As elsewhere in western Europe, Flanders has known increased migration and international mobility in the last few decades. Today, about 10 % of all Flemish inhabitants do not have the Belgian nationality, and an additional 15.2 % are Belgians with a migrant background (Statbel, 2022). In this context of growing ethnic and linguistic diversity, ethnic minority groups often face discriminatory behaviour on the part of the majority group, which is evidenced by ample research by scholars in sociology, labour economics and personnel psychology. Heylen and Van den Broeck (2016), for example, investigated the influence of ethnicity on a person’s chances on the Flemish housing market and found that candidates with a Moroccan or Turkish name, especially males, less often received an invitation for a house tour than candidates with a Flemish name after expressing their interest in the house via mail or telephone. As for employment, Baert et al. (2013) demonstrated by means of so-called correspondence experiments that resumes signed with a Turkish name received significantly fewer invitations for a job interview than resumes of candidates with a Flemish name, even though both resumes were otherwise identical. The potential role of language in such processes of discrimination, however, remains understudied. To the best of our knowledge, Van der Bracht et al. (2014) is the only study to have investigated language-based discrimination in Flanders. In their telephone survey on the rental housing market in the cities of Antwerp and Ghent, they found that candidates with an Arabic name who were proficient in Dutch and had no noticeable accent were discriminated against as often as Arabic candidates with lower proficiency in Dutch.

The question whether an accent, just like a name, can influence perceptions of ethnic minority members is by no means trivial, since foreign accents are a common aspect of second language acquisition (Flege et al., 1995; Weber et al., 2011). Especially for late L2 learners, mastering (near-)native pronunciation is a persistent difficulty (Bongaerts et al., 1997; Hyltenstam, & Abrahamsson, 2003; Dollmann et al., 2009). Although foreign accents can hamper intelligibility (Dragojevic et al., 2017), they do not necessarily hinder successful communication with L1 speakers (cf. Hansen Edwards, 2008; Ramjattan, 2019). While accent strength has generally been found to correlate with negative attitudes towards non-native speech (Dragojevic et al., 2017; Gluszek, & Dovidio, 2010; Hendriks et al., 2021; Ryan et al., 1977; Simon et al., 2022), even mild non-native accents that do not hamper intelligibility have been found to provoke negative attitudes in native listeners. This downgrading is most consistent for status-related features: in most experimental attitudinal studies, non-native speakers are considered less credible (Lev-Ari, & Keysar, 2010) and less superior (Grondelaers, & Van Gent, 2019; Lindemann, 2003) than native speakers. In addition, the prestige or stigma that a particular foreign language or a group of non-native speakers may be associated with can also affect native speakers’ judgements of these accents and speakers. Especially the foreign accents of speakers of languages which are labelled as “non-prestigious” generally elicit negative evaluations (Lindemann, 2005; Torstensson, 2010). In Western contexts, the latter type of accents mostly concerns non-Western accents such as Arabic or Turkish, which are associated even less with status- or solidarity-related traits (such as intelligence, power or friendliness) than Western accents such as French or American English, instead provoking perceptions of speakers as “aggressive”, “anti-social”, “low-skilled” and “unintelligible” (Grondelaers et al., 2015; Torstensson, 2010). These negative attitudes towards foreign-accented speech can create a disadvantage for L2 speakers in many aspects of their social life: their accents can prevent qualification for high-status, well-paying job positions (Creese, & Kambere, 2003; Hosoda, & Stone-Romero, 2010), and impact judgement in court hearings through stereotypical attributions on credibility and perceived guilt (e.g. Bayard et al., 2001; Rodman, 2002).

Even though Flanders has an extensive tradition of sociolinguistic (and principally experimental) research on attitudes towards endogenous language variation (i.e. the use of Standard vs. non-standard Dutch), attitudes towards exogenous language use belong to relatively uncharted waters. An example of a study conducted in the Flemish context that does focus on the latter is Simon et al. (2022), who measured attitudes towards two Chinese speakers with high proficiency in Dutch, and concluded that the speaker who was rated as having a stronger (foreign) accent was also evaluated more negatively on status and comprehensibility. Rosseel (2021) investigated attitudes of university students and staff towards the requirement of a proficiency at CEFR level B2 for Flemish university courses in Dutch, and in particular how successful they estimated participation in higher education would be for L2 speakers of Dutch who represented CEFR pronunciation levels B1, B2 and C1. Overall, Rosseel found positive evaluations of all foreign-accented speakers’ chances of success in higher education, and a positive, albeit non-significant, trend in the evaluations whereby an improved level of pronunciation led to increased appreciation of the speakers’ language use.

Although exogenous accents mostly seem to provoke negative attitudes, and Flanders seems to be no exception to that trend, some studies in the Flemish context have found associations between certain exogenous accents used by ethnic minority members, many of whom are not foreign but Flemish born, and a more “informal” or “modern” kind of prestige. Grondelaers and Marzo (2022) investigated evaluations of Turkish and Italian Limburgian colloquial Dutch and Turkish and Italian-accented Citétaal, an urban vernacular that is primarily used by the ethnically diverse Limburgian youth in the former mining area. These speech styles were compared to Limburgian-accented Standard Dutch in a speaker evaluation experiment. Grondelaers and Marzo found that whereas Italian and Turkish-accented colloquial speech scored lower than standard speech but still positively on superiority (features such as “good grades” and “well-paid job”), Citétaal received negative scores on this dimension. On integrity (i.e. the features “sincere”, “warm”, “friendly”), Italian Limburgian colloquial speech scored higher than all other varieties. Both Italian and Turkish-accented Citétaal scored highest on dynamism, or “streetwise prestige” (i.e. the features “cool”, “loud”, “provocative” and “tough”). While the Italian and Turkish-accented colloquial speakers hovered around neutrality on this dimension, standard speech scored negatively on dynamism. For Netherlandic Dutch, Grondelaers et al. (2015) found that Moroccan speakers from Nijmegen and Amsterdam were downgraded on status and beauty compared to native speakers from the same areas. In Grondelaers and Van Gent’s (2019) speaker evaluation experiment, which combined a Dutch or Moroccan name with Moroccan, Limburgian and Randstad accents of varying strengths, both the Moroccan names with a Moroccan accent and those without an audible exogenous accent were categorically downgraded on superiority, leading to the conclusion that unaccentedness does not compensate for the stigma of a Moroccan name (Grondelaers, & Van Gent, 2019). Simultaneously, however, Moroccan-flavoured Dutch was found to reflect dynamism, in that it was associated with the traits “hip”, “tough” and “cool”. In a follow-up-experiment that included female speakers, Grondelaers, & Van Gent (2022) found that dynamism associations only pertain to male Moroccan-Dutch speakers. Broad Moroccan accents were downgraded on superiority for both male and female speakers, but female speakers with a Moroccan name and no audible Moroccan accent were judged no less superior than the indigenous speakers without an audible regional accent. Stereotypes can thus clearly play out differently for male vs. female ethnic minority members.

Whereas evaluative differences between endogenous vs. exogenous accents have thus far been little studied in Flemish experimental attitudinal research, the use of standard vs. nonstandard varieties, marked by morphosyntactic, lexical and phonological differences, has in comparison received much more scholarly attention. The description of Grondelaers and Marzo’s (2022) findings above already hinted at the distinction that Flemish linguists make between Standard and Colloquial Dutch (Geeraerts, & Van de Velde, 2013). The latter variety is in Flanders most often termed tussentaal (‘in-between language’), as it is, in terms of its formal features, situated in between Standard Dutch and the dialects. Colloquial Dutch has over the past few decades gained ground in both the private and the public spheres (Absillis et al., 2012; Geeraerts, & Van de Velde, 2013; Plevoets, 2008, 2013; Vandekerckhove, 2005) and has been shown to elicit different evaluations compared to Standard Dutch. Those studies demonstrate that while Standard Dutch is still the overtly prestigious variety in Flanders, Colloquial Dutch is typically associated with social attractiveness (e.g. Lybaert, 2017; Impe, & Speelman, 2007) and dynamism (Grondelaers, & Speelman, 2013; Grondelaers et al., 2020; Rosseel et al., 2019). It remains an open question whether the same associations for Standard and Colloquial Dutch can be found when a speaker has an audible exogenous accent, since, as Rosseel (2021, p. 225) notes, most attitudinal studies investigating differential perceptions of native and non-native speakers have up to now focused on use of the standard language.

Indications that the use of standard vs. colloquial language by ethnic minority speakers can lead to different perceptions are provided by Lybaert et al. (2022), who investigated the influence of ethnicity (a Flemish vs. Moroccan name), a marker of religious affiliation (wearing a hijab vs. not wearing one) and language variety (Standard vs. Colloquial Dutch) on students’ evaluations of a female university teacher, but did not include accent as a marker of ethnicity (although the speaker used a mild Antwerp accent to enhance the naturalness of the speech samples). The authors found that university students evaluated the hijab-wearing teacher with a Moroccan background who spoke the standard variety more positively than her Colloquial Dutch speaking counterpart and all other guises (without hijab and with a Flemish or a Moroccan name). Lybaert et al. hypothesised that the veiled teacher’s proficiency in Standard Dutch “surprised” the students in that they possibly held lower expectations with regard to her language use, which may have led them to better appreciate her efforts to use the standard (Lybaert et al., 2022). These results suggest that the use of Standard vs. Colloquial Dutch does influence attitudes towards ethnic minority members, at least in the context of higher education and thus for highly educated speakers (university lecturers). One could hypothesize that in less formal, non-professional contexts, the use of a colloquial variety might lead to positive perceptions of non-native speakers. Especially in situations where native speakers of Dutch typically use colloquial features, knowledge of these features might be perceived as an index of “social integration” on the part of foreign-accented speakers, in that it could be interpreted as a sign of their effort and willingness to speak like the locals. Colloquial Dutch might thus incite positive solidarity-related evaluations of those speakers.

The present study aims to contribute to the budding research on exogenous variation in Flanders by conducting a speaker evaluation experiment examining the combined influence of ethnicity, accent and variety on the attitudes of ethnic majority Flemings towards non-native Dutch speech. The specific type of non-native Dutch speech we will investigate is that of Turkish L2 speakers of Dutch. Together with Moroccans, Turks constituted the largest part of the foreign guest workers who migrated to Belgium in the post-WWII decades, as a result of which today these two groups still constitute the largest non-EU ethnic minority communities in Belgium (Gsir et al., 2015; Kaya, & Kentel, 2008). Turkish is thus a common exogenous accent in Flanders. Although empirical evidence on the evaluation of this accent is as yet limited, Turkish is generally considered to represent what Jaspers (2009) has called “plebeian multilingualism”: a type of multilingualism that is associated with the multi-ethnic and mostly low-educated urban working classes, and that is assumed to be non-prestigious and stigmatised (Blommaert, 2011). In view of the lack of empirical evidence on the evaluation of Dutch spoken by L2 speakers of Turkish descent, our experiment was designed to study the combined effect of a speaker’s name (Flemish vs. Turkish), accent (native vs. Turkish) and variety (standard vs. colloquial) on the attitudes towards that speaker.

The Factor Analysis revealed that three factors provided a well-fitting model for the ratings on eleven traits. Self-assured scored high on two factors instead of one, hence we excluded this trait from further analyses. Based on a visual examination of the scree plot (Figure 1), the Factor Model was reduced to the first three factors. After applying the Varimax rotation these three factors explain 64.5 % of the total variance.

Figure 2 shows how the personality traits are related to the three factors, with the digits representing the so-called loadings of the traits on the factors. It is clear that the first factor PA1 represents the four Social Attractiveness traits Likeable, Helpful, Warm and Reliable together with the trait Easy-going, which we originally classified as a Dynamism trait. The second factor PA2 groups the four Superiority traits Highly-educated, Well-paid, Intelligent and Leader. Finally, the third factor PA3 represents the Dynamism traits Tough and Cool. Our factor model thus largely recovers the three dimensions we had envisioned. Easy-going, which we selected as a Dynamism trait, appears to be grouped together with Social Attractiveness traits. Of the four original traits we selected for Dynamism, only Cool and Tough make up the third factor, reducing it to a specific “streetwise” or “macho” type of Dynamism (cf. Grondelaers, & Van Gent, 2019; see discussion).

When analysing the respondents’ scores on PA1, Social Attractiveness, the combinations of Variety, Accent and Name (i.e. the three-way interactions between them) proved to have statistically significant differences and there was also a significant main effect for Scenario (Marginal R² = 11.617 %, Conditional R² = 43.894 %). None of the other controlling variables were statistically significant; see Table 4 for a table of the most important effects (and Table A.1 in the Appendix for all the estimated effects). The estimated means for Variety, Accent and Name are visualised in Figure 3. Confidence intervals that do not overlap indicate statistically significant differences. We see that Standard Dutch spoken with a native accent is evaluated as less socially attractive if the speaker has a Turkish name than if the speaker has a Flemish name, whereas the opposite is true for Colloquial Dutch (i.e. comparing the first panel to the second panel). When the accent is Turkish, a Turkish name is evaluated as more socially attractive than a Flemish name for both Standard and Colloquial Dutch (i.e. the third vs. fourth panel). The combination of a Turkish name and a Turkish accent is not downgraded vis-à-vis the combination of a Turkish name and a native accent (for either variety), but for Standard Dutch the Turkish name with either accent does score lower than the combination of Standard Dutch with a Flemish name and a native accent. Furthermore, the combination of a Flemish name and a Turkish accent receives the lowest scores of all guises.

The difference between Scenario 1 and 2 on Social Attractiveness is shown in Figure 4. On the whole Scenario 2 is evaluated as less socially attractive than Scenario 1, but the difference appears to be only borderline significant. This effect is a consequence of having the same speaker operate under various guises. As Table 3 in 2.3 shows, the difference between Scenario 1 and 2 is essentially that the same speaker (i.e. either a native speaker or a Turkish speaker) switches varieties in his two scripts (i.e. Standard Dutch becomes Colloquial Dutch and vice versa) and appears with names of a different ethnic origin (i.e. Flemish names become Turkish names and vice versa). We do not have a plausible explanation for this effect.

In the analysis of PA2, i.e. the Superiority dimension, only the interaction between Variety, Accent and Name was statistically significant; none of the controlling variables were (Marginal R² = 16.394 %, Conditional R² = 32.493 %). Table 5 and Figure 5 show the most important effects (see table A.2 in the Appendix for all estimated effects). Standard Dutch scores higher on Superiority than Colloquial Dutch for both a speaker with a Flemish name and a native accent (first panel) and a speaker with a Turkish name and a Turkish accent (last panel). In the other two combinations of accent and name, there is no significant difference in the evaluations of the two varieties. For both varieties, we find clear evidence of downgrading of speakers with a Turkish name and a Turkish accent compared to speakers with a Flemish name and a native accent, and compared to speakers with a Turkish name and a native accent. Speakers with a Flemish name and a Turkish accent (third panel) again receive the lowest scores (regardless of variety). Finally, keeping the accent the same (i.e. comparing the first two panels with each other and the last two panels with each other), there are no significant differences in how either Standard Dutch or Colloquial Dutch is evaluated on Superiority when the speaker has a Flemish name or when he has a Turkish name.

In the analysis of PA3, the Dynamism dimension, the three-way interaction between Variety, Accent and Name was not statistically significant and neither were any of the controlling variables. As a consequence, we refitted the model with the three-way interaction decomposed into the three two-way interactions, i.e. between Variety and Accent, between Variety and Name, and between Accent and Name, of which only the interaction between Variety and Name was statistically significant (Marginal R² = 8.665 %, Conditional R² = 43.287 %). The results can be seen in Table 6 and Figure 6 (see Table A.3 in the Appendix for all estimated effects). Standard Dutch spoken by someone with a Flemish name is evaluated as less dynamic than the combination of a Flemish name with Colloquial Dutch, and than a Turkish name combined with either variety. The latter three combinations themselves are evaluated more or less the same, since the differences between them are not statistically significant. Overall, the Turkish name received high ratings on this dimension, with no significant effect of Variety.

The goal of this experiment was to gauge the combined influence of name, accent and variety on speaker evaluations. Focusing on name-based biases first, we could not observe any consistent downgrading of the Turkish name on any of the three dimensions. We did find some significant differences between Flemish and Turkish names for the Social Attractiveness dimension: Turkish names with a native accent were downgraded vis-à-vis Flemish names with a native accent for Standard Dutch, but upgraded for Colloquial Dutch. In addition, Turkish-named speakers, regardless of the variety or accent they used, scored significantly higher on Dynamism compared to standard speakers with a Flemish name. The latter finding is similar to that of Grondelaers and Van Gent (2019) for Moroccan Dutch in the Netherlands: in their study, any Moroccan accent (regardless of strength) was deemed more dynamic than indigenous (Limburg or Randstad) accents. In contrast to their observation that Moroccan speakers were consistently downgraded on Superiority on the basis of their Arabic names, however, our results suggest prima facie that the Flemish participants in this study do not have such a bias against a Turkish name. Another possibility, however, is that our respondents were aware of the fact that name-based discrimination is socially unacceptable. Awareness of this issue among the general public has been raised in recent years, e.g. by mediated debates on the use of so-called correspondence tests and field experiments to measure discrimination on the labour and housing market (see section 1). Awareness of the undesirability of name-based discrimination may thus have caused our respondents to underreport negative attitudes towards Turkish names.

Regardless of what may have caused the absence of a name-based bias, however, we did find clear evidence of an accent bias, which penalises the exogenous accent. Whereas on Social Attractiveness, a Turkish accent spoken by a speaker with a Turkish name was not downgraded vis-à-vis a native accent spoken by a speaker with a Turkish name (neither for Standard nor Colloquial Dutch), the combination of a Turkish name, a Turkish accent and Standard Dutch was downgraded vis-à-vis the combination of a Flemish name, a Flemish accent and Standard Dutch. The accent bias is however clearest for the Superiority dimension, on which the Turkish accent was categorically downgraded. This finding is in accordance with the research overviewed in section 1, showing that foreign-accented speech generally provokes negative evaluations on status-related traits. The findings for Social Attractiveness and Dynamism moreover suggest that in Flanders the Turkish accent does not carry a more informal or “covert” type of prestige.

Our respondents’ attitudes were most negative towards speakers with a Flemish name and a Turkish accent, a combination which was consistently downgraded on both Social Attractiveness and Superiority. It is possible that the Flemish name created the expectation of the speaker being an L1 speaker of Dutch and that divergence from this expectation provoked negative reactions (cf. Language Expectancy Theory, see Burgoon and Miller, 1985 and Burgoon et al., 2002). In present-day Flanders, multilingual speakers of mixed ethnic descent are however increasingly less rare, and especially in urban, multi-ethnic and multilingual contexts ethnic majority youth have been observed to adopt speech patterns associated with their ethnic minority peers (Jaspers, & Van Hoof, 2017; a mediatised example of this phenomenon is provided by the TV series Grond, in which the character JB is of Flemish descent and speaks a Moroccan-flavoured street slang). The combination of a Flemish name and an exogenous accent are therefore not in se unrealistic. Nevertheless, the combination of a Flemish name and a Turkish accent was clearly evaluated negatively by our respondents.

As for the impact of variety, we found that speakers with a Turkish name and a Flemish accent are considered more socially attractive when speaking Colloquial Dutch than when speaking the standard, whereas the reverse is true for speakers with a Flemish name and accent. The latter finding contrasts with previous research, which found fairly consistent associations between Colloquial Dutch and Social Attractiveness (e.g. Ghyselen, 2009, Impe, & Speelman, 2007; Lybaert, 2017). Our findings for Superiority do show an upgrading of Standard Dutch vis-à-vis Colloquial Dutch, but this tendency only reached significance for the Flemish speaker with a native accent and the Turkish speaker with a Turkish accent. For Dynamism, we found the expected positive attitudes towards Colloquial Dutch compared to the standard only in the Flemish-named speaker, with no effect of Accent. The fact that Turkish-named speakers received positive evaluations on Dynamism regardless of the variety they spoke, suggests that Turkish men are in general thought of as “cool” and “tough”. We did not find a similar effect for the Turkish accent.

It can be noted that our factor analysis showed that only “cool” and “tough” clustered as a separate, third dimension in this study (similar to Grondelaers and Van Gent’s (2019) findings). Both traits are examples of what Grondelaers and Van Gent (2019) called “macho Dynamism”. The other traits that we selected, viz. “self-assured” and “easy-going”, which bear more resemblance to what Grondelaers and Van Gent call “yuppie Dynamism”, appeared to be more closely related to the traditional status and social attractiveness features in our study. The fact that we failed to categorise all traits we originally selected as ‘dynamic’ under one dimension casts doubt on the assumption that both types of traits can be viewed as a single evaluative dimension (Grondelaers, & Van Gent, 2019, p. 10). More research seems needed to map the precise relation between the two types of traits which have both been classified as “dynamic” in the literature.

A few caveats are in order when interpreting our results. We need to bear in mind that in our speech clips a single feature marked the difference between Standard Dutch and Colloquial Dutch, viz. the second person singular pronoun. Although that morphosyntactic feature appeared to be salient enough to evoke differential evaluations, colloquial language use in natural settings mostly includes multiple markers at once, as do most attitudinal studies focusing on endogenous variation (cf. Grondelaers & Speelman, 2013; Rosseel et al., 2019). It cannot be ruled out that for instance Dynamism associations are typically triggered by other markers. It therefore remains to be seen if similar associations as the ones we found in this study can be evoked when employing different markers, or several markers at once. In addition, it should be noted that we did not ask our respondents to identify the ethnic background of the speakers they evaluated (cf. Lindemann, 2003 for a critique of this modus operandi). As a consequence, we cannot be certain that the respondents in the speaker evaluation experiment recognised the names and the accents as being of Turkish origin, and therefore can only attribute the associations we found to the non-native, foreign character of the names and accents we deployed, and not to their Turkish provenance specifically. Hence, no solid conclusions can be drawn on the presumed stigmatised (or ‘plebeian’) character of the Turkish accent in Flanders on the basis of these data.

In this paper, we presented a speaker evaluation experiment in which we investigated the combined influence of name (Flemish vs. Turkish), accent (native vs. Turkish) and language variety (standard vs. colloquial) on attitudes of Flemish ethnic majority L1 speakers of Dutch towards other speakers of Dutch. In this experiment the Turkish accent was evaluated as inferior to the native accent in terms of status, and did not carry the covert prestige that is sometimes ascribed to the Dutch of ethnic minorities (cf. a Moroccan accent in the Netherlands). In contrast to this accent bias, we found no consistent name-based bias against Turkish speakers. Even though the latter finding could be the result of a social desirability bias, these results suggest that L1 speakers of Dutch are less reluctant to overtly differentiate between people on the basis of an exogenous accent than on the basis of ethnic origin (as indexed by a name). Note that this finding holds for L2-accented speakers who exhibit native-like grammatical control in Dutch and knowledge of ‘endogenous’, colloquial features, as the speakers in our stimuli did.

It remains an open question whether the accent bias found in this study could also be retrieved in real-life settings in Flanders. Given the limited ecological validity of an experiment such as ours, which tested fairly decontextualised stimuli in a lab setting, our finding calls for further research investigating accent bias in more naturalistic contexts. Especially gatekeeping contexts such as job interviews, in which accent bias could potentially pose a serious threat to ethnic minority members’ chances at success, merit further attention in the Dutch-speaking language area (cf. Hosoda and Stone-Romero (2010), Sharma et al. (2019) and Schmaus and Kristen (2021) for experimental studies on accent bias in job recruitment contexts in the US, the UK and Germany respectively).

Finally, it remains uncertain whether the accent bias we found in this study was caused by the foreignness of the accent in general, or by the Turkish accent in particular. Further research is needed to get into view the impact of the recognisability of accents on speaker evaluations and to answer the question whether accents which are categorised in the literature as prestigious versus non-prestigious elicit different evaluations in the Dutch language area as a whole, and in Flanders in particular (cf. Lindemann, 2003). As more languages and accents enter the Flemish linguascape as a result of migration, and the number of L2 speakers of Dutch continues to grow, the social relevance of this type of sociolinguistic research will only increase.