Multilingual writing practices as code choices: Dutch alongside French in private family letters

1 Introduction

Studies in historical multilingualism have examined different phenomena in a vast number of contact settings in Early and Late Modern Europe (e.g. Braunmüller and Ferraresi 2003; Franceschini et al. 2023; Rutten et al. 2017). Two common phenomena in multilingual societies are language choice (Rindler Schjerve 2003; Rjéoutski and Frijhoff 2018) and code-switching (Glaser et al. 2021; Schendl and Wright 2011), which have often been investigated as two discrete objects of research. This alleged opposition is rooted in different methodological approaches and research traditions, which typically treat language choice as a macro-level phenomenon in institutional or societal contexts, i.e. in terms of a sociology of language, whereas code-switching is examined more commonly in (socio)linguistics as a micro-level phenomenon of individual multilingualism. In recent years, however, the field of historical sociolinguistics has been shifting towards more holistic views on multilingualism and multilingual historiography (Bennett and Cattaneo 2022; Krogull 2021; Pahta et al. 2018a; Pavlenko 2023). As pointed out by Pahta et al. (2018b: 3), “[m]ultilingual societies were composed of multilingual individuals who used more than one language in their daily lives, even within a single utterance”. These individuals “deployed their linguistic resources according to the pragmatic constraints of the relevant community at the relevant point in time, often moving fluently from one language to another” (Pahta et al. 2018b: 15). In line with current developments, foregrounding the multilingual repertoires of individuals and social groups, we argue that it may be necessary to reassess linguistic choices on the one hand, and multilingual practices on the other, as contact phenomena that are more closely intertwined than discrete (Schendl 2012: 523–524). Our paper therefore sits at the intersection of what would traditionally be labelled as ‘language choice’ and ‘code-switching’.

A case in point is Dutch-French multilingualism in the Early and Late Modern history of the Northern Low Countries, i.e. the present-day Netherlands. Against the background of enduring contact between Dutch and French, and widespread multilingualism among certain groups, viz. the urban elites in the Netherlands, we seek to establish the proportion of monolingual versus multilingual practices in their private sphere. Loosely inspired by Fishman’s (1965) question “Who speaks what language to whom and when?”, and applied here to historical written data, we are particularly interested in who ‘mixed’ Dutch and French with whom and when, and under which circumstances. In order to identify the sociolinguistic profiles of those multilinguals who actually used Dutch alongside French, or French alongside Dutch, we apply two complementary methodological approaches. First, based on a substantial dataset of private family letters, we provide solid quantitative evidence on mono- and/or multilingual letter writing practices, taking into account diachronic and sociolinguistic factors. These findings form the baseline of our qualitative analysis, in which we zoom in on the phenomenon of multilingual (i.e. Dutch-French) letter writing without a clearly dominant main language, which exemplifies the multilingual practices also addressed in recent research. Beyond the concrete case study of Dutch-French family correspondence, this paper also seeks to suggest some methodological and theoretical implications for research into multilingualism and multilingual practices in historical, written texts.

In Section 2, we first outline the historical background of the multilingual Low Countries. In Section 3, we introduce our dataset of family letters and the methodological framework. We then present the quantitative and qualitative analyses of linguistic practices in Sections 4 and 5, respectively. Finally, the main findings are discussed in Section 6.

2 Historical background

The history of the Low Countries is characterised by extensive multilingualism. Low Countries is a historical description of the territory presently covered by the Netherlands, Belgium and Luxembourg. The Germanic-Romance language border cuts through this territory, rendering Belgium and Luxembourg multilingual states. We focus on the Netherlands or Northern Low Countries here, located north of the language border, yet also characterised by broad historical multilingualism.^[1] In addition to the range of Dutch (i.e. Franconian and Saxon) and Frisian dialects, the main written languages in medieval times were Dutch, Frisian, Latin and French, and Dutch, Latin and French from the early modern period onwards. Economic and ideological migration brought a range of languages to the Netherlands: labour migrants, for example, often came from French, (High and Low) German and Scandinavian areas, while religious migration also included speakers of Spanish, Portuguese and Yiddish (Lucassen and Penninx 1994: 30–35).

French was the main contact language in the Early and Late Modern period in the Low Countries.^[2] Medieval education was dominated by Latin and to a lesser extent Dutch, but in the late medieval urban trade centres, French became increasingly important. So-called French schools, initially targeting middle-class boys to prepare them for a career in trade, emerged in the fifteenth and sixteenth centuries (Dodde 2020: 59–60). They developed into fully-fledged educational institutions in the course of the seventeenth and eighteenth centuries, offering many other subjects such as German, geography and history in addition to French, arithmetic and bookkeeping (Dodde 2020: 79). An important impetus to the French school system came from the many Huguenot refugees who fled to the Netherlands: the First Refuge, associated with the massacre of St. Bartholomew’s Day (1572), brought at least 100,000 people to the north, where they constitued c. 10 % of the population (Frijhoff 2003: 131, 136). The Second Refuge, linked to the Revocation of the Edict of Nantes (1685), brought another 35,000 migrants to the north (Frijhoff 2003: 136). Dozens of Francophone churches were established in towns and cities across the Netherlands (Bots et al. 1985).

French was a functional choice in international relationships (trade, diplomacy, but also science and journalism) and a contact language through migration, but it also developed into a prestige language, as almost everywhere in Europe (Rjéoutski et al. 2014). The acquisition and use of French were associated with upward social mobility and with elite in-group communication, distinct from the Dutch-dominant lower ranks (Frijhoff 2015: 129; van Strien-Chardonneau 2014: 169–171). This means that French was used among the upper-middle and upper ranks, even in ego-documents such as private letters and diaries (Ruberg 2005; van Strien-Chardonneau and Kok Escalle 2017).

The widespread use of French led to anti-French discourses, often labelled as discourses of frenchification (Rutten et al. 2015). The peak of the supposed frenchification is usually situated in the seventeenth and eighteenth centuries. A recent exploration of ego-documents such as diaries and autobiographies reveals an increase in the use of French in the eighteenth century, while Dutch is the dominant language choice overall (Krogull et al. 2023: 245–250). The increase is most visible in the language choice of female writers, in line with previous assumptions about French being a women’s language in the first place in this period (van Strien-Chardonneau 2018). In the present paper, we zoom in on another type of ego-document, viz. private letters.

3 Data and methodology

In order to study Dutch-French language choice in private correspondence, we collected manuscript letters from Dutch family archives across the Netherlands. These letters were digitised in the form of photographs or scans but not transcribed in the sense of a searchable corpus, which is why we use the term dataset instead. Diachronically, the dataset covers the period 1600–1899, thus containing letters from three consecutive centuries (see Table 1 at the end of this section for the distribution of data). Aiming for a wide geographical coverage of the Netherlands, we selected twelve cities, all of which represent urban centres, also historically, in their respective provinces (Lourens and Lucassen 1997): Amsterdam (North Holland), Arnhem (Guelders), Den Bosch (North Brabant), Groningen (Groningen), Haarlem (North Holland), Leeuwarden (Friesland), Leiden (South Holland), Maastricht (Limburg), Middelburg (Zeeland), The Hague (South Holland), Utrecht (Utrecht), and Zwolle (Overijssel). We strove for a balanced representation of texts from all these cities across our dataset (see e.g. Puttaert et al. 2022: 266 for the nineteenth-century data).

The socioeconomical and demographic profiles of families living in Dutch cities are fairly homogeneous, representing the educated and cultured strata of society. The upper-middle and upper classes are typically associated with Francophonie in the Low Countries, “where knowledge of more than one language makes choice possible” (Offord 2020: 14). In addition to categorical monolingual choices, e.g. Dutch or French, such communicative situations also make multilingual practices possible. In principle, we selected those families which fit the profile of the urban elites, irrespective of their linguistic profile. This means that particularly multilingual family archives were never prioritised over families with documents primarily written in Dutch.^[3] Within the selected family archives, however, we prioritised private correspondence among the immediate family, i.e. between parents and children, siblings, and spouses. More distant relations of the extended family, such as grandparents, aunts and uncles, were only considered when data from the immediate family proved limited.

Although the collected letters were not transcribed in the sense of an electronically searchable corpus, we still compiled searchable inventories for each family archive. These databases contain the same set of metadata, such as information on the archival location, the name of the families and their associated city. They also contain information on the letters, including the date and place of writing, and the place of receiving. These metadata were then operationalised into variables, both temporal and spatial. With regard to individual writers, basic biographical data were added, such as gender, and the date and place of both birth and death. Taking into account the specific communicative setting of letter writing, we also incorporated metadata on the relationship between senders and addressees. These sociolinguistic variables comprise familial relationships (e.g. child–parent, parent–child, siblings, spouses), gender constellations (e.g. male–male, male–female, female–male), as well as the number of senders and addressees (e.g. single–single, single–multiple, multiple–single). Whenever known, we added the addressees’ names to the database.

The fundamental variable is language (or code) choice, which was manually assigned to each letter in the database. In the case of Dutch-French multilingualism, language choice in private correspondence is best viewed as a continuum of linguistic practices without clear-cut boundaries, although we operationalise these practices as categories for practical reasons, as illustrated in Figure 1. Five different choices seem a sufficiently fine-grained yet still workable categorisation for our linguistic analyses.

Figure 1:

Continuum of linguistic practices and choice categories

The continuum has two idealised monolingual poles for each of the two languages, which we refer to as the two monolingual categories, viz. (i) ‘Dutch’ and (v) ‘French’. It needs to be emphasised, though, that the actual texts assigned to these categories may not be entirely monolingual. We do, for example, allow for loanwords (Assendelft et al. 2023; Rutten and Assendelft, this issue), single-word and even short multi-word switches functioning as lexical units or conventionalised expressions, as shown in (1)–(6).^[4] These single- or multi-word switches may also be dates, proper names (e.g. of places or people), or titles^[5] (e.g. of books or theatre plays), as in (5) and (6). Other-language elements are sometimes ‘flagged’ visually by writers, mostly by underlining (see (3), (4), (6)).

1. En het is uyt die hoofde, dat ik tegenwoordig zeer en peine ben, (begrypende dat myne affaires zeer incertain zyn).

   ‘And it is for that reason that I am much in pain these days (understanding that my affairs are very uncertain).’

2. Mijne huidige Standplaats is zeer lekker niet te warm en niet te Koud, Juste milieu.

   ‘My current location is very nice, not too warm and not too cold, happy medium.’

3. j’ai vraiment eu peur en voyant myn voorgeevel en profil moi qui ne l’avois jamais vu qu’en face ce voorgeevel qui ne me plaisait pas infiniment.

   ‘I was really scared when I saw my [big] nose^[6] in profile, and I had only ever seen that [big] nose from the front, which I really didn’t like.’

4. Pierre commencoit a avoir un commencement d’engelsche ziekte je lui ai aussi tot fait mediciner et il est bien mieux

   ‘Pierre was starting to have a beginning of the English disease [rickets], so I had him medicated immediately and he is much better.’

5. Ook had grootma graag dat je La vie privée des animaux mee bragt om uit te teekenen

   ‘Also, grandma would like you to bring The Private Life of Animals to draw from’

6. Er zal komedie gespeeld worden le Medecin malgré lui van Molière, en natuurlijk gedanst worden.

   ‘A comedy will be played The Doctor in Spite of Himself by Molière, and there will of course be some dancing.’

The two categories located closer to the middle of the continuum in Figure 1, i.e. (ii) and (iv), still have a clearly dominant main language.^[7] By that, we first and foremost mean the prevalence of a language at the most visible level (i.e. the lexical rather than, for instance, the syntactic level), which seems a suitable conceptualisation of ‘main’ (and ‘non-main’) languages for our specific Dutch-French constellation (Assendelft 2023). Crucially, the texts in these categories also contain a fair amount of text in the other language, primarily in the form of multi-word switches (as a rule of thumb, at least three lines of other-language text, either consecutive or spread out). In the case of category (ii) ‘Dutch/French’, Dutch is the main language, with sizeable instances of code-switching into French, as shown in (7)–(9). The main language in category (iv) ‘French/Dutch’ is French, but with considerable switches into Dutch, as in (10)–(12). Multi-word switches of this type do not necessarily have to be creative, but could also be proverbs or reported speech, see (9), (11), (12).

1. Zy is niet geestig, maar heeft gezond verstand, doch niet ontwikkeld. C’est une education morale à faire et votre mere ne vous a pas – je le crains –, fait un cadeau; maar dat is myne schuld niet.

   ‘She is not witty, but has common sense, but not developed. There is some moral education to be done and your mother has not – I’m afraid – done you a favour; but that is not my fault.’

2. Is de slaapkamer ook wat vochtig. Zoudt gij ook eens ruilen met Gideon? Die stevige slotenspringer zal er beter tegen kunnen dan mijn zwakke vrouwtje. Ce n’est qu’un idée, mais ne la rejetez pas sans y réfléchir.

   ‘Is the bedroom also a bit damp. Would you also swap with Gideon? That sturdy ditch-jumper will handle it better than my weak wifey. It’s just an idea, but don’t dismiss it without thinking about it.’

3. Ik zal maar zeggen zoo als Martine de meid in les femmes savantes “quand on se fait entendre, on parle toujours bien, Et tous les biaux dictons ne servent pas de rien.”

   ‘I will just say like Martine the maid in The Learned Ladies “when we are understood, we always speak well, and then all fine diction serves no purpose.”’

4. la mélancholie est grand, het heimwee nog erger, enunmot ce bon couple ne jouit pas, ils voyagent pitoyeusement, et comme toujours, de thuis blyvende lagchen hun uit.

   ‘Melancholy is great, homesickness even worse, in short, this good couple is not enjoying it, they travel pitifully, and as always, those staying at home laugh at them.’

5. je voudrois que tout est que rester sur l’ancien pied, mais cela ne se pouvant pas, je m’en consolle, et comme dit fort bien le Proverbe, die zijn Gad brand moet op de Blaare zitten.

   ‘I would like everything to remain as before, but that is not possible, I console myself, and as the proverb says so well, he who burns his butt must pay the price.’

6. il nous a loue, mais avec menagemend en disant dat is allegaar goed en wel als ‘t maar niet is dat nieuwe beezems goed veegen, j’espere que non.

   ‘He praised us, but with circumspection, saying that is all good and well as long as it’s not that new brooms sweep well, I hope not.’

Right in the middle of the continuum sits category (iii) that we refer to as ‘50/50’. It represents the most multilingual texts in our dataset. ‘50/50’ was applied whenever one single main language could not be identified. This means that both Dutch and French occur to more or less the same extent lexically, as constant (inter- or intrasentential) alternations throughout the text, or longer but more separated parts. Irrespective of its type, the dominant language cannot be established with certainty, which is why we treat ‘50/50’ as a category in its own right. We return to the different types of ‘50/50’ in Section 5. We should add that the continuum in Figure 1 is still a simplistic representation of the multilingual situation in the Netherlands, as it disregards contact with languages or varieties other than Dutch and its main contact language French. It seems an appropriate compromise for the specific purpose of this study, though.

In order to make a balanced, representative selection of all private letters collected from the family archives, a set of well-defined criteria had to be applied. This is a crucial step due to the uneven distribution of archival data at three different levels, viz. families, writers, and sender-addressee relationships. At the level of family archives, we defined a maximum of forty letters per family, drawing on all inventoried letters. At the level of individual writers, we aimed to include at least five different writers from the same family, ideally also balanced across genders and generations. Finally, at the level of sender-addressee relationships, no more than four different addressees per sender were selected, and no more than three letters per sender to the same addressee. The final dataset thus never contains more than twelve texts by the same individual.

As part of this procedure, language choice was represented at the level of unique sender-addressee relationships. The importance of role relations in (potentially multilingual) family communication was already noted by Fishman (1965: 76), who pointed out that “interacting members of a family […] are hearers as well as speakers (i.e., that there may be a distinction between multilingual comprehension and multilingual production)”, and “their language behaviour may be more than merely a matter of individual preference or facility but also a matter of role-relations”. Without going into too much detail here, our selection procedure takes into account three prototypical constellations of language choice (see Puttaert et al. 2022 for a comprehensive account of the methodological framework). First of all, if all inventoried letters within a sender-addressee relationships had the same language choice, for instance ‘Dutch’, then three texts of this category were selected. Secondly, if two linguistic categories were used within the same sender-addressee relationship, one of which occurred more frequently than the other, then two texts in the prevalent option and one in the minor option were selected. Thirdly, if more than two categories were attested within the same sender-addressee relationship, then one text of each category was selected, irrespective of their proportions.

Table 1 shows the final dataset of private correspondence used in this paper, spanning the seventeenth to the nineteenth centuries.^[8] This dataset comprises letters from 56 families, 566 individual writers, and 819 unique sender-addressee relationships. In total we selected 1,827 letters based on the criteria described above.

4 Results (I): Quantitative analysis

We give a general quantitative description of the dataset before we zoom in on multilingual practices. Our dataset comprises 1,827 letters. The distribution of these letters over the five categories of language choice discussed in Section 3 above is as follows:

1. ‘Dutch’: 1,289 letters (70.6 %),

2. ‘Dutch/French’: 52 letters (2.8 %),

3. ‘50/50’: 24 letters (1.3 %),

4. ‘French/Dutch’: 62 letters (3.4 %), and

5. ‘French’: 400 letters (21.9 %).

Dutch is clearly the dominant language choice. Together, the monolingual categories (i) and (v) make up 92.5 %, while the three multilingual choices (ii), (iii) and (iv) combined constitute 7.5 %. In Figure 2, we present the diachronic development of the two monolingual options separately and the multilingual categories together. Figure 2 demonstrates that the three multilingual categories together are always below 10 % across the entire timespan. Category (i) ‘Dutch’ is most often between 70 and 80 %, except for 1800–1849, when it represents 66.6 %, and the period 1750–1799, when it drops to 56.9 %. French, on the contrary, peaks in this particular period (i.e. 1750–1799) with 36.6 %. This peak echoes the traditional idea of a frenchified eighteenth century (Krogull et al. 2023).

Figure 2:

Dutch (i), French (v) and multilingual categories (ii, iii, iv) across time

As we are interested in multilingual practices rather than in differences between the monolingual categories, we now group together the two monolingual options (i) and (v), comparing them to the multilingual categories (ii), (iii) and (iv). Leaving possible regional effects aside (see however Puttaert et al. 2022: 261–263), we concentrate on social variables (gender and familial relationships) in what follows.

Table 2 gives the absolute numbers and the proportions of monolingual and multilingual letters across gender and across gender constellations. It shows that both men and women produce mostly monolingual letters. There does not appear to be a gender difference. There is also no difference between gender constellations: whether men write to men or to women, and whether women write to men or to women does not influence the distribution of monolingual versus multilingual communication. This distribution is always approximately 90 % versus 10 %, which is in line with the diachronic results presented in Figure 2. We therefore do not have any reason to assume gender differences with respect to the use of multilingual practices in private communication.

Our dataset also permits an analysis in terms of familial relationships. In Puttaert et al. (2022: 271–272), we suggested a difference between parents and children with respect to the use of French when compared to Dutch as children seemed to use more French than parents in nineteenth-century private letters. With respect to monolingual versus multilingual communication, however, there do not appear to be any differences between children and parents (Table 3). In both cases, the same 90 % versus 10 % distribution is found, as is the case for letters exchanged between siblings and spouses.

Our results so far show an increase in the use of French, particularly in the period 1750–1799, but when comparing monolingual discourse, i.e. categories (i) and (v), to multilingual discourse, i.e. categories (ii), (iii) and (iv), the social factors built into the database do not seem to influence the results. Neither gender and gender constellations nor the type of relationship affects the rough overall distribution of 90 % monolingual and 10 % multilingual letters over the different periods of time.

Within the three multilingual categories, the central one, category (iii), represents the most multilingual repertoire, in that it is impossible to decide which language dominates the discourse. This category is the most intriguing one from the perspective of multilingual practices. In our dataset, there are 24 letters with this ‘50/50’ pattern (1.3 % of the total number of letters). Most of these, viz. 22 letters, are from the nineteenth century. There are two examples from the seventeenth century and none from the eighteenth. It may seem remarkable that the century with the highest proportion of French monolingual letters, i.e. category (v), lacks examples of this mixed ‘50/50’ category, but this imbalance is most likely an effect of the structure of the dataset: 72.7 % of the letters in our dataset are from the nineteenth century (see Table 1). The 22 examples from the nineteenth century are roughly equally distributed over the periods 1800–1849 and 1850–1899: eight letters are from the first half of the century, twelve from the second, and another two are not dated, although we know they are nineteenth-century letters on the basis of biographical and other contextual information.

In this subset of texts, too, gender and gender constellations do not seem to influence the distribution at the macro level: there are fifteen category (iii) letters written by men (1.9 % of all the letters written by men) and seven by women (1.3 % of all the letters written by women), and the addressees are mostly men (twelve times), but also women (nine times) or, in the case of one letter sent to the writer’s parents, a man and a woman. Similarly, looking into social relationships, the proportion of ‘50/50’ letters is always 1–3 % between children and parents, siblings and spouses, representing a handful of letters in each case. Finally, we also looked into the regional distribution, which confirms the impression of a general phenomenon occurring everywhere to some extent. Letters in the ‘50/50’ category are found in ten out of twelve cities represented in the database (Amsterdam, Arnhem, Den Bosch, Leeuwarden, Leiden, Middelburg, Leeuwarden, The Hague, Utrecht, and Zwolle), and with letter writers linked to thirteen different families.

After this general, quantitative description of the dataset along the major social factors built into the dataset, we move to a qualitative analysis of multilingual letters, focusing on the ‘50/50’ letters.

5 Results (II): Qualitative analysis

We zoom in on private family letters with ‘50/50’ as the assigned language choice, for which we continue to use the 22 texts from the nineteenth-century dataset. We briefly recall here that the ‘50/50’ category was applied whenever it was impossible to clearly determine only one main language that dominates the overall language choice of the text (see Section 3). While Dutch and French are roughly used to the same extent in these cases, ‘50/50’ is no homogeneous category.

On closer inspection of the letters under scrutiny, different types of ‘50/50’ can be identified. We refer to the two major prototypes as Type 1 and Type 2, respectively. In Type 1, long passages written in Dutch and French are neatly separated from each other halfway through the text with only one or very few instances of code-switching from one language to the other. Type 2 cases are characterised by multiple instances of code-switching between Dutch and French throughout the text, both inter- and intrasentential. Naturally, there are also cases which fall in-between these prototypes, not as clearly separated as Type 1 and not to the degree of hybridity as Type 2.

One of the prototypical examples of Type 1 is a private letter dated 1851, written by Walrave Boissevain (1833–1854), descendant of a wealthy patrician family from Amsterdam, to his father Gideon Jeremie in Paris. The first half of this letter is written in French, but continued in Dutch a few days later, i.e. clearly separated (also temporally). In a similar vein, a letter by Anna Odilia Bieruma Oosting (1844–1862), nine years old at the time of writing, to her brother Johannes consists of two parts in two languages. The first part in French, introduced with the salutation “Cher Jean” (i.e. Dear John), is followed by a second part in Dutch, with “Lieve Johan” (i.e. Dear John). French and Dutch are thus used as distinct, in a way monolingual choices within the same multilingual document.

Intriguingly, some Type 1 texts contain metalinguistic comments by the letter writers. Example (13) is taken from a letter written by Agatha Victoria Brantsma (1817–1897), born in the Frisian capital of Leeuwarden, to her husband Jan Bieruma Oosting. Her letter from 1855 is marked by one intersentential instance of code-switching from French (first two pages) to Dutch (last two pages). However, it is only on the final page that she notices her switch to Dutch and comments on this, describing it as a ‘foolishness’.

1. De gekheid waarom ik in eens in het Hollandsch ben begonnen bemerk ik nu pas. Maar dat is vast gekomen dat ik tusschen beiden die Hollandsche brief van Ma kreeg.

   ‘The foolishness of starting in Dutch is something that I only realise now. But that must have been because I received that Dutch letter from Mum in between.’

This strongly indicates that her switch from French to “Hollandsch” was unconscious and perhaps unwanted, though triggered by an incoming letter in Dutch from her mother. The unconsciousness of the switch signals even more the writer’s multilingual repertoire as well as the various possible code choices within this sender-addressee relationship. In fact, other letters from Agatha to Jan in our dataset are in ‘Dutch’, ‘Dutch/French’, and ‘French’.

Generally, multilingual practices are a recurring phenomenon in the correspondence of the Frisian Cats and Bieruma Oosting families. In example (14), dated 1842, Jan Bieruma Oosting (1816–1885), the addressee of the above-cited letter, writes to his mother Sjucke Cats. Again, the first part of the letter is in French, with an intrasentential switch to Dutch near the middle of the text. Unlike his wife’s switch in (13), this one is conscious and explicitly motivated by a topic change.

1. A présent un instant je dois vous parler d’affaires en duidelykheidshalve zal ik dit in ’t Hollandsch doen.

   ‘Now I must talk to you for a moment about some matters and for the sake of clarity I will do this in Dutch.’

A different metalinguistic comment, shown in (15), is found in a letter dated 1827 kept in the Beucker Andreae archives, a notable family of lawyers from Leeuwarden. The 15-year-old Johan Henrik (1811–1865) writes to his father Daniël Hermannus, starting his letter in French, explicitly framing his choice as a language-learning exercise, not uncommon for boys of his age and social background. Towards the end of the first page, however, Johan Henrik decides to switch to their “vaderlandse taal” (i.e. Dutch), apparently motivated by his laborious struggle to write in French.

1. A Présent je veux écrire une fois à vous une lettre en francais. Je sais bien que vous en trouverez une quantite de fautes, mais je ferai mon possible, que j’aurai du moins mon touts de fautes d’ecrivain. […] Ha Ha daar heb 1/2 uur op gezeten. maar nu zal ik onze vaderlandse taal gebruiken

   ‘Now I want to write you [once] a letter in French. I know you will find a lot of mistakes in it, but I will do my best to at least have my share of writer’s mistakes. […] Ha ha, I have spent half an hour on that. But now I will use our patriotic language.’

Concentrating on Type 2 cases, we can provide evidence of private letters in which writers move fluently from one language to another, in this case from Dutch to French, and vice versa. A case in point comes from the Brantsen archive, an important regent family from Arnhem in the eastern province of Guelders. In a letter dated 1860, Karel Maria Brantsen (1834–1909) writes to his brother Theodoor Leopold, in the French harbour city of Toulon.

1. Mon cher Theodore! Bien merci pour votre aimable lettre par laquelle vous me felicitez du nouvel emploi qui m’est tombé en partage. Vous direz que j’ai tardé bien longtemps à vous répondre mais vous comprendrez j’espère que mes occupations avant mon départ ont été fort nombreuses. Je suis donc ici depuis mercredi passé le 17 du courant quand a eu loué mon installation zooals gij begrypt met zeer veel deftigheid en staatsie. Pour le moment je suis encore logé à Bingerden mais la semaine prochaine je vais prendre mon intrek à l’hôtel “Het wapen van Bingerden” où j’ai loué une chambre qui fonctionnera comme salon, salle à diner, chambre à coucher, etc etc. Ik heb my daar besteed voor 25 stuivers per dag behalve vuur en licht. Gij begrypt dus dat daarbij niet veel vet in de pot kan komen en ik wel mager zal worden. In het voorjaar denk ik een kasteel te bouwen echter niet om er mijn intrek te nemen met vrouw en kinders. Mon depart du Zijp m’a coûté bien de la peine car vous me connaissez le grand faible que j’avais pour cette cherie terre natale mais je comprenais qu’on m’en chasserait pourtant tôt ou tard et mieux vaut partir libre que d’être chassé. Ik zal dus nu maar moeten zien tant bien que mal myn eigen huishouden op te rigten. Au Zyp pour le reste blijft alles bij het oude: […]

   ‘My dear Theodore! Thank you very much for your kind letter congratulating me on my new job that I have been offered. You will say that it has taken me a long time to reply, but I hope you will understand that I was very busy before my departure. I have been here since last Wednesday the 17th, when I rented my residence, as you understand with very much refinement and stature. For the moment I am still staying in Bingerden but next week I will take my residence in hotel “The arms of Bingerden” where I have rented a room that will function as living room, dining room, bedroom, etc. etc. I have taken up residence for 25 stuivers there, fire and light not included. You understand that I will not save a lot of money and I will become skinny. In spring I think I will build a castle, but not to live there with my wife and children. My departure from Zijp cost me a lot of pain because you know the great weakness I had for this dear homeland, but I understand that I would be chased away from it sooner or later and it is better to leave freely than to be chased away. I will thus have to see now that I set up, as best I can, my own household. At Zijp, for the rest, everything stays the same.’

This fragment displays various phenomena, single-word switches and onomastic elements, but mostly multi-word switches, both at the inter- and intersentential level. The effortlessly ‘mixed’ use of the writer’s linguistic resources on the first page is followed by a longer passage in Dutch, before resuming a multilingual Dutch-French code on the final page. It is worth mentioning that multilingual practices are found across the Brantsen family correspondence. For example, Karel’s brother Theodoor’s letters to their mother Jacoba Charlotte Juliana van Heeckeren van Kell are written in ‘French’, ‘French/Dutch’, ‘Dutch/French’, and ‘50/50’. We find the same four categories also in mother Jacoba’s letters to her son Theodoor.

The next example comes from the archive of the well-to-do Schorer family from Middelburg in the southwestern province of Zeeland. This letter from 1834 is written by Johan Cornelis Schorer (1801–1856), during one of his many European travels, to his sister Jacoba Elisabeth.

1. Je suis arrivé ici, ma chère Lise, lundi matin par un tems abominable, aujourd hui le tems est magnifique et c’est une vraie réjouissance : je me trouve ici seul et passablement logé et extrêment isolé, j’ai mes chevaux avec moi, dans une huitaine Bogaard viendra ici à la même ferme où je me trouve. Ici on doit se pourvoir de tout, et ce genre de ménage me commence à m’ennuyer excessivement, je profite donc de votre offre pour m’envoyer qq^e chose. Mais qq^e choice de solide p. e. eene groote gebradene Osse rib, waaraan ik het meeste behoefte heb, bruine boonen zoo er die nog zijn, groente is hier niet besteed, daar onze kok zeer matig kookt, ingelijde paling of zoo iets voor ’s avonds, gerookte tong, den Baas heeft nog wel van die Steen mandjes, want het moet gedragen kunnen worden, en ook tegen het rijden kunnen, enfin vous arrangerez cela (aimmablement, je vous demande sans réserve), puisque je sais que vous aimez à m’obliger. Addresser het aan mijn adres in den Grooten Boom maar zet er bij aan den Heer Leut adjudant Wagner te Schoondijken die het verder aan mij zal laten bezorgen. Je me rejouis pour vous que Meier vdam est encore chez vous, il vous restera qu’une huitaine que vous serez seule […] je reçus ce matin le paquet de Melan ([illegible word], ayez la bonté de m’envoyer aussi billet van afvaart van de Schelde). en nog één paar Onderhemden die goed zijn, want ik heb er hier die ik niet meer kan aandoen, geheel aan stukken.

   ‘I arrived here, my dear Lise, on Monday morning in abominable weather, today the weather is magnificent and it is a real pleasure: I find myself here alone and fairly well housed and extremely isolated, I have my horses with me, in about a week Bogaard will come here to the same farm where I am. Here we have to provide for everything, and this kind of household is beginning to bore me excessively, so I am taking advantage of your offer to send me something. But something solid, e.g. a large roasted ox rib, which I need most, brown beans, if there are any left, vegetables are out of the question, as our cook cooks very moderately, pickled eel or something like that for the evening, smoked sole, the boss has some of those stone baskets, because it must carriable, and also be able to withstand riding. At last, you will arrange it (kindly, I ask you unreservedly), since I know you like to oblige me. Send it to my address in the Grooten Boom but add to Mr Lieut[enant] adjutant Wagner at Schoondijken who will have it delivered to me. I’m delighted for you that Meier v[an] Dam is still with you, so you’ll only have a week or so left to yourself. This morning, I received a package from Melan ([illegible word], be so kind as to also send me a ticket of departure from the Scheldt), and one more pair of vests that are good, because I have some here that I can’t wear anymore, entirely to pieces.’

Like in (16), code-switching occurs at the inter- and intrasentential level. Our dataset comprises another ‘50/50’ letter of the siblings, dated 1839, but written by Jacoba Elisabeth to Johan Cornelis, who was in Paris at that time. Although she uses Dutch and French roughly to the same extent, the linguistic resources in her letter are more separated (i.e. Type 1), unlike the more hybrid nature of her brother’s writing.

The final example of ‘50/50’ letters in their most hybrid form can be found in the archive of the notable De Constant Rebecque family from the court city of The Hague. In a letter dated 1853, Charles Theodore Jean De Constant Rebecque (1805–1870) employs multilingual practices when writing to his wife Juliana Frederica d’Ablaing van Giessenburg.

1. Mercredi j’ai eu mon diner, il y avait M^r et M^d d’Escury, M^r et M^d Teylingen Jet et Henriette, j’avais fait venir Naatje pour ce jour, et commande quelques plats chez le Cuisinier Lambert, une majonnaise de hommards, un vol au vent, aspic et des petits pates aux huit[r]es. le diner a très bien réussi, Jet was er van uit, alles zoo netjes en goed en zonder drukte, die stumpert zit er in, zy moet een van die dagen de Minister te eeten hebben, en weet in het geheel niet wat ze doen moet. J’avais prie Jet de faire les honneurs du salon, au grand bonheur de [illegible name] qui avoit une peur terrible que Henriette s’en mélat, j’ai averti cette derniére que Jet s’etait chassé du commandement demaniére que tout s’est trés bien passé. à 8 heures tout était denouveau en ordre et rien de cassé. les dames sont restées boire le thé, et Doly a lu une partie du journal de témoin. Jeudi heb ik je broer Willem met Willem Boreel ten eeten gehad, om de klikken te helpen opeeten. Apropos hebben wy één of twée zilveren kleine pianoblakertjes? heb je er een mede?

   ‘On Wednesday I had my dinner, Mr and Mrs d’Escury were there, Mr and Mrs Teylingen, Jet and Henriette, I had Naatje come for the day, and ordered some dishes from the cook Lambert, a lobster mayonnaise, a vol-au-vent, aspic and small patties with oysters. The dinner went very well, Jet was out of it, everything so neat and good and no fuss, that dullard is involved, she must invite the Minister over for dinner one of these days, and doesn’t know at all what to do. I had asked Jet to do the honours in the drawing room, much to the delight of [illegible name], who was terribly afraid that Henriette would get mixed up in it. I warned her that Jet had left the command so that everything went very well. At 8 o’clock everything was back in order and nothing was broken. The ladies stayed to drink tea, and Doly read part of the witness’s diary. On Thursday I had your brother Willem with Willem Boreel for dinner, to help eat the leftovers. Apropos, do we have one or two little silver piano sconces? Do you have one with you?’

Interestingly, no fewer than five ‘50/50’ texts from the De Constant Rebecque family are included in our dataset, written by three different members, all men. Two letters are Type 2, with one additional text showing characteristics of Type 2. If we take into account the representative selection of the family’s letters in the dataset, we can establish that multilingualism, including multilingual practices, was ubiquitous among the De Constant Rebecque family. While 60 % of the individual letters is monolingual (three quarters in French, one quarter in Dutch), a remarkable 40 % is multilingual (i.e. categories (ii), (iii), (iv) combined), which is far above average in family correspondence (see Section 3). This indicates that multilingual practices were not just individual code choices, but potentially part of a social communicative practice within the family – at least during the nineteenth century, and at least in written communication.

Now that we have looked at our ‘50/50’ category from a qualitative level, it is worth reconsidering the sociolinguistic variable of gender and more specifically the gender of letter writers. Our general (quantitative) results in Section 4 did not reveal any gender differences as regards monolingual versus multilingual practices. They also suggest that multilingual ‘50/50’ choices for private correspondence are made by both male and female writers. However, the more fine-grained (qualitative) distinction between types of ‘50/50’ allows us to add some important nuances.

Table 4 shows that out of the 22 letters under investigation, all six Type 2 cases, in which writers switch back and forth between Dutch and French, are produced by men. Two in-between cases with certain characteristics of Type 2 are also by male writers. Six other texts by men can be categorised as Type 1, more neatly separating their linguistic resources.

In contrast, out of all seven ‘50/50’ letters written by women, six unambiguously represent Type 1, and one more can be located somewhere between Types 1 and 2. None of the ‘50/50’ texts by female family members shows the same level of hybridity as the Type 2 texts produced by male members. Although we shy away from making generalising claims based on the limited data sample, this may mean that women, while using both Dutch and French in their private letters, tend to keep their multilingual resources separated from each other, whereas men also ‘mix’ them in a more hybrid way. This potential gender difference is a topic for further research. No patterns can be signalled with regard to the gender of addressees. ‘50/50’ texts of both types have male and female family members as addressees. For Type 2, four addressees are women, two are men. This also holds true for different familial relationships: the multilingual practices of Type 2 can be identified in letters between siblings (three letters), children and parents (two letters), and spouses (one letter).

6 Discussion and conclusion

Multilingual practices have been the main focus in this paper. In order to sufficiently contextualise historically attested multilingual practices in Dutch-French family correspondence, we have first offered a quantitative description of our dataset along the social factors built into it. Previous research within the same project as the present research has shown that various social factors influence language choice, such as gender and gender constellations, familial relationships, and regional background (Puttaert et al. 2022). Whereas the present research repeats the increased importance of French in the eighteenth century (Krogull et al. 2023), we could not establish an influence of gender and gender constellations, familial relationships, and regional background. Overall, we conclude that multilingual practices constitute a low-frequency phenomenon, which is however widespread in that it occurs among men and women, across all gender constellations and social relationships, and in most of the cities represented in the dataset. It is infrequent in all those contexts, but it does occur almost everywhere. For this reason, we believe the multilingual practices studied here, including category (iii) with the ‘50/50’ letters, constitute code choices by themselves and should not be seen as a mere mixture of other, more monolingual choices nor as the remarkable habit of specific individuals. This generalisation depends on the extensive dataset used and its quantitative description above, which allows us to interpret a seemingly marginal phenomenon such as the ‘50/50’ letters as low-frequent, yet widespread.

Against the background of this baseline evidence, we have taken a qualitative view on multilingual practices, zooming in on the ‘50/50’ cases from our nineteenth-century dataset. The sociodemographic profile of writers fully deploying their Dutch-French repertoires in family correspondence can be described as the urban elites, i.e. the educated and cultured, often mobile and well-travelled layers of society, both men and women. On closer inspection, the ‘50/50’ category could be further divided into two prototypes, revealing different strategies of using Dutch alongside French. Multilingual individuals could opt for neatly separating their resources as two distinct codes within the same letter (Type 1), or ‘mix’ Dutch and French into a hybrid code (Type 2), with various gradations in between.

It is at this qualitative level that the gender dimension comes into play again. Though admittedly based on a small amount of such letters (but nonetheless a representative sample of this phenomenon), we have observed that Type 1 is used by both men and women. The hybridity of Type 2, however, is exclusively found in men’s letters. This suggests that male writers made use of switching between Dutch and French in a more fluid manner, whereas female writers appear to have used Dutch and French more as distinct resources, even in multilingual letters.

These tentative, yet still intriguing observations open up additional questions about sociolinguistic gender patterns in relation to code-switching (e.g. Gardner-Chloros 2012; Nurmi and Pahta 2012). If we adopt the view that code-switching or, more generally, multilingual practices are ‘non-standard’ behaviour (Cheshire and Gardner-Chloros 1998: 13–14), we may cautiously interpret the absence of Type 2 letters by women as historical evidence of multilingual female writers tending to avoid linguistic hybridity. At the same time, our results do not confirm that women wrote fewer multilingual letters than men per se, at least in quantitative terms. Neatly separating linguistic resources in letter writing, however, possibly echoes traditional (modern-)sociolinguistic assumptions about gender, “with women code-switching less than men in order to conform with a more purist or socially acceptable speech style” (Cheshire and Gardner-Chloros 1998: 14). Also in the light of our small data sample, this clearly requires further investigation and more empirical evidence, for instance, by taking into account multilingual texts from categories (ii) and (iv), too.

Irrespective of potential gender patterns, we have also shown that multilingual practices were not just individual preferences. Obviously, multilingual letters had to be understood by at least receptively multilingual addressees (and accepted as socially appropriate choices within these sender–addressee relationships). In fact, ‘50/50’ practices were often embedded in highly multilingual family correspondence, with various members making use of their repertoires in letters to various relatives. Multilingual family letters can therefore be interpreted as social communicative practices, highlighting their status as code choices in their own right.

For the study of historical multilingualism this implies that ‘language choice’ and ‘code-switching’ are not necessarily discrete objects of research, but closely intertwined, certainly in the private domain. We may view Type 1 strategies of multilingual writing as co-occurring but ultimately distinct language choices within the same communicative act, which do not affect the (largely) monolingual nature of the codes themselves. Type 2 strategies, in contrast, foreground written code-switching as language or code choices beyond monolingualism in the sense of either-or choices. Both strategies constitute multilingual writing practices, attested at the level of individuals and families, within the same genre, domain, and (historical-)sociolinguistic space.

The combination of quantitative baseline data and qualitative analyses has proven to be crucial for uncovering such phenomena and interpreting underlying sociolinguistic factors (or the lack of clear patterns). This combined approach has also allowed us to arrive at a fine-grained account of the linguistic resources, practices, and choices of users. As shown, monolingual and multilingual practices in reality form a continuum of practices rather than clear-cut choices, also in written text sources from the past. Our suggested model in Figure 1 (Section 3) appears to be a useful conceptualisation of the letter writing practices under investigation. The idea of a continuum should remind researchers in historical sociolinguistics, including ourselves, that the categories we create may be practical for analytical purposes, but can be fuzzy and heterogeneous, too, and therefore need to be acknowledged as such.