European movers’ language use patterns at home: a case-study of European bi-national families in the Netherlands

Language use patterns of migrant populations are often considered to indicate migrants’ integration into the host society (Alba et al.2002; van Tubergen and Kalmijn 2009). Research showed, for example, that migrants’ proficiency in the dominant language is often important for their successful incorporation into the labor market (e.g. Chiswick and Miller 1995; Shields and Price 2002) and for their educational career (e.g. Rumberger and Larson 1998; Entorf and Minoiu 2005). It is thus not surprising that a considerable number of studies focused on language use patterns and proficiency of first, second and third generation migrants and their offspring (e.g. Chiswick and Miller 1996; Espenshade and Fu 1997; Shields and Price 2002; van Tubergen and Wierenga 2011; Vervoort et al. 2012; Rebhun 2015).

In contrast, studies focusing on language use patterns among intermarried couples, consisting of a native and foreign partner, are rather scarce. Mixed unions, however, form an interesting group, as the majority and minority language come together in one household, which potentially leads to complex negotiations with respect to which language should be used with whom. In this paper, we investigate which individual and environmental factors are involved in the choices for the use of the majority, minority and/or third language in such families, using the Dutch data from the EUMARR survey, a unique data set on European bi-national couples. Although we acknowledge the existence and relevance of other forms and definitions of mixed families (for a detailed discussion, see Deprez et al.2014), we thus focus on couples consisting of two partners of a different nationality. This specific scope allows us to investigate how individual and familial trajectories in combination with the migration history of one of the partners influence language use patterns in mixed families (Unterreiner 2014).

Our contribution to the literature is threefold. First, the few available studies on language use in bi-national families in Europe mainly focused on a few non-European migrant groups such as Turks or Moroccans (e.g. van Tubergen and Kalmijn 2009; Vervoort et al.2012), despite the fact that European migrants form a substantial share of the migrant population in Europe today (Van Mol and de Valk 2016). Intra-EU movers can be considered to be a specific group of migrants, as they have the right to freedom of movement. As European citizens, no language requirements and/or tests are imposed on them, and they are considered to be ‘unproblematic’ in terms of social cohesion in receiving societies (Koelet et al.2017). Hence, their experiences with multiple languages might well be different compared to migrant families and bi-national couples involving a non-European partner. This might particularly hold true as in contrast to non-EU migrants, for whom knowledge and proficiency in the majority language are considered crucial for their societal integration, multilingualism of intra-EU movers is often considered an asset by European policy-makers (European Commission 2012). In addition, most research on immigrants’ language use focused on English-speaking countries such as Australia (e.g. Chiswick and Miller 1996), the United Kingdom (e.g. Kirsch 2012; Parameshwaran 2014) and the United States (e.g. Velázquez 2014; Rebhun 2015). Second, we include the ‘market value’ (Bourdieu 1991) of different languages in our study. It can be expected that not all languages will be equally appreciated, which potentially influences the choices migrants make regarding language use. The status of English as a global language might, for example, significantly influence migrants’ willingness to learn the majority language in non-English-speaking countries. To our knowledge, there are as yet no studies empirically investigating the connection between language status and language use patterns within bi-national families. Finally, we connect language use patterns within mixed families to the broader microsystem in which individuals are situated. We start with the assumption that language choices within families are – apart from negotiations within the family – connected to the social environment of family members in the Netherlands and abroad. Our paper thus also advances our empirical understanding of the social embeddedness of language use patterns.

In sum, we adopt a micro-level approach and examine which factors influence language use patterns in European bi-national families in the Netherlands, the embeddedness of these practices in social environments, and the role of language status in the development of language practices.¹

The family context is an important domain to consider when studying language choice patterns of migrant families (Stevens 1992; Alba et al.2002; Pauwels 2005; Lanza 2007). For many migrants, the home offers a protective environment in which customs and traditions can be continued, including the use of their own language. The minority language not only serves as a means of communication, but can also be considered to be a fundamental component of the minority culture (Smolicz et al.2001), which migrants generally want to pass on to the next generation. After all, languages are ‘best able to express most easily, most exactly, most richly, with more appropriate over-tones, the concerns, artefacts, values, and interests of [a given] culture’ (Fishman 2007: 72). Languages thus express culture as well as reproduce it, as words are linguistic ‘symbols’ of culture (Jaeger and Selznick 1964). As such, languages form an essential tool for passing on cultural knowledge and practices.

Three types of cultural transmission can be distinguished: vertical, horizontal and oblique (Trommsdorff 2009). Vertical socialization processes point to purposeful transmissions between generations within a household, whereby cultural elements are passed on to the next generation. Intergenerational transmission is considered to be vital for maintaining the heritage culture over generations (Phalet and Schönpflug 2001). Horizontal transmission, in contrast, refers to indirect influences within generations, such as from peers and wider social connections. Finally, oblique transmission refers to cultural transmission between generations which is not related to the household (e.g. from parents’ friends to children). In this paper, we directly assess horizontal (language use between partners) and vertical (language use with children) transmissions, and investigate which individual and contextual factors influence these language use patterns. Indirectly, oblique transmission is also considered, as we investigate whether the composition of parents’ friendship network is related to patterns of language use with their offspring.

In a migratory situation, language transmission might not always be easy. The minority parent might, for example, have a wish to use the minority language in the household in order to maintain connections with the country of origin and/or pass on cultural practices. Simultaneously, however, they might be strongly in favor of acquiring and transmitting the majority language to their children, insofar as it facilitates adaptation and inclusion in the host society (Kirsch 2012). The vertical transmission of minority languages thus appears challenging (e.g. Tuominen 1999), especially when a native partner is involved. Research in the United States (Grenier 1984; Stevens 1992; Alba et al.2002) shows, for example, that exogamous marriages significantly increase the probability of using the majority language in the household for first, second and third migrant generations.

Several individual factors have been documented to influence language use patterns of migrants. First, socio-economic status may play an important role in the maintenance or disappearance of a minority language. The direction of this relationship, however, is unclear, and conflicting results have been reported in the European context. A study among Greek, Italian, Russian and Turkish migrant families in Germany, for example, revealed a strong effect of educational level on minority language retention (Nauck 2001). The direction of this effect, however, differed according to migrants’ origin: Greek, Italian and Turkish families with lower education levels had higher retention levels, whereas for the Russian-speaking families, higher educational levels were related to high language retention levels. For the Netherlands, van Tubergen and Kalmijn (2009) reported that higher levels of education of both respondents and their partners lead to higher usage of Dutch in the household. Furthermore, socio-economic status is related to families’ financial capacities enabling, for example, frequent visits to the country of origin which enhance vertical language transmission (e.g. Pauwels 2005; Parameshwaran 2014). Increased financial capacity also enables to purchase books and minority language material enhancing the bilingual education of children (Tuominen 1999). Consequently, we expect the propensity of mainly using the minority language in bi-national families to be correlated with higher educational levels of both partners (hypothesis 1).

Second, vertical language transmission might be gender-specific. Several qualitative studies indicate a central role of mothers in intergenerational language transmission processes (e.g. Kirsch 2012; Velázquez 2014), as they are more inclined toward using the minority language with their children compared to their husbands (Castonguay 1982; Grenier 1984). Several quantitative studies, however, contradict these findings (Stevens 1985; De Houwer 2007), and do not find any differences between parents’ language use with their offspring. In this paper, we test the hypothesis that when the mother is from another EU country, the minority language will be significantly more used in the family (hypothesis 2).

Finally, it has been suggested that the employment status of the minority parent can influence language use at home, especially with the children. Okita (2002) showed, for example, that the work demands of majority group fathers meant that they invested less time in their children’s (majority) language development. Consequently, in these couples, the mothers had more freedom to transmit their own (minority) language to their children. Conversely, employed minority parents might be more exposed to the majority language in their workplace and have less time to spend with their children, leading to a potential decrease in their use of the minority language with their offspring. As a result, we expect that minority parents who are not in full-time employment might have more time to spend on language learning of the children, leading to a greater use of the minority language in their communication with them (hypothesis 3).

Apart from considering individual characteristics, language use patterns within bi-national families might also be related to the social connections of individual family members. Individuals’ decisions, actions and experiences interact with broader micro-, meso- and macro-systems. This might also hold true for language use patterns. Research on macro factors revealed, for example, that the use of and/or proficiency in the majority language is related to the size and concentration of a minority group (Grenier 1984; Stevens 1992; Chiswick and Miller 1995; van Tubergen and Kalmijn 2009; Vervoort et al.2012) as well as to a country’s minority language policies (e.g. Portes and Hao 1998; Lanza 2007). Here, we focus on the microsystem, namely the connection between language use patterns in the bi-national family and social relationships with family and friends in the Netherlands and the country of origin.

Bi-national unions involving a native partner might often be confronted with competing demands and expectations of the native partner’s and the migrant’s family. The family in the destination country might, for example, place strong expectations on the use of the majority language in the family, increasing the odds of using this language at home. Therefore, we expect that the main use of the majority language will be positively correlated with the frequency of contact with the local (Dutch) family (hypothesis 4). However, the household members might simultaneously perceive strong preferences on the part of the transnational family to maintain the minority language, especially when children are involved (Stoessel 2002). Tuominen (1999), for example, suggested that vertical language transmission is related to frequent phone calls to the country of origin. Kirsch (2012) and Stoessel (2002) on their turn showed that migrants themselves also place a lot of importance on transmitting the language to their children, as they often think this facilitates good relationships with the family abroad and teaches their children to comfortably move within the cultural environment of the country of origin. Consequently, we expect that the main use of the minority language will be positively correlated to frequency of contact with the transnational family network (hypothesis 5).

Apart from the family, friends also play a role (e.g. Li 2006), although perhaps to a lesser extent. Social connections with friends who share the same mother tongue in the host country might, for example, be helpful for maintaining and vertically transmitting the minority language. Several studies showed that contact with peer groups valuing the heritage language can be important for maintaining that language (Okita 2002; Li 2006), as they can offer external language support (Schaberg and Barkhuizen 1998). Okita’s study of Japanese migrants in the United Kingdom, for example, revealed that contact with other Japanese migrants was conducive to maintaining the use of Japanese at home (Okita 2002). Nevertheless, social connections can also have the opposite effect. Migrants might, for example, receive negative feedback from members of the majority group on minority language use, making them cautious about overtly using their native language outside their home (Okita 2002). To our knowledge, the studies of Braun (2010) and Vervoort et al. (2012) are the only studies investigating the relationship between majority language proficiency and social contacts with ethnic and native networks. Both studies showed that social contact with natives is positively correlated with majority language use and proficiency, while an opposite direction was found when considering social contact with co-ethnics. For our case study of European bi-national couples, we hence expect similar results. The higher the share of natives in the core friendship network, the higher the probability of using Dutch in the household (hypothesis 6). Similarly, the higher the share of friends with the same mother tongue, the higher the probability of using the minority language for communicating at home (hypothesis 7).

Finally, not all languages have an equal status (e.g. de Swaan 1993; Crystal 2003). Languages that are spoken by a large population in the world as well as languages with a high prestige might be more important resources, potentially yielding profitable returns (Rössel and Schroedter 2014). Indeed, the limited body of existing research shows that the ‘market value’ (Bourdieu 1991) of a language in a specific context plays a role in family language policies (Schaberg and Barkhuizen 1998; Kirsch 2012). Considering our study context, the Dutch government ‘attaches great importance to the Dutch language’ (Benedictus-van den Berg 2012: 162). This is reflected, for example, in the requirement of passing an exam on knowledge of Dutch language for non-EU migrants. Consequently, Dutch clearly has the highest status for participating in most areas of societal life. English, however, is also highly valued in the Netherlands. It is the first foreign language pupils learn at primary school, and the number of schools offering bilingual programs in Dutch and English is on the rise (Benedictus-van den Berg 2012). English is the most widely known foreign language in the Netherlands, with 90% of the population claiming to be able to speak it, followed by German (71%), and French (29%) (European Commission, 2012). Therefore, in this paper we consider English, German, French and Spanish separately, given the probably different utility status for European migrants compared to some smaller European languages (e.g. Danish, Luxembourgish and Slovene). These smaller languages may be subjectively perceived as being less useful, and partners might put less effort into speaking or transmitting that language (hypothesis 8).

Our analyses are based on the EUMARR survey. This unique survey collected data (2010–2012) among European bi-national unions, consisting of a European and a native partner, in Belgium, the Netherlands, Spain and Switzerland. The survey contains detailed information on what language(s) respondents speak, as well as what language they use for communicating with their partners, children and colleagues. Although the EUMARR survey contained information on language use, it did not capture language proficiency levels, which is a common limitation for surveys of this type (van Tubergen and Kalmijn 2009). Nevertheless, for our study this is not a key issue, as we seek to understand which factors influence language use patterns in European bi-national families.

The Dutch data were collected through an online survey in The Hague and Amsterdam. Couples were sampled through the municipal population register GBA (Gemeentelijke BasisAdministratie) of both cities. The two most common bi-national combinations were Dutch–German and Dutch–UK couples, followed by Dutch–French, Dutch–Belgian, Dutch–Spanish and Dutch–Italian. In terms of gender balance, in our sample, Dutch men are more likely to be in a relationship with a European women than European men being in a relationship with Dutch women. Both the bi-national combination and gender balance are in line with the overall composition of European bi-national couples in the Netherlands (see van Wissen and Heering 2014). Ninety-three percent of the respondents completed the questionnaire online. Respondents could answer the questionnaire in one of three languages: Dutch, English and French. The overall response rate was 37.1%, which is in line with response rates of this type of survey in the Netherlands (see e.g. Groenewold and Lessard-Phillips 2012). All respondents were between 30 and 45 years old. This age criterion was established in line with the broader aims of the EUMARR-project, namely securing a homogeneous sample of respondents who started their unions after the establishment of the right to freedom of movement. We filtered out respondents with a non-European first nationality, as well as Flemish respondents, as Flemish is a very similar language to Dutch and including these respondents in the analysis would bias the findings. As a result, our analyses include 627 individuals – 306 Dutch nationals who have a relationship with an EU-national (presented as Dutch–EU in the tables), and 321 EU-nationals having a relationship with a Dutch national (presented as EU–Dutch).

The dependent variables measure respondents’ self-reported main language of communication. A first variable measures the main language of communication between partners, a second the main language of communication with their children (if applicable). These two variables are based on the questions ‘Which languages do you use when talking to your partner/spouse?’ and ‘Which languages do you use when talking to your children?’. Both questions were open-answer questions, whereby respondents could indicate a language under the option ‘I usually speak’. Based on the languages respondents indicated, we created two dichotomous variables, indicating whether the respondent usually spoke (1) Dutch (reference category); (2) her/his native European language; or (3) a third language. This third category represents languages that are not the mother tongue of neither of the partners.

First, gender is included as a dichotomous variable (0 = male, 1 = female). Second, the employment status of the respondent and of their partner is measured by two dichotomous variables (0 = unemployed, 1 = employed). As we do expect this not to influence communication between partners, we only include this variable in the models on communication with children. Third, socio-economic status is measured by two variables. As we do not have a reliable estimator of respondents’ household incomes, the subjectively assessed social position of respondents in the Netherlands is used as a proxy, based on the question ‘When you consider your household income from all sources and the wealth you and your partner may have accumulated, could you tell on which step you would place yourself’, ranging from 0 to 10 (0 = lowest level, 10 = highest level). Besides social position, we included a variable indicating respondents’ educational level. Educational status is measured by an ordinal level variable ranging from 1 to 9 (1 = less than primary, 9 = doctoral or equivalent). We recoded this variable into three categories, based on the International Standard Classification of Education (ISCED 2011), namely a lowly (ISCED level 0–4), a medium (ISCED level 5–6) and a highly (ISCED level 7–8) educated group.

First, we include a variable indicating the share of friends in the core friendship network with the same mother tongue as the European partner. This variable is measured as the share of co-national friends among the closest friends in the Netherlands for European respondents or the share of friends having the same nationality as their partner for Dutch respondents, of which respondents could indicate up to a maximum of five. Along with co-national friends, we included international friends who would be expected to speak the same mother tongue – such as Austrian friends for Germans, or American, Australian or Irish friends of British migrants. This is based on the assumption that it is the use of the same language rather than nationality that might motivate migrants to use their mother tongue. Second, we constructed a variable measuring the share of Dutch friends, as this can be expected to increase the tendency of using Dutch. Third, we included two variables indicating the frequency of contact with the family network in the Netherlands as well as with the transnational family network. Both variables are based on a Principal Component Analysis with a fixed one-factor-solution, combining the frequency of contact with these respective networks in terms of (1) traveling to meet members of this network; (2) receiving visits from members of the network; and (3) having phone conversations (including conversations over internet). Respondents could rate each of these items from 1 (rarely) to 8 (daily). We use the regression scores of both variables.

We included a categorical variable indicating whether the native language of respondents or their European partner is (1) German; (2) English; (3) French; (4) Spanish; or (5) another European language (reference category). We distinguished between these five categories as English, German, French and Spanish are the four most widely spoken foreign languages in the European Union (European Commission 2012). For the interviewed European nationals, this variable is based on the question ‘In which language were you raised?’. For the Dutch nationals, this variable is based on the nationality of their partner.

We included a range of control variables in the models that have been found relevant in earlier studies. First, age and age at first birth (where applicable) are measured as two continuous variables. Second, migration duration is included as a continuous variable indicating the number of years a migrant spent in the Netherlands. We included this information as it has been shown that the longer a minority partner lives in a host country, the more likely it is the majority language will be spoken among bi-national couples (Varro 1984). As a considerable number of respondents (n = 110) did not provide this information, we had to manually estimate this number based on other variables. We distinguished between four periods of migration duration (2–5 years, 6–10 years, 11–15 years and 16 years or more). Third, two variables specifically relate to language use. First, a variable indicating whether the respondent is able to fluently speak the language of their partner (0 = no, 1 = yes), as this logically has implications for the available language choices. Second, we take the language used between partners into account in our analyses focusing on communication with children.

We initially conducted multinomial regression models on language use patterns. Nevertheless, for the models on communication between partners, parameters were seriously biased because of the little variability in third language use when the European partner’s native language was English. Therefore, we use separate binary logistic regressions to investigate which factors are correlated with use of the minority or third language, excluding English from the models on third language use. Furthermore, only two respondents communicate with their children in a third language (see Table 1). As a result, we only focus on Dutch and the minority language in the models on communication with children. In addition, only eight of the Dutch individuals in our sample speak the minority language with their children. Consequently, we only focus on the minority parent when discussing vertical transmission processes within mixed families.

Finally, an analysis of the independent variables indicated that more than 10% of our sample had missing information on one or more of the variables included in the models (see Table 1). As a result, we used multiple imputations to decrease possible bias (Allison 2001), based on the Markov Chain Monte Carlo Method. Five data sets were imputed, and the average estimates across these five data sets are reported (Little and Rubin 2002).

The descriptive statistics (see Table 1) show that Dutch turns out to be used most often for communication between partners. No differences between EU-nationals and Dutch individuals were found, suggesting a shift of EU-nationals toward Dutch language use with the partner. Interestingly, Dutch respondents use the language of their partner significantly less often in the household compared to the surveyed Europeans (χ²= 27.79, p < .001). Two-thirds of the non-Dutch EU-nationals indicated to use their native language as the main language of communication with their children, whereas almost one-third of the EU movers indicated a shift to the use of Dutch as the main language of communication (χ²= 235.18, p < .001). No differences could be detected between the Dutch and European respondents regarding the mother tongue of the non-Dutch partner.

Furthermore, descriptive analyses of the individual characteristics reveal differences in the gender composition of both groups (χ²= 76.19, p < .001). The surveyed Dutch nationals are more likely to be male. Significant differences can also be detected considering respondents’ self-assessed position in society (t(575) = 4.04, p < .001). EU-nationals place themselves lower on the scale compared to their Dutch counterparts. No differences were found between the two groups, however, regarding their educational attainment. Considering employment status, EU-nationals are more likely to be unemployed compared to their Dutch counterparts (χ²= 7.65, p < .01). Similar findings are observed with regard to the employment status of the partners of respondents (χ²= 12.20, p < .001).

Significant differences can also be observed between both groups in terms of social network characteristics. EU-nationals have more frequent contact with the transnational family network (t(577) = 5.11, p < .001) and less frequent contact with the Dutch family (t(580) = 4.30, p < .001) compared to Dutch individuals partnering with an EU-national. Both groups also differ considering the share of friends speaking the mother tongue of the European partner (t(573) = −10.76, p < .001), and the share of friends from the majority population (t(573) = 13.28, p < .001). Logically, Dutch respondents have considerably more native friends, and they have fewer friends who speak the mother tongue of their partner. Interestingly, EU-nationals have proportionally more Dutch friends than friends speaking their mother tongue in their close friendship network in the Netherlands.

Finally, our analysis revealed differences considering several control variables. Whereas no differences could be detected regarding age, significant differences are detected considering age at first birth (t(481) = 2.01, p < .05) and migration duration (χ²= 13.42, p < .01). EU-nationals are slightly younger at their first childbirth compared to Dutch nationals, and the EU partners of the Dutch nationals in our sample have resided longer in the Netherlands than the surveyed EU-nationals.

Next, we present binary logistic regression models on the main language of communication between partners in Table 2.

Model 1 shows that male respondents are less likely to use the minority language. However, no other significant relationships between individual characteristics and the use of Dutch or the minority language could be detected in models 1 and 3, contradicting hypothesis 1 on the relationship between socio-economic background and language use. Interestingly, for European respondents, the propensity to use the minority partner’s language seems to be related to social relationships with friends sharing the same mother tongue in the Netherlands, in line with hypothesis 7. The higher the share of such friends in the European respondent’s social environment, the higher the probability that the European language will be used for communication between partners. Interestingly, the analysis reveals no relationship between the share of Dutch friends, frequency of contact with the Dutch and transnational family network and the use of the European language or Dutch as the main language of communication between partners, contradicting hypotheses 4, 5 and 6. Furthermore, our analysis confirms hypothesis 8 as the language status of the minority language seems to play a key role in language use patterns. If the minority language is English, French or Spanish, the probability of using this language between partners is much larger compared to situations where a lesser-spoken European language is involved. This particularly holds true for English, as model 1 shows that Dutch respondents are also more likely to communicate in English if their partner originates from an English-speaking country. In contrast, when the minority language is German, the propensity of using of Dutch is greater for Dutch individuals.

Models 2 and 4 overall show the same pattern, as most of the individual characteristics are not significantly correlated with the use of a third language, except for educational attainment. When the European partner is lower educated, the likelihood of using a third language decreases. Furthermore, the results do not reveal any significant correlations in terms of social environment characteristics. Finally, the findings show that language status of the mother tongue of the minority partner also plays a role in using a third language. Compared to the other European languages, the propensity to use a third language is much lower in bi-national households involving a German-speaking partner, whereas for Spanish-speaking partners, the likelihood is higher. In addition, it is worth noting that for all models, migration duration is positively correlated with an increasing use of the majority language. The longer the European partner resides in the Netherlands, the higher the likelihood of using Dutch for communication with their Dutch partner.

In Table 3, we present two models on the main language of communication with children. Model 1 reveals that with increasing levels of parental education, the propensity of using the minority language with children grows, confirming hypothesis 1 for language use in bi-national households with children. Furthermore, the model reveals a significant correlation between the share of mother tongue friends in the local friendship network and the use of the minority partner’s language for communication with children. In addition, confirming hypothesis 8 again, the status of languages shows to play an important role in using the minority language with children. The findings strongly suggest that when the minority parent’s native language is spoken by a larger number of people in Europe, the propensity of using that language with their offspring increases.

In model 2, we control for the language used between both partners. Most of the relationships identified in model 1 persist in model 2. The choice for the use of the minority language with children is still correlated with parental education as well as the status of the European minority language. This particularly holds true for German and English. Furthermore, model 2 supports the idea that vertical language transmission is gender-specific. Mothers are found to be more likely to use the minority language with their children compared to fathers, confirming hypothesis 2. Contrary to model 1 and our findings on language use between partners, however, the relationship between the share of mother tongue friends in the local friendship network and minority language use with children disappeared. Interestingly, model 2 reveals that the most significant factor related to the use of the European language other than Dutch for communication with children is the use of the minority partner’s language or a third language between both partners. As such, language practices between parents have their repercussions on language use with their children. If bi-national partners use a language different from the majority language in their relationship, the likelihood of using that language with children clearly increases. Finally, our results did not confirm hypothesis 3, namely that non-Dutch parents who are not in full-time employment might have more time to spend on their offspring’s language learning, increasing the propensity of using the minority language.

Languages can be considered to be essential components of cultures; they both reflect and reproduce culture. As such, language is an important tool for ensuring cultural continuity over the generations, particularly in migrant families in surroundings where the main language of communication is different from their mother tongue. In this paper, we were interested in the individual and environmental factors that shape language use patterns in European bi-national households, and took the Netherlands as a case study. We specifically focused on the horizontal, oblique and vertical transmission of languages in European mixed families.

The results indeed revealed patterns of horizontal (between partners and social networks) and vertical (between parents and children) transmission of language. We did not find convincing empirical evidence, however, of oblique transmission. Furthermore, the findings suggest that language use patterns between partners and with children are influenced by a different set of factors. This does not, however, mean that these patterns are uncorrelated: we provide strong evidence for the connection between the main language used between partners and the language used for communication with their offspring. The following conclusions concerning patterns of language use among European bi-national couples can be drawn.

First, surprisingly most of the individual characteristics which we expected to correlate with the use of the majority or minority language between partners did not prove significant in the analyses, except that Dutch male respondents are least likely to communicate in the minority language with their partner, and lowly educated Europeans are less likely to use a third language. This last finding can simply be explained from the perspective that such individuals probably dispose of less ‘language capital’ in terms of mastering a foreign language, thus excluding the possibility of using one. This result is largely in line with existing research on social class, second language knowledge and European identity (e.g. Fligstein 2008). Individual characteristics, however, appear to be more important when considering language use patterns with children. Our results revealed that minority parents more often opt for their native language if they are more highly educated. Furthermore, the findings illustrate the central role mothers often play in the vertical transmission of language, as we showed that the propensity of using the minority language with children decreases if the minority parent is male.

Second, our results partially support the idea that language use patterns, particularly between partners, are embedded within social environments wherein individuals, couples and families operate. Particularly for European respondents, the results indicate that the propensity to use a non-Dutch European language with the native partner increases with the share of mother-tongue friends in the surrounding network. Although a similar pattern with regard to the share of mother-tongue friends could be detected for communication with respondents’ offspring, this relationship did not persist when controlling for language use patterns between partners. The presented results hence reveal the horizontal transmission of language in social environments, but do not provide convincing empirical support for the existence of oblique transmission.

Third, our results suggest that the status of languages plays a role in choosing which language to use in European bi-national families. Households involving a German-speaking partner appear to be less likely to use a third language compared to those involving other European languages, and are more likely to use German for communication with their children. This finding might be related to the fact that Dutch and German are quite similar languages. Consequently, German movers in the Netherlands might face fewer barriers in learning the majority language, and find it easier to pass on this language to their offspring in an environment characterized by a language with a similar structure. Interestingly, English-, French- and Spanish-speaking intra-EU movers are also more likely to use their own language for communication with their partner. This might be due to the ‘market value’ of these languages in the world, particularly English, as well as to the fact that these languages are often spoken by other European populations. The propensity to meet and partner with a European, knowing one of these languages is higher compared to other languages. Similar findings were found for communication with children. For those minority parents speaking English or German, the two most widely spoken languages in Europe (European Commission 2012), a significant correlation with the use of this language with their offspring is detected. This indicates that vertical transmission of the mother tongue to the offspring might also be related to the relative status of the EU-mover’s native language on a European and global scale.

Although our data shine a unique light on language transmission processes of European migrants, some limitations apply. First, our data did not allow us to investigate language proficiency levels. Such insights, however, would be relevant, as research already indicates that speaking two languages at home does not guarantee full bilingualism for the children (Li 2006; De Houwer 2007). Future research among European migrant populations could therefore take into account language proficiency levels of parents and children. Second, we only dispose of data of European migrants with a native partner. It might be interesting to compare these couples with bi-national couples involving a non-European partner or consisting of two Europeans of different nationalities, as the patterns of language use in such families might be even more complicated. Regarding the latter, for example, they have to negotiate among three or possibility four languages (two minority languages, a possible third language and the majority language). Third, our analyses are based on cross-sectional data. As a result, we cannot detect any causal relations. Fourth, partners and children were not surveyed, which forces us to rely upon the self-reported data of one partner in the couple only. For a more complete understanding of language transmission processes, however, it would be relevant to map the preferences, proficiency level and usage of all members of bi-national families.

Altogether, our results show that language use within bi-national households is not taking place in a vacuum. Instead, the findings presented in this paper suggest that language use patterns should be situated within the wider environments in which they are formed. We observed a tendency among intra-EU movers to mainly use the majority language as the main language of communication in bi-national unions involving a native partner. Returning to the classical integration question often connected to languages use and proficiency, this result might indicate that these migrants are likely to linguistically integrate into the receiving society. At the same time, however, intra-EU movers are also similar to non-European migrant groups that were studied before: irrespective of origin, migrants intend to preserve and pass on their native language on to their children in a foreign context. Given the high value that the European Commission places upon multilingualism, as well as the expected link between multilingualism and European identity (see e.g. Fligstein, 2008), bi-national families should hence be considered as an important group for further investigations on the construction of a European society from below.

No potential conflict of interest was reported by the authors.

Christof Van Mol is Assistant Professor at the Department of Sociology at Tilburg University, the Netherlands. He is interested in intra-European migration processes, determinants, experiences and outcomes. He extensively published on these issues in leading journals including Geoforum, Journal of Ethnic and Migration Studies, and Population, Space and Place. His book on intra-European student mobility (2014, Palgrave Macmillan) received the 2016 Best Book Prize in Sociology of Migration (RC31) of the International Sociological Association.

Helga A. G. de Valk is theme group leader Migration and Migrants at the Netherlands Interdisciplinary Demographic Institute (NIDI, the Hague) and professor of Migration and the life course at the University of Groningen. Her research focuses on migration and integration, the transition to adulthood of immigrant youth, union and family formation, the second generation, intergenerational relationships in immigrant families, and European mobility. She was editor-in-chief of the European Journal of Population (2014-2018) and is recipient of the European Demography Award 2016.

Christof Van Molhttp://orcid.org/0000-0001-9275-101X