60 and .70,

a composite mean was calculated. For the two variables where cronbach's alpha was not high

enough (listener accent and speaker accent), the items were analyzed separately.

Procedure

Participants filled in an online questionnaire on Qualtrics.com. The participants were

informed that participation is voluntary and the answers would be processed anonymously.

After agreeing to partake in the study, participants started with answering the background

questions. The participants were then told to imagine they were a member of a hiring

committee responsible and they were hiring a new marketing lecturer. First they were
presented with instructions, or no instructions, depending on which prejudice control group

they were assigned to. They then listened to an audio clip of a lecturer with either a slight or a

moderate accent. Afterward they filled in the questionnaire assessing all the variables. They

first gave a first impression of the speaker, and then filled in 7-point likert scale questions

assessing their attitudinal evaluations of the speaker. Finally, the participants were also asked

about their own proficiency and asked to complete the Lextale proficiency test. Finally, the

participants were asked more background questions and assessed on other characteristics,;

prejudice, familiarity with the Dutch language. gender, work experience, hiring experience

and in which language their degree programme was taught. On average, the experiment took

about 17 minutes. Participants were reimbursed for filling out this questionnaire.

Statistical treatment

To test the hypotheses and answer the research question, we used a two-way ANOVA

for the main dependent variables to test whether there is an interaction between accent and

prejudice control, and whether there are main effects of accent and prejudice control. If there

was an interaction, a follow-up one-way ANOVA was done to determine whether there was a

significant effect of accent in the different prejudice control groups, or if there was a

significant effect of the prejudice control groups in the different accent groups.
Results

The purpose of this study was to investigate whether implicit or explicit instructions about

accent discrimination can reduce negative evaluation regarding slight and moderate accented

speakers.

Manipulation checks

As can be seen in Table 1, the mean of ‘speaker accent 1’, which is the variable

testing whether participants thought the speaker sounded like a native speaker of English, was

higher for the moderate accent group (M = 4,77; SD = 1.98) than the slight accent group (M

= 2.86; SD = 1.58) (see table 3). The scores were measured on a scale from completely

disagree (1) to completely agree (7). This difference was significant (F (193, 94) = 11.23, p <

.001). This means that the manipulation check was successful; the participants thought the

speaker with the moderate accent sounded less native than the speaker with the slight accent.

Table 3. Evaluation of the nativeness of the English speaker (from completely disagree

(1) to completely agree (7)

Accent N M SD

Moderate 94 4.77 1.98

Slight 100 2.86 1.58

The mean ‘speaker accent 2’, which measured whether participants though the

speaker had a strong foreign accent in English, was higher for the moderate speaker group (M

= 4.79; SD = 1.81) compared to the slight accent group (M = 3.11; SD = 1.93), but not

significantly (F(192) = 0.344, p = .558) (see table 4).


Table 4. Evaluation of the accentedness of the English speaker (from completely disagree

(1) to completely agree (7)

Accent N M SD

Moderate 94 4.79 1.80

Slight 100 3.11 1.93

In the questionnaire, listeners were also asked to indicate the origin of the speakers.

To examine if speakers in the two accentedness conditions were actually recognized as Dutch

(for the moderate and slight conditions), a Chi-square analysis was carried out to examine the

relation between accentedness and correct identification of the speakers’ origin. The relation

was significant (Χ² = 63.25, p < .001); the origin of the moderately accented speaker was

identified correctly by relatively more participants (83%) than the origin of the slightly

accented speaker (26%).

Speaker evaluations

Status

A two-way ANOVA for status with accent and prejudice control as factors showed a

significant main effect of accent (F(1, 188) = 15, p <.001), no significant main effect of

prejudice control (F(2, 188) = 2.57, p = .079), and a significant interaction (F(2, 188) = 3.72,

p = .026) etc.).

After splitting the file on accent, a one-way ANOVA with prejudice control as a

factor showed a significant effect of prejudice control on the status of moderately accented

lecturers (F (2, 91) = 4.16, p = .019). Participants in the implicit control condition gave

higher status scores to the speakers (M = 5.04, SD = 0.71) than those in the control condition
(M = 4.36, SD = 1.22; p = 0.25). A one-way ANOVA showed no significant effect of

prejudice control on the status of slightly accented lecturers (F (2, 97) = 1.87, p = .16).

Table 5. Descriptive statistics of the variable ‘status’ of the moderately accented group after

splitting the file on accent

Prejudice control N M SD

Implicit 34 5,04 0,71

Explicit 34 4,46 1,06

Control 32 4,36 1,22

After splitting the file on prejudice control, a one-way ANOVA with accent as a

factor showed a significant effect of accent on the status of the speaker for participants who

had received no prejudice control instructions (F (1, 64) = 15.21, p <.001). The status scores

were significantly higher for the speaker with a slight accent (M = 5.40, SD = 0.89) than the

speaker with a moderate accent (M = 4.36, SD = 1.22). There was no significant effect of

accent on the status of the speaker for the participants who had received implicit prejudice

control instructions (F (1, 64) < 1) and for the participants who had received the explicit

prejudice control instructions (F (1, 60) = 3.76, p = .057).

Table 6. Descriptive statistics of the variable ‘status' for the control group after splitting the

file on prejudice control

Accent N M SD

Moderate 34 4.36 1.22

Slight 32 5.49 0.89


Likeability

A two-way ANOVA for likeability with accent and prejudice control as factors showed no

significant main effect of accent (F (1, 188) = 0.85, p = .357) and no significant main effect

of prejudice control (F (2, 188) = 1.11, p = .333), but a significant interaction effect (F (2,

188) = 3.08, p = .048).

After splitting the file on accent, a one-way ANOVA showed a significant effect of

prejudice control on the likeability of moderately accented lecturers (F (2, 91) = 3.23, p

= .044). Participants in the implicit instruction conditions gave higher scores (M = 5.37; SD

= 0.70) than those who were not instructed (M = 4.87; SD = 0.81).

Table 7. Descriptive statistics of the variable 'likeability' for the moderate accent group

Prejudice control N M SD

Implicit 32 5.37 0.70

Explicit 28 5.09 0.93

Control 34 4.87 0.77

There was no significant effect of prejudice control on the status of slightly accented

lecturers (F (2, 91) = 1.02, p = .363)

After splitting the file on prejudice control, a one-way ANOVA showed a significant

effect of accent on the likeability scores of the speakers in the control group that received no

instructions (F (1, 64) = 6.30, p = .015) The moderate accented speaker were scored

significantly lower on perceived likeability (M = 487, SD = 0.77) than the speaker with a

slight accent (M = 5.37, SD = 0.85).


Table 8. Descriptive statistics of the variable 'likeability' of the control group

Accent N M SD

Moderate 34 4.87 0.77

Slight 32 5.37 0.85

There was no significant effect of accent on the likeability of the speaker for the

people in the implicit instructions group (F (1, 64) = 0.95, p = .33) and the explicit prejudice

control group (F(1, 60) = 0.00, p = .98).

Comprehensibility analysis

A two-way ANOVA for comprehensibility with accent and prejudice control as factors

showed no significant effect of accent (F (1, 188)= 0.21, p = .650), no effect of ojudice

control (F (2, 188) = 0.88, p = .416), and no interaction effect (F (2, 188) = 0.98, p = .376).

Competence

A two-way ANOVA for competence with accent and prejudice control as factors showed no

significant effect of prejudice control (F (2, 188) = 1.18, p = .310), no significant effect of

accent (F (1, 188)= 3.8, p = .053), and no interaction effect (F (2, 188) = 2.59, p = .078).

Dynamism

A two-way ANOVA for dynamism with accent and prejudice control as factors showed no

significant main effect of accent (F (1, 188) = 0.48, p = .49), no significant main effect of

prejudice control (F (2, 188) = 0.03, p = .973), and no significant interaction effect (F (2,

188) = 2.90, p = 0.06).


Teaching quality

A two-way ANOVA for teaching quality with accent and prejudice control as factors showed

a significant main effect of accent (F(1, 188) = 7.76, p = .006); participants in the slight

accent condition scored higher on perceived teaching quality (M = 5.29, SD = 0.96) than

participants that listened to a moderately accented teacher (M = 4.92, SD = 0.88). However,

there was no significant main effect of prejudice control (F(2, 188) = 0.35, p = .707), and no

significant interaction effect (F(2, 188) = 2.30, p = .052).

Hireability

A two-way ANOVA for hireability with accent and prejudice control as factors showed a

significant main effect of accent (F(1, 188) = 11.06, p = .001); the participants that listened

to a slight accent gave higher scores for the speaker’s hireability (M = 5.35, SD = 1.17) than

the participants how listened to a moderate accent (M = 4.80, SD = 1.09). There was no

significant main effect of prejudice control (F(2, 188) = 1.30, p = .276), and no significant

interaction effect (F(2, 188) = 1.69, p = .188).


Discussion

The aim of this study was to determine how native Dutch listeners evaluate native Dutch

English-speaking lecturers with a moderate accent or a slight accent in an EMI context on

status, competence, likeability, dynamism, teaching quality and hirability. The results showed

that in general, in the control group, people evaluated moderate accented speakers as lower in

competence and teaching quality. However, for the variables status and likeability, there was

no difference in prejudice between the different accents for the group that received implicit

and explicit instructions, meaning the prejudice control did work for those conditions and

variables.

Manipulation checks

In this study, we conducted manipulation checks. We examined whether participants really

perceived the accents of the moderate speakers as non-native and the slightly accented

speakers as native. The results showed that participants in general thought the moderately

accented speaker was non-native more often than the slightly accented speaker, and the other

way around; in general they thought more often that slightly accented speakers were native

than not.

For the second manipulation check we researched how participants identified the

origin of the accent, namely whether it was a Dutch accent or not. The origin of the moderate

accent was identified correctly by more participants than the origin of the slight accent, which

means that the manipulation was successful.

The results of the manipulation checks are therefore in line with previous research, as

Lev-Ari and Keysar (2010) also showed that slight accents are more often perceived to be

native than moderate or strong accents.


Comprehensibility

The results showed that participants found the speaker with the slight accent easier to

understand than the speaker with the moderate accent. This finding is in line with previous

research; in the study by Nejjari et al., (2012) the comprehensibility of the standard accent

was evaluated higher than the slightly and moderately accented English, which was the case

in our study as well. The study by Roessel et al. (2019) also shows that slight accents are

regarded as more understandable than moderate accents. This result is therefore in line with

previous research. This result is also is interesting, however, as it may be possible that the

comprehensibility of the speaker might also influence the evaluations of the listeners;

according to Reber et al. (2004) and Lev-Ari and Keysar (2010), when speakers are more

difficult to understand, attitudinal evalutions of the speaker will also be more negative.

Attitudinal evaluations

Status

Regarding the variable status, there was an interaction effect of accent strength and prejudice

control on the ratings. The results showed that prejudice control did have an impact on the

perceived status of the moderately accented lectures; participants in the implicit instructions

condition gave higher scores than those in the control condition. However, this effect did not

occur for the group that listened to the slightly accented speaker. So, the prejudice control

intervention did have effect when the accent was moderate, but not when the accent was

slight. Furthermore, in the group who received no instructions, accent did have a significant

effect on the status evaluations, but this was not the case for the groups who received implicit

and explicit instructions.


These results are in line with previous research. For example, Nejjari et al. (2012)

investigated the influence three different accents (a slight Dutch accent, moderate Dutch

accent and Standard British accent) had on, among other things, the status of the speaker.

They found that the standard British accent evoked more status than the accented speakers,

and slight accent had higher status reports than the moderate accent. The meta-analysis by

Fuertes et al., also revealed that the effect of using a non-standard accent compared to a

standard accent on the status evaluations is very high. Thus, our research confirms previous

studies in regards to the effect of accents on status evaluations. However, our study also

shows that a prejudice control intervention eliminates this effect when participants listen to a

moderate accent. This is in harmony with previous research by Roessel et al., which used a

similar prejudice intervention.

Likeability

Similarly to status, the prejudice control intervention did have an effect on the evaluation of

the participants on the speaker’s likeability; for the group who listened to a moderate accent,

participants in the implicit instruction condition gave higher scores than participants who did

not receive instructions. This effect did not occur in the group of people who listened to

slightly accented lecturers. Moreover, there was a significant effect of accent on the

likeability scores in the group that did not receive instructions. Previous research supports

this; as said before, the prejudice control intervention through raising awareness was effective

in the study by Roessel et al. (2017), as well. Moreover, Nejjari et al. (2012), found that

degrees of speaker accentedness matter in the participants' evaluation of a speaker;

moderately accented speakers command less affect, a variable comparable to likeability, than

the slightly accented speaker. These results show that a moderate accent tends to make
speakers less likeable, but also that this effect can be minimized by raising awareness about

the effects of accent discrimination.

For the remaining variables (competence, dynamism, teaching quality, hirability, and

personal concerns) neither implicit nor explicit instruction reduced accent discrimination.

This is not in line with previously mentioned Roessel et al. (2017), who found an effect of

their prejudice control intervention on the attitudinal evalutions of the speaker and the

hirability of the speaker. A possible explanation for this might be that in the study by Roessel

et al., the participants were German and the accent was German as well. It is possible that the

German accent has different effects on evaluations and might therefore be eliminated

differently than the Dutch accent. Another explanation for the fact that the

prejudice control intervention did not have an effect for these varaibles is that the intervention

(implicit or explicit instructions) was done in the beginning of the experiment. The

experiment took on average 17 minutes and it is possible that the information read before

evaluating might have been forgotten already.

Thirdly, it might simply be the case that accent discrimination happens unconsciously,

as most biases are unconscious, and is difficult to overcome.

Accent did have an effect on most of the attitudinal evaluation variables, showing accent

discrimination, namely for the variables teaching quality and reliability, as is expected from

previous literature showing an effect of accent on attiduinal evalutions (Fuertes et. al., 2012;

van Meurs et. al., 2012; Roessel et al., 2017; Inbar-Lourie, & Donitsa-Schmidt, 2020).

However, interestingly enough, for some of these variables, no accent discrimination

took place; for dynamism, competence and personal concerns (learning ability), the slightly

accented speaker was not rated significantly different from the moderately accented speaker.
For these variables, it seems that participants do not take accent to be a deciding factor in

how they evaluate persons. A possible explanation for this might be that other factors are

weighed more heavily. For example, how much expression is used in speech, the contents of

the speech and used words might be other factors influencing the evaluations.

This result is not surprising, however, when looking at previous research done by

Hendriks et al. (2016), which showed that slightly accented speakers were perceived to have

less competence than moderately accented speaker. However, one must take into account that

this study was done with German listeners and Dutch speakers, which may have been

influenced by a positive attitude from Germans toward Dutch people.

However, the result does not concur with other research by Roessel et al. (2017) and

Hendriks et al. (2017), which showed that speakers with a strong accent were evaluated more

negatively on their competence.

Limitations and directions for future research

One limitation of our study can be found in the use of the verbal-guise method, meaning that

different speakers were used, but the text was the same. It is possible that other differences

between the speakers, like voice quality, delivery style, voice sound, could have influenced

the judgemnt of the listeners. For future studies, it is recommended that a speaker is found

that can employ different accents, so that other confounding factors can be eliminated.

A second limitation is that the instructions and the recordings were shown in the

beginning of the experiment only. The instructions might have been forgotten by the time

participants were halfway through the experiment, and the way the accent sounded might

have been forgotten as well. This may have lead participants to have given unreliable

answers. We recommend that in future research, the researchers keep the short memory of

participants in midn adn play the audio multiple times, or keep the questionnaire rather short.
Furthermore, as this study was rather short-term and online, we recommend future

research to also be focused on possible long-term interventions, and in a real life setting. This

way, the research may be more generalizable to other, real-life situations, and the long-term

effects may be studied. It is possible that this intervention only works for a short time, as

most biases are embedded unconsciously and may resurface when the instructions are

forgotten.

Conclusion

This study examined the effect of a prejudice control intervention on the attitudinal

evaluations of participants listening to different accents in a hiring setting. The results

showed that instructing people on the effect of discrimination and to not discriminate

eliminates accent discrimination when participants were evaluating the speakers on status and

likeabiblity. However, the prejudice intervention did not have succes with the variables

competence, personal concern, teaching quality, hireability and dynamism.

Even though previous studies have been conducted to investigate the effect of

different accents on evaluations, few studies have started to investigate the possible

interventions that may combat accent discrimination. This study thus fills an existing gap in

the research where an instruction intervention is tested in an EMI-hiring context, specifically

with Dutch accents. Not only does this study provide more insight into accent discrimination

in an EMI context and a hiring context, it also lays the groundwork for the development of

interventions that will combat accent discrimination. This study provides the necessary

knowledge that is needed for the development of these interventions.


References

Boyd, S. (2003). Foreign-born teachers in the multilingual classroom in Sweden: The role of

attitudes to foreign accents. International Journal of Bilingual Education and

Bilingualism, 6(3-4), 283-295. DOI: 10.1080/13670050308667786

Dragojevic, M. (2020). Extending the fluency principle: Factors that increase listeners’

processing fluency positively bias their language attitudes. Communication

Monographs, 87(2), 158-178.

European Commission. (2008). Discrimination in the European Union: Perceptions,

experiences and attitudes (Special Eurobarometer 296, Wave 69.1). Retrieved from

http://ec.europa.eu/COMMFrontOffice/PublicOpinion/index.cfm/Survey/getSurveyDetail/

yearFrom/1973/yearTo/1994/surveyKy/656

Fuertes, J. N., Gottdiener, W. H., Martin, H., Gilbert, T. C., & Giles, H. (2012). A meta‐

analysis of the effects of speakers' accents on interpersonal evaluations. European

Journal of Social Psychology, 42(1), 120-133.

Freynet, N., & Clément, R. (2019). Perceived accent discrimination: Psychosocial

consequences and perceived legitimacy. Journal of Language and Social Psychology,

38(4), 496-513.

Giles, H. (1970). Evaluative reactions to accents. Educational review, 22(3), 211-227.

Hansen, K., Rakić, T., & Steffens, M. C. (2014). When actions speak louder than words:

Preventing discrimination of nonstandard speakers. Journal of Language and Social

Psychology, 33(1), 68-77.

Hendriks, B., Van Meurs, F., & Hogervorst, N. (2016). Effects of degree of accentedness in

lecturers’ Dutch-English pronunciation on Dutch students’ attitudes and perceptions of

comprehensibility. Dutch Journal of Applied Linguistics, 5(1), 1-17.


Hendriks, B., van Meurs, F., & de Groot, E. (2017). The Effects of Degrees of Dutch

accentedness in ELF and in French, German and Spanish. International Journal of

Applied Linguistics, 27(1), 44-66.

Hendriks, B., van Meurs, F., & Usmany, N. (2021). The effects of lecturers’ non-native

accent strength in English on intelligibility and attitudinal evaluations by native and

non-native English students. Language Teaching Research, 1362168820983145.

Hideg, I., Shen, W., & Hancock, S. (2022). What is that I hear? An interdisciplinary review

and research agenda for non‐native accents in the workplace. Journal of

Organizational Behavior, 43(2), 214-235.

Inbar-Lourie, O., & Donitsa-Schmidt, S. (2020). EMI Lecturers in international universities:

is a native/non-native English-speaking background relevant?. International Journal of

Bilingual Education and Bilingualism, 23(3), 301-313.

Jesney, K. (2004). The use of global foreign accent rating in studies of L2 acquisition.

Calgary, AB: University of Calgary Language Research Centre Reports, 1-44.

Lambert, W., Anisfeld, M. and Yeni-Kosmshian, G. (1965). Evaluational reactions of Jewish

and Arab adolescents to dialect and language variations. Journal of Personality and

Social Psychology, 2, 84-90.

Lev-Ari, S., & Keysar, B. (2010). Why don't we believe non-native speakers? The influence

of accent on credibility. Journal of Experimental Social Psychology, 46(6), 1093-1096.

Lippi-Green, R. (1994). Accent, standard language ideology, and discriminatory pretext in

the courts. Language in Society, 23(2), 163-198.

Macaro, E., Curle, S., Pun, J., An, J., & Dearden, J. (2018). A systematic review of English

medium instruction in higher education. Language Teaching, 51(1), 36-76.


Meurs, F. van, Hendriks, B., & Planken, B. (2012). Studying the effects of non-nativeness in

a business communication context. Experimental studies as input for an advanced level

bachelor course, 37-45. Retrieved from https://hdl.handle.net/1887/21883

Nejjari, W., Gerritsen, M., Van der Haagen, M., & Korzilius, H. (2012). Responses to Dutch‐

accented English. World English, 31(2), 248-267.

Reber, R., Schwarz, N., & Winkielman, P. (2004). Processing fluency and aesthetic pleasure:

Is beauty in the perceiver's processing experience?. Personality and Social Psychology

Review, 8(4), 364-382.

Roessel, J., Schoel, C., Zimmermann, R., & Stahlberg, D. (2019). Shedding new light on the

evaluation of accented speakers: Basic mechanisms behind nonnative listeners’

evaluations of nonnative accented job candidates. Journal of Language and Social

Psychology, 38(1), 3-32.

Santana-Williamson, E., & Kelch, K. (2002). ESL students’ attitudes toward native and non-

native speaking instructors’ accents. CATESOL Journal, 14(1), 57-72.

You might also like