English from Scratch: Preadolescents’ Developing Use of English Lexical Resources in Belgian Dutch

Schuring, Melissa; Zenner, Eline

doi:10.3389/fcomm.2021.788768

ORIGINAL RESEARCH article

Front. Commun., 24 February 2022

Sec. Psychology of Language

Volume 6 - 2021 | https://doi.org/10.3389/fcomm.2021.788768

This article is part of the Research Topic Englishes in a Globalized World: Exploring Contact Effects on Other Languages View all 17 articles

English from Scratch: Preadolescents’ Developing Use of English Lexical Resources in Belgian Dutch

Melissa Schuring*

Eline Zenner

QLVL, Faculty of Arts, KU Leuven, Brussels, Belgium

Working within the framework of the socio-pragmatic turn in anglicism research, this paper adds a developmental sociolinguistic perspective in investigating preadolescents’ use of English lexical resources in Belgian Dutch. The so far largely undocumented role of English in the linguistic transition from childhood to adolescence is analyzed through a fieldwork corpus of 15,465 utterances, collected during sociolinguistic interviews with 26 (12 boys, 14 girls) Belgian Dutch preadolescent (6–13 years/o) respondents from a local hockey club. All English lexical material in the corpus was identified and categorized following a three-step identification protocol. This protocol introduces a distinction between recognizable unavoidable English (RUE) and recognizable avoidable English (RAE). Results reveal that, overall, 9.7% of the utterances contain recognizable English (RUE + RAE), with RUE being significantly more frequent than RAE. Our findings further indicate only limited stratification according to traditional socio-demographic parameters and display a number of outliers in the respondent profiles. Closer inspection of these outliers allows the conclusion that in the community of practice studied, English is an emerging youth language marker, typically used when talking about gaming or girl-oriented activities. In sum, we conclude that preadolescents in our sample instrumentalize English for incipient identity work, both on the micro-level (being a gamer, a soon-to-be teenage girl) as on the macro-level (through ingroup and outgroup marking).

1 Background

This paper contributes to the ongoing socio-pragmatic shift in anglicism research by tapping into the current attention for the acquisition of variation in sociolinguistics, sometimes referred to as “developmental sociolinguistics”. Particularly, results are presented from an investigation into the development of English lexical resources in the speech of Belgian Dutch preadolescents. In Section 2, we discuss the creation of our preadolescent corpus, elaborating on the community of practice under scrutiny and the chosen fieldwork method of sociolinguistic interviews. Next, we introduce our protocol to identify and classify English insertions in the corpus. Section 3 presents the results of this paper, the implications of which will be discussed in Section 4. First, we provide the necessary background on anglicism research and developmental sociolinguistics (1.1), followed by an outline of the Dutch-English contact situation in Flanders (1.2).

1.1 Anglicism Research Featuring Developmental Sociolinguistics

In the current climate of ever-increasing globalization (Blommaert, 2010), English continues to diffuse into Europe’s linguistic landscape, stimulating a tradition of anglicism research that focuses on the resulting English borrowings in European domestic languages. Initially, a large part of the studies in the field have charted all possible manifestations or “types” of English and proposed classification strategies and algorithms to answer the question of how English should be defined and counted in receptor language corpora. For instance, Onysko and Winter-Froemel (2011) introduced the labels of “catachrestic” (∼necessary or unavoidable) and “non-catachrestic” (∼non necessary, avoidable) loanwords to account for the absence (∼catachrestic) or presence (∼non-catachrestic) of receptor language alternatives. Other strategies to classify anglicisms involve etymology, adaptation to receptor language morphology or phonology and listedness in dictionaries (Gerritsen et al., 2007; Roberts et al., 2021). Moreover, English loans are usually tagged for part of speech and for the semantic field they belong to (e.g., Yang, 1990). As such, English insertions are shown to be clustered around i.a. IT/gaming, business and sports, thus being labeled as “English-prone” semantic fields (Onysko, 2007; Balteiro, 2018; Hunt, 2019).

In a second wave of anglicism research, the focal point was shifted to the social and pragmatic meanings assigned to these manifestations of English. Vaattovaara and Peterson (2019), for instance, rely on a mixed methods approach to uncover the indexical link between English swearing in Finnish and “urbanity”. A more pragmatic perspective was taken in Andersen (2017) who observed a change in discourse function for the Norwegian anglicism jobb from negative to more neutral or positive. This “socio-pragmatic turn” in anglicism research (cf. Andersen et al., 2017) foregrounds the perspective of the language user in a quest for the how and why of the language user’s choice (not) to include English foreign material in specific contexts, thus achieving a certain social or pragmatic outcome (Onysko and Winter-Froemel, 2011; Peterson and Beers Fägersten, 2018).

For one thing, studies adopting this socio-pragmatic focus reveal how English is a clear youth language marker occurring in game talk, social media interaction and rap- and hip-hop lyrics, making the teenager one of the prototypical English language users (Leppänen, 2007; Pennycook, 2007). However, it is still unclear how and when these teenagers get to the point of using English in the first place: what happens in preadolescence, viz. the transition from childhood to adolescence, remains undocumented. This issue provides the point of departure for this paper and will be tackled against the background of the emerging field of “developmental sociolinguistics”.

Developmental sociolinguistics is an upcoming interdisciplinary framework situated at the intersection of language acquisition and sociolinguistics. The term was first used by Entwisle (1966) when investigating how children learn to identify and employ socially meaningful variation patterns (De Vogelaer et al., 2017). The vast majority of studies in the field provide insight into the acquisition of social meaning, studying the stratification of standard and vernacular use according to children’s socio-demographic profiles. As concerns age, results point to the standard being used more by younger children (Roberts, 1994) with an increase of non-standard variants when growing older (Smith et al., 2007). Next, the findings on gender patterns are largely inconsistent (Smith and Durham, 2019) and can be categorized into three groups, according to their contradictory conclusions (Nardy et al., 2013): (1) girls using more standard variants than boys (Romaine 1984), (2) boys oppositely using more standard variants than girls (Chevrot, 1991; Roberts, 1997); and (3) absence of a gender effect (Chabanal, 2001; Foulkes et al., 2001). Finally, concerning socio-demographic background, scholars observe a tendency of higher standard use for higher social class (Chevrot et al., 2000; Nardy, 2008).

In this paper, we apply the methods and principles of developmental sociolinguistics, usually targeting standard and vernacular variation, to the unresolved questions about English use by (preadolescent) children. An optimal setting for our intents can be found in Flanders, where contact between English and Dutch is fundamentally present and, moreover, well documented.

1.2 Contact Between English and Belgian Dutch

(Belgian) Dutch is one of Belgium’s official national languages, alongside French and German. It is primarily spoken in Flanders, the northern and most populated part of the country. Where Brussels, the capital, has an intensive business, political and also personal-based contact situation between (native) speakers of English, French and Dutch (Mettewie and Janssens, 2006), Flanders has no such level of bi- or multilingual communication: despite expected domain loss in typical areas such as international business and tertiary education, contact with English remains primarily remote and indirect (Rys et al., 2019). It is notably through mass media such as the Internet, pop music and English spoken TV shows and films that contact with English is established (Booij, 2001). Given this indirect nature of the contact situation (see Onysko, 2009), the influence of English on Belgian Dutch is largely limited to the introduction of loanwords and -phrases to the receptor language lexicon.

Recent studies taking a production perspective and targeting Belgian Dutch adults corroborate these findings. A case in point is Zenner and Van De Mieroop (2017) who analyzed the language use of three participants in a Dutch reality TV show. Their use of English is characterized as having a locally emergent and highly dynamic social meaning, amongst others indexing masculinity and brotherhood. Alongside face-to-face conversations, computer-mediated communication too has been the subject of inquiry, with growing attention for the analysis of tweets. For instance, research on the ooit/ever construction (Zenner et al., 2018) and the deconstructionalization of the pimped ride (De Pascale et al., 2022 forthc., this issue) has demonstrated a developing creativity with English on Dutch Twitter.

Turning to English as a youth language marker, primarily the study of De Decker and Vandekerckhove (2012) is worth mentioning. In their analysis of more than 200,000 MSN chat messages, they identified at least one English insertion in 13.3% of the posts. Interestingly, markedly different results are found in Zenner and Van De Mieroop (2021)’s study on parent-child interactions involving dinner table conversations between 16 parents and 18 children aged 1 to 7, complemented by sociolinguistic interviews with the parents. Here, less than 1% of the utterances contain minimally one English word and parents, furthermore, report to have no socialization aim towards English. Between preschoolers’ non-production (in parent-child interactions) and teenagers’ considerable production of English, a transition must take place.

What is more, the results of SLA research in Flanders provide further evidence for the presence of this as of yet undocumented transition. Although English tuition only starts at the second year of secondary school (at age 13), Flemish children already have high receptive English vocabulary knowledge before the age of 12, viz. prior to formal instruction¹ (Peters et al., 2019; Puimège and Peters 2019; De Wilde et al., 2020; De Wilde et al., 2021). The respondents’ level of English proficiency is, however, largely idiosyncratic, depending on contextual factors and being related to the types of English input teenagers have access to (ibid.). Indeed, Bollansée et al. (2021) found a positive correlation between productive word knowledge of English and the frequency of playing video games. Since boys are more regularly engaged in English-themed gaming activities than girls (e.g., Kuppens, 2010), this gaming pattern additionally points to a gender effect in overall English use, with boys having a higher expected English production than girls. Following the aforementioned studies, it is then likely that since children younger than 12, through high extramural exposure, receptively understand a lot of English words, they would also produce them. Whether this assumption holds true is what this study aims to address.

1.3 Research Questions

This paper makes a case for applying a developmental sociolinguistic perspective to socio-pragmatic studies in anglicism research, aiming to uncover when, why and how preadolescents use English lexical resources in Belgian Dutch. Three research questions are addressed:

RQ1. How many English insertions do we find in the language use of Belgian Dutch preadolescents overall, taking into account the type of English used?

Following the aforementioned work of Zenner and Van De Mieroop (2021) involving preschoolers in parent-child interactions (less than 1% of English) and De Decker and Vandekerckhove (2012) targeting adolescents (13.3% of English), we hypothesize to find an intermediary frequency, situated in between those two percentages, for the preadolescent age group. As previous studies have shown varying results for different types of English, we insist on categorizing the English elements found in terms of e.g. “avoidability” (cf. catachrestic and non-catachrestic loans, Onysko and Winter-Froemel, 2011) or listedness in dictionaries (Roberts et al., 2021).

RQ2. To what extent do we find stratification by age and gender in Belgian Dutch preadolescents’ use of English lexical resources in Dutch, taking into account the type of English used?

For the age trajectory, we advance two conflicting hypotheses: either preadolescents evolve in using English gradually with age, as was found for non-standard variants in previous developmental sociolinguistic studies (Roberts, 1994; Smith et al., 2007); or, following the idiosyncratic nature of English vocabulary learning (i.a. Puimège and Peters, 2019; De Wilde et al., 2021), the preadolescent age group presents a high amount of individual variation with no clear age pattern to be identified. Similarly, the impact of gender on the use of English insertions in the studied transition period is difficult to assess, given the conflicting results in earlier work in developmental sociolinguistics (see Section 1.1). We could, nonetheless, cautiously hypothesize to find more English lexical material in boys’ speech because of their prototypical gamer’s image and the earlier demonstrated importance of gaming (i.a. Bollansée et al., 2021, Section 1.2).

RQ3. How can an in-depth analysis of English lexemes in well-targeted individual users help explain the patterns found in RQ1 and RQ2?

A more fine-grained analysis of individual users can possibly throw light on additional parameters, aside from gender and age, steering the insertion of English lexical material in Belgian Dutch. Taking into account the English-prone semantic fields discussed in Section 1.1, the amount of English lexical material can for instance also depend on the topic that is discussed. We therefore expect to see a topic effect and resulting English hotspots for e.g. English-prone gaming and sports.

2 Methodology

The research questions are addressed through sociolinguistic fieldwork in a cohesive community of practice (Lave and Wenger, 1991, see Section 2.1) where we conducted sociolinguistic interviews (Section 2.2). The resulting corpus of 26 hours and 28,998 utterances, 15,465 of which originating from our preadolescent respondents, was transcribed (Section 2.3) and then mined for English insertions for which we created a phased identification protocol (Section 2.4).

2.1 Community of Practice

Respondents were recruited in a hockey club in a middle-sized town in Flanders, Belgium. We included children from all boys’ and girls’ teams with ages ranging between 6 and 13 years old (M age = 9;8², SD = 1;8). The resulting sample consists of 26 preadolescent respondents of whom 12 are boys (M age = 9;2, SD = 1;7) and 14 are girls (M age = 10;2, SD = 1;8; see Table 1, Section 2.3). The respondents are mainly monolingual speakers of Dutch³ with an educational track situated between the first year of primary school and the first year of secondary school. As a result, with two exceptions, these preadolescents have not had any formal instruction of English yet⁴ (cf. Section 1.2). All children provided assent to participate in the study and were given partial disclosure of the research purpose, parents provided informed consent and were given full disclosure⁵.

TABLE 1

TABLE 1. Corpus composition—ordered by respondents’ gender and ascending age.

The reason behind choosing a hockey club to recruit respondents is threefold: firstly, a sports club in general, as opposed to a school, is a more informal and therefore more favorable context to collect production data for a youth language phenomenon. Secondly, hockey players in Flanders traditionally have middle-to upper-class backgrounds which allows us to keep the SES variable in this study relatively stable. Thirdly, the hockey club presents an ideal community of practice that is both cohesive, because children all join hockey practice, and dynamic, since social networks do not completely overlap given that the children attend different schools.

The above-described community of practice was recruited for a larger research project investigating the development of social meaning of English. A total of 114 hours of conversational data was collected in the project, both in individual sessions as in group interactions, amounting to 7 hours of data per respondent. In this paper, we zoom in on a subset of the data collection that consists of individual sociolinguistic interviews.

2.2 Sociolinguistic Interviews

Our goal was to track English insertions through preadolescents’ unmonitored, casual speech. Since it is the very essence of the sociolinguistic interview to elicit this type of vernacular-like speech, we decided to rely on this staple of Labovian sociolinguistic fieldwork methodology. Prototypically, the sociolinguistic interview involves a near-natural one-on-one casual conversation on everyday topics and emotional memories between a researcher and a language user from a local community (Meyerhoff, 2016). In the strict Labovian tradition (Labov, 1984), the sociolinguistic interview consists of multiple components including the actual interview questions, a reading task, a word list and a list of minimal pairs, each targeting the same vernacular variable (Becker, 2013). In the remainder of this paper, we use “sociolinguistic interview” to refer to the looser definition of making one-on-one recordings of only the interview part of the Labovian format.

As for the practical details of the study, respondents individually participated in the sociolinguistic interviews in March and April 2021. The interviews were conducted online⁶ through MS Teams, on a laptop or computer that was set up by a parent in a separate room. All preadolescent children were interviewed by the same 22-year-old Belgian Dutch-speaking female researcher.

For the content and structure of the sociolinguistic interview, we started from the traditional Labovian version (1984) which was then adapted in two ways. On the one hand, we made changes in light of our research goal (identifying English). We designed a semi-structured interview with topic control including i.a. English-prone IT/gaming and sports and more Dutch-prone classroom stories and leisure activities. The choice for these semantic fields was made by combining information from an extensive literature review on English loanwords (cf. Onysko, 2007; Balteiro, 2018; Hunt, 2019) and through a large-scale pretest survey targeting perceptions of and attitudes towards English insertions in Belgian Dutch (see Schuring et al., 2021). On the other hand, we tailored the research method to our “young” target group by (1) including personal narratives of the researcher, as is proposed in the Conversational Map Eliciting Procedure (Peterson and McCabe, 1983); and by (2) reworking and updating traditional sociolinguistic questions (cf. Holmes-Elliott, 2021). As such, we transposed Labov’s near-death experience narrative (1972, 1984) into a more child-friendly gaming context: “have you ever been in a very dangerous situation when you were gaming?” These procedures resulted in a protocol in which the questions gradually become more personal and challenging, starting with social and demographic information, moving on to emotional experiences and finishing with hypotheticals.

2.3 Corpus and Annotation

The resulting corpus consists of 26 one-hour sociolinguistic interviews containing a total of 28,998 utterances. Half of the utterances were produced by the researcher and will be disregarded for further analysis⁷. The remaining 15,465 preadolescent utterances are the core research object for this study. Table 1 provides an overview of the corpus in terms of Child ID and alias, Age (M = 9;8, SD = 1;8), Number of Utterances (M = 595, SD = 113) and Mean Utterance Length (in words per utterance, M = 8.92, SD = 1.38).

The corpus was manually transcribed and annotated following the CHAT conventions of the CHILDES project (MacWhinney, 2000). Excerpt (1) presents an example of what the final transcriptions look like:

(1) ^∗INT: &-eumh en welke muziek luister jij graag?%eng: &-umh and what music do you like to listen to?^∗YAS: ik vin(d) Lizzo wel heel tof.%eng: I really like Lizzo.^∗YAS: Billie Eilish minder want (.) ik vind da(t) zo droevig ofzo.%eng: Billie Eilish not so much because (.) I find that like so sad.^∗YAS: ik word daar nie(t) happy van.%eng: it doesn’t make me happy.

The transcription of an utterance always starts with a three-letter speaker ID preceded by an asterisk (^∗INT for interviewer, ^∗YAS for Yasmine) or with “%eng”, which is used for the English translation of the utterance. In the utterance itself, (…)⁸ and &-euh are used for silent and filled pauses respectively and brackets indicate that (parts of) words are not pronounced as a result of phoneme deletion for round brackets and assimilation for square brackets.

2.4 Identifying English

The identification of English lexical material has been continuously raised as a methodological concern in loanword studies focusing on anglicisms. As mentioned in Section 1.1, a variety of algorithms, strategies and parameters have been created to answer the question of what should count as English. Inspired by Roberts et al. (2021), we synthesize the methods proposed in the literature into a phased identification protocol involving three levels (see Figure 1): etymology, recognizability and relative markedness (Levinson, 2000; Winter-Froemel, 2013). A decision task on each of the levels results in a classification into four borrowing types, ordered on a scale from low Englishness to high Englishness: no English, unrecognizable English (UE), recognizable unavoidable English (RUE) and recognizable avoidable English (RAE).

FIGURE 1

FIGURE 1. 3-step English identification protocol.

As a first step, we need to decide what lexical material from the transcripts is analyzed through the protocol. The point of departure are words or phrases containing at least one free morpheme found in English dictionaries. Hereby, we do not take into account the traditional borrowing-codeswitching dichotomy but instead consider them to be “two outer poles on a continuum” (Zenner and Van De Mieroop, 2021: 8; also see; Backus, 2014). Both English-listed words (cool) as phrases (oh my god) are therefore referred to as “insertions”. Additionally, the fact that only one free morpheme has to be found in the dictionary is a way to account for the possible adaptation to Dutch morphology, especially in the case of verbal inflection for which morphological integration seems to be inevitable (Winter-Froemel, 2008), and for which the result (Dutch gebottleflipt vs. bottle flipped) would not be listed in English dictionaries.

For each of these units identified as candidates for the identification protocol, we then make a first major division on the etymological level: we verify whether an insertion has English origin according to het Etymologisch woordenboek van het Nederlands (Philippa et al., 2018). A “NO” answer on the decision task at this level results in a classification into “No English”. As etymological information can be ambiguous, we choose for an inclusive approach by opting for “YES” when at least one of the suggested etymologies is English.

If an insertion proves to be of English origin, it trickles down in the identification protocol to the recognizability level: an English insertion should be recognizable as English to native speakers of Dutch, and this because “the non-Dutch character of a word can only exert influence on the language user’s behavior when the expression at issue is identifiable as a non-Dutch word” (Geeraerts and Grondelaers, 2000: 56). Ideally, the non-Dutch character would be operationalized by our respondent’s own perception, or by extension by Dutch native speakers’ perception, of what characterizes a non-Dutch insertion. However, this would require a series of perception studies that are beyond the scope of this paper. We hence decided to include two parameters that serve as a proxy for the English character of an insertion. The first parameter concerns grapheme-phoneme mapping (Onysko, 2007) and probes whether the spelling of an insertion leads to a transparent and correct pronunciation in the receptor language, following the grapheme-phoneme mapping rules of that same receptor language. For example, an insertion as challenges (Excerpt 2) would get a “YES” answer on the recognizability level because Dutch pronunciation (/χa’lɛnɣəs/) would sound very different from English pronunciation (/’tʃæləndʒəz/), whereas the insertion sport (/’sport/), see Excerpt (3), gets a “NO” answer since naïve Dutch pronunciation matches English pronunciation. The second parameter that leads to English recognizability and proxies the non-Dutch character is the presence of English chargrams (Andersen, 2005; Andersen, 2012; Zenner and Van De Mieroop, 2021). English chargrams can be defined as “a string of n characters within a certain word that have English etymology” such as word-initial c in cornflakes (Excerpt 4). If both grapheme-phoneme mapping and English chargrams lead to a “NO” answer on this level, the insertion is categorized as unrecognizable English (UE).

(2) ^∗ADA: &-eumh één doet allemaal challenges en de andere maakt slijm.%eng: &-umh one does all of these challenges and the other makes slime.(3) ^∗ELI: &-euh ik kijk nie(t) zo speciaal naar de sport op tv.%eng: &-umh I don’t particularly watch sports on television.(4) ^∗CHA: je kon cornflakes eten croissants enzo.%eng: you could eat cornflakes croissants and so on.(5) ^∗FLO: ma(ar) buiten hockey onthou(d) ik dan ook nie(t) echt veel van mijn weekends.%eng: but aside from hockey I don’t remember much of my weekends.

A recognizable English insertion must finally go through the last step of our protocol which is based on the principle of relative markedness (Levinson, 2000; Winter-Froemel, 2013). This principle corresponds to the labels of “catachrestic (∼necessary) and “non-catachrestic”(∼non necessary) loanwords (Onysko and Winter-Froemel, 2011, cf. Section 1.1) as it accounts for the onomasiological difference between a “marked” insertion because of presence of a Dutch alternative⁹ and an “unmarked” insertion because absence of the same. In our protocol, the existence of Dutch alternatives was verified through three Dutch dictionaries (Van Dale, Het Algemeen Nederlands Woordenboek (ANW) and woorden.org). An insertion was labeled “marked” when at least one of the dictionaries provided a Dutch lexical alternative and, to check for a minimal level of entrenchment (cf. Zenner et al., 2014), when that alternative lexicalization had more than 2,000 hits on Google (Dutch language settings). As a result, words like hockey and weekends (Excerpt 5, no Dutch alternatives) are categorized as non-marked and get labeled recognizable unavoidable English (RUE), whereas words like happy (Excerpt 1, Dutch alternative: blij), challenges (Excerpt 2, Dutch alternative: uitdagingen) and cornflakes (Excerpt 4, Dutch alternative: ontbijtgranen) get a “YES” answer for relative markedness which gives them the label of recognizable avoidable English (RAE). Note that in the case of “proper nouns”, which we define as a string of words that designates a person, place or thing that has been claimed in the physical world, the decision of relative markedness always leads to a “NO” answer as proper names are ipso facto unavoidable insertions.

The above-described identification protocol is applied to all preadolescent utterances in our corpus. In the analysis of the results, we do not take into account unrecognizable English (UE), precisely because of its unrecognizable nature and resulting low Englishness. Instead, in the rest of this paper, we focus on the RUE and RAE borrowing types. All analyses are conducted in R, with the utterance level as a point of departure: for each utterance, we indicate whether or not it includes any instances of RAE or RUE.

3 Results

The results are divided into three sections, corresponding to our three research questions. We first present the overall frequency of English lexical resources in the preadolescent respondents’ speech (Section 3.1, RQ1), followed by an analysis in terms of stratification by age and gender (Section 3.2, RQ2). Finally, we explain the patterns found in Sections 3.1–3.2 by zooming in on three well-targeted individual language users (Section 3.3, RQ3): Kobe (8;11), Max (11;6) and Camille (12;5).

3.1 Overall Frequency of English Lexical Resources

Table 2 presents the overall use of English insertions in our corpus: 9.7% of the utterances contain recognizable English insertions (RAE + RUE). This percentage corresponds to 1,502 utterances out of the corpus total of 15,465. Recognizable unavoidable English (RUE) and recognizable avoidable English (RAE) have a two to one ratio, RUE (6.7%) being significantly more frequent than RAE (3.0%) (results for Wilcoxon rank sum exact test: W = 28, p < 0.001), for which a large effect (WilcoxonR = 0.7) was found.

TABLE 2

TABLE 2. Overall use of English insertions—total number of utterances 15,465.

The corpus contains 1,695 English tokens and 581 English types, resulting in a type/token ratio of 34.3% (same type/token ratio for RAE and RUE individually). The three most frequent types for RAE are match (N=58), team (N=23) and stick (N=19); the most frequent RUE types are hockey (N=127), computer (N=48) and Fortnite (N=24); for further reflection on these individual types see Section 4. It immediately stands out that four of these six frequent insertions are related to sports. This naturally follows from our choice for hockey players as a respondent group and confirms the previous studies’ labeling of sports as an English-prone semantic field (see Section 1.1). A final comment on these numbers relates to the proportion of names in our corpus, such as Fortnite¹⁰ (N=24). As mentioned in Section 2.4, all English-inspired proper names figure in the RUE borrowing type. Overall, 216 types or 54.7% of RUE are proper names. As the example of Fortnite illustrates, the proper names in the corpus mainly refer to objects and concepts such as PlayStation and For Girls Only. We chose to retain these names because, although unavoidable, they are still recognizable as being English. In the next section, we investigate how the English insertions discussed above are distributed across respondents.

3.2 Stratification

Figure 2 shows the relative utterance-based frequency of English insertions per type (RUE and RAE) and per respondent. A table version of the plot can be found in Supplementary Material S1. The y-axis represents the percentage of utterances in the corpus containing (recognizable) English insertions and ranges from 0 to 15%. The x-axis portrays the individual respondents with a code consisting of their alias and corresponding age (in years and months). The youngest respondent is placed on the left of the x-axis, followed by the other respondents in the order of ascending age. Further, the color scheme (both in the graphs as in the x-axis labels) represents the gender parameter¹¹ with light blue coding for boys (B) and dark blue coding for girls (G). Last, the full line represents the relative frequency of RUE and the dashed line portrays the relative frequency of RAE.

FIGURE 2

FIGURE 2. Relative frequency of English per type per respondent (utterance-based). Full line: recognizable unavoidable English (RUE)—Dashed line: recognizable avoidable English (RAE).

Figure 2 reflects four stratification patterns, relating to (1) the RAE/RUE ratio across respondents, (2) age, (3) gender and (4) outliers. The corresponding descriptive results can be found in Table 3 below. We refer to Supplementary Material S2 for an overview of the boxplots.

TABLE 3

TABLE 3. Descriptive results for stratification by Age and Gender in % per utterance– *= significant.

First, the graph resonates the findings of the difference between RUE (M = 6.7%, SD = 1.8%) and RAE (M = 3.0%, SD = 2.1%), reported in Section 3.1. The relative frequency of RUE is higher than the relative frequency of RAE for all respondents (except for Kobe, cf. Section 3.3.1). Moreover, the two graphs mostly follow a parallel trajectory, with respondents who generally use proportionally more RUE, also use more RAE and vice versa.

Second, Figure 2 shows no clear stratification by respondent age, as going from the left on the x-axis (youngest respondents) to the right (oldest respondents), the graphs are roughly stable and show no clear upward trend. A Kruskal Wallis rank sum test based on a categorization into three age groups (6–8 years/o, 9–10 years/o and 11–13 years/o), indeed, shows no stratification by age for RAE (H(2) = 0.483, p > 0.05), RUE (H(2) = 0.087, p > 0.05) nor for RAE + RUE (H(2) = 0.039, p > 0.05).

Third, we discuss gender stratification patterns. Gender does not play a role in the overall English use of our respondents (RUE + RAE, H(1) = 0.024, p > 0.05). Zooming in on the specific borrowing types, gender shows no significance for RUE (H(1) = 1.400, p > 0.05) but becomes a factor in explaining the RAE variation (H(1) = 5.357, p = 0.02, effect size η²= 0.182), with girls (M = 3.1%, SD = 0.8%) using significantly more recognizable avoidable English (RAE) than boys (M = 2.9%, SD = 3.1%). The difference in the means between girls and boys is, however, only 0.2%, indicating that we may not want to overstate the importance of this effect, which also loses its significance when applying Bonferroni correction.¹² In contrast, support for the presence of a gender effect is found when excluding 8-year-old outlier Kobe from the analysis: the gender stratification becomes more apparent as can be seen in the RAE mean for girls (3.1%, SD = 0.8%) and the new REA mean for boys (2.1%, SD = 0.9%; H(1) = 8.104, p = 0.004¹³, effect size η² = 0.296).

Fourth, the reinforcement of the gender pattern upon Kobe’s exclusion already signals the presence and importance of outliers in the sample. The most apparent outlier indeed is Kobe (8;11) who shows the highest relative frequency for both RUE and RAE leading to his profile standing out in Figure 2. Otherwise, we observe the opposite pattern when analyzing Max’s (11;6) numbers: the lowest frequency for RAE and one of the lowest frequencies for RUE. Additionally, we can identify an outlier in the profile of Camille (12;5) who has the highest RAE frequency and second-highest RUE frequency on the girls’ side.

We conclude, answering our second research question, that, in our corpus, RUE is significantly more frequent than RAE for all respondents. At the same time, Figure 2 and Table 3 show fairly limited stratification according to traditional socio-demographic parameters, with age not showing any significance and gender reaching robustly significant differences for RAE when outlier Kobe is excluded from the analysis. Indeed, Figure 2 reveals a number of outliers, individuals just like Kobe, that show a pronounced use of RAE and/or RUE and deserve closer scrutiny in the next section (RQ3).

3.3 An In-Depth Analysis of Three Individual Users

3.3.1 Kobe

Despite being one of the younger respondents in the sample, Kobe (8;11) is the top user of English for both RUE (11.6%) and RAE (12.4%) in our corpus. To explain why this is the case, we performed a topic analysis on all recognizable English insertions (N = 150, token count) identified in Kobe’s recorded speech. Our method consisted of tagging the conversation topic in relation to which each insertion was uttered. This was done according to a binary distinction between “gaming-related” and “non-gaming-related” utterances in the conversation. Our choice for this distinction was based on gaming being a highly English-prone semantic field, following the previous research discussed in Section 1.2 (cf. Puimège and Peters, 2019; De Wilde et al., 2020; De Wilde et al., 2021; Bollansée et al., 2021) and following insights from a qualitative exploration of Kobe’s interview. The results of the topic analysis, as well as an account for the total number of tokens for RAE and RUE are given in Table 4.

TABLE 4

TABLE 4. Kobe.

Table 4 shows that out of a total of 150 English tokens identified in Kobe’s speech, 103 tokens or 68.7% were uttered when talking about gaming. The percentage even goes up if we only consider RAE: 62 out of 81 tokens, corresponding to 76.5%. Excerpt (6) clarifies what kind of gaming-related insertions we are referring to. Kobe is talking about his favorite video game “Fortnite”:

(6) ^∗KOB: der is ook zo iemand een boss in de game.%eng: there is also like someone a boss in the game.^∗KOB: en der zijn twee (.) &-euh aan elke toren is er zo een boss.%eng: and there are two (.) &-uh at each tower there is like a boss.^∗KOB: en ja die kan je dan killen.%eng: and yes then you can kill him.^∗KOB: want der is ook in Sweety Sand(s) is er zo iemand da(t) jou altijd iets gratis geeft.%eng: because there is also in Sweety Sands there is like someone that always gives you something for free.^∗KOB: soms geeft die jou ook een medic weapon.%eng: sometimes he also gives you a medic weapon.

The excerpt confirms Kobe’s frequent use of recognizable avoidable English when he says boss (Dutch alternative: baas), game (Dutch alternative: spel), killen (Dutch alternatives: vermoorden/doden) and medic weapon (Dutch alternative: medisch wapen). Sweety Sands is the name of a “location” in Fortnite and therefore gets the label of RUE. What is more, the excerpts reveals that, in general, Kobe is a frequent gamer which, in turn, is confirmed by Kobe’s further answers to the questions in the sociolinguistic interview. Apart from explicitly mentioning that he likes gaming and that he games for several hours a day, he also steers the conversation towards the gaming topic, even when he is asked a gaming-neutral question, as can be seen in Excerpt (7):

(7) ^∗INT: en maak jij ook wel es ruzie (.) me(t) jouw broer?%eng: and do you sometimes argue (.) with your brother?^∗KOB: &-euh ja veel.%eng; &-uh yes a lot.^∗INT: ja en waar gaat dat dan over?^∗eng: yes and what is it about then?^∗KOB: omdat die da(t) wapen eerst wilt enzo.%eng: because he wants that weapon first and so on.^∗KOB: want ik had ne keer ik was daar alleen geland en ik heb da(t) per ongeluk gedropt.%eng: because I had once I landed there once alone and I dropped it by accident.^∗KOB: en dan had die da(t) die gouden shot gun en dan wouk [: wou ik] da(t) terug.%eng: and then he had that that golden shot gun and then I wanted it back.

Kobe’s tendency to talk about games all the time (namely in 196 out of his 533 utterances), accompanied by his high frequent English use when doing so, implies a strong topic effect of gaming. This could explain why Kobe is an outlier, producing a much high number of English insertions than the other respondents in the sample. To verify if the topic effect of gaming can indeed account for the variation, we turn to Max (11;6).

3.3.2 Max

Max is nearly 3 years older (11;6) than Kobe (8;11), yet uses very few English insertions (both RAE and RUE). We performed the same topic analysis as described in 3.3.1, which led to the following results:

Table 5 shows that only 10 tokens, or 22.2%, of Max’s English insertions were produced when discussing games. Comparing this to Kobe’s 68.7% and knowing that Max did not talk much about gaming (cf. in only 45 utterances vs. Kobe’s 196 utterances), we would be inclined to conclude that talking about games is indeed a predictor for the amount of English lexical material in preadolescents’ speech. However, this statement has to be nuanced since Max talked about gaming a little bit in the sociolinguistic interview, yet when doing so, did not produce English insertions, as becomes clear in Excerpt (8):

(8) ^∗MAX: &-euh een pinkend hartje een spelletje zowa(t) met een vogel da(t) een ja (.) er zijn zo allemaal streepjes.%eng: &-uh a pounding heart a game with like with a bird that a yes (.) there are like all of these stripes.^∗MAX: en dan zijn er een paar streepjes gevuld en ééntje is over.%eng: and then there are a couple of filled stripes and one is left.^∗MAX: en dan is er nog zo één blokje vrij en die moe(t) je dan na(ar) boven naar onder na(ar) boven naar onder en zo tot je dood bent.%eng: and then there is only one empty block and you have to (get it) up and down up and down and so on until you’re dead.

TABLE 5

TABLE 5. Max.

In sum, when preadolescent children talk about gaming, they do not necessarily use a lot of English. Max talks about gaming but solely uses Dutch in doing so. This is probably due to the fact that Max’s parents do not allow him to play games (see Excerpt 9), which does not give him the opportunity to get familiar with the ingroup gaming code.

(9) ^∗ INT: game jij ook wel es?%eng: do you sometimes play games?^∗MAX: ah nee mijn mama en papa zijn daar tegen.%eng: ah no my mom and dad are against that.^∗MAX: ma(ar) ik ik zou ik vin(d) het leuker zonder gamen.%eng: but I I would I find it more fun without gaming.

Therefore, it would seem that rather than only a topic effect of gaming, English use is also connected to frequently engaging in gaming activities. We would then find high-frequent English use for high-frequent gamers. Camille (12;5) has the ideal profile to check whether this assumption holds true.

3.3.3 Camille

Camille shows the highest RAE frequency and second-highest RUE frequency of the girls, of which she, herself is one of the oldest (12;5). From the sociolinguistic interview, we know she likes to play video games like Minecraft and Mario Kart. She also does this frequently and plays together with her friends. Table 6 presents the topic analysis (cf. Section 3.3.1) in terms of English tokens produced in gaming- and not-gaming-related utterances.

TABLE 6

TABLE 6. Camille.

Contrary to our expectations, merely 34.0% (N = 35) of Camille’s insertions were produced when she was talking about video games (97 utterances). This percentage is only half of Kobe’s 68.7% (Table 4) and is situated closer to Max’s 22.2% (Table 5). From these numbers, it appears that “frequently playing games” cannot entirely account for Camille’s peak in English use. In what follows, we first throw light on Camille’s low frequency of gaming-related English insertions after which we provide an alternative hypothesis to account for her outlier profile.

Although Camille frequently plays games, she does not produce a large number of gaming-related English insertions. Camille provides a possible explanation for this herself, as Excerpt (10) illustrates:

(10) ^∗INT: en game jij ook wel es?%eng: and do you sometimes play games?^∗CAM: ja ik heb een Nintendo Switch.%eng: yes I have a Nintendo Switch.^∗CAM: die heb ik gekregen in quarantaine.%eng: I got it during quarantine.^∗CAM: &-euh omda(t) je omda(t) ik mij toch redelijk snel verveel enzo.%eng: &-uh because you because I get bored fairly quickly and all.^∗CAM: daar heb ik wel een aantal spelletjes op staan.%eng: I do have some games on there.^∗CAM: ik speel vaak me(t) mijn vrienden Mario Kart.%eng: I often play Mario Kart with my friends.^∗CAM: en ja (.) tis nie(t) da(t) ik zo echt serieus game ofzo.%eng: and yes (.) it’s not like I’m really serious about gaming or anything.^∗CAM: ma(ar) ik vind da(t) wel leuk gewoon als ontspanning.%eng: but I like that just to relax.

Since Camille states she’s not too “serious about gaming”, her not being committed to gaming in the way for example Kobe is, can presumably account for her lower gaming-related English use.

This still leaves us with Camille’s 103 English tokens to discuss, a number that is already closer to Kobe’s 150 tokens, yet not to the same extent connected to gaming. A clarification for this can be found in the nature of the English insertions Camille produced. Consider Excerpts (11) and (12) on this account:

(11) ^∗INT: en heb jet [: je het] ondertussen wel al uitgepraat?%eng: and have you talked things out by now?^∗CAM: ja ma(ar) das [: dat is] zo die heeft da(t) al zo vaker gedaan he.%eng: yes but that’s like she has done that before hey.^∗CAM: ma(ar) dan me(t) kleinere dingen zo echt zo last minute zo nee zeggen ofzo.%eng: but then with smaller things like really like last minute like say no or something.(12) ^∗INT: en ik vroeg mij af of (.) jij ook nog weet wa(t) je dit weekend allemaal gedaan hebt?%eng: and I was wondering if (.) you also still remember what you did during the weekend?^∗CAM: &-euh ja ik ben zeg maar met die vriendin met Fara ben ik naar Antwerpen geweest.%eng: &-uh yes I went like with that friend with Fara I went to Antwerp.^∗CAM: dus zaterdag waren wij in Antwerpen.%eng: so on Saturday we were in Antwerp.^∗CAM: en dan hebben wij daar zo een beetje gaan shoppen enzo.%eng: and then we have done like a little bit of shopping and stuff.

In Excerpt (11), Camille is talking about an argument she had with her friend about participating in a summer camp, stating that her friend decided at the last minute not to come, which made her quite angry. Next, in Excerpt (12), Camille discusses her weekend in which she did a shopping visit to Antwerp, together with her friend Fara. Both last minute and shopping were produced when talking about stereotypically girl-oriented¹⁴ activities. In fact, 37.8% (14 out of 37) of Camille’s RAE tokens, and 9.1% (6 out of 66) of her RUE tokens, can be traced back to these girl-oriented topics, with additional English insertions like pony (RUE) and playbacken, tie dyen and slash (RAE). The peak in English use for Camille would therefore not result from a topic effect of gaming alone, nor would it stem solely from her frequently playing video games. Camille’s high frequent use of English additionally seems connected to her being a preadolescent girl, talking regularly about girl-oriented topics.

To corroborate this, we briefly discuss two other girls in the sample, Charlotte (9;10) and Sarah (12;8), who report on an argument with their friends (Excerpt 13) and on the boys’ extracurricular activities on the playground (Excerpt 14):

(13) ^∗CHA: en nu gaan die ruzies zo meestal over van “bemoeit u me(t) uw eigen zaken” ofzo.%eng: and now those arguments are like usually about like “mind your own business” or something.^∗CHA: en dan beginnen wij te diss(en) te(gen) allez beginnen wij zo scheldwoorden tegen and elkaar te zeggen.%eng: and then we start dissing to well we start saying like swear words to each other.^∗CHA: omda(t) &-euh ja wij denken dan van “we gaan (i)kga mij we gaan ons nie(t) laten doen” enal.%eng: because &-uh yes we then think like “we’re going to I’m not letting me we’re not letting us be pushed around” and all.

(14) ^∗SAR: en &-eumh hij had zijn vrienden hadden daar zo op gebottleflipt.%eng: and &-umh he had his friends had like bottle flipped on that.^∗SAR: das [: dat is] zo me(t) waterflesje en dan moet ge zo draaien dat da(t) zo blijft staan ofzo.%eng: that’s like with a water bottle and then you have to like turn so that it like stays up or something.

Here, Charlotte and Sarah too talk about girls’ activities. In Excerpt (13), Charlotte uses the word dissen (to diss) to explain how an argument with her friends usually ensues. Dissen/To diss has a slang origin and is an informal way of saying you disrespect someone by insulting them in a certain way (cf. Cambridge Dictionary, 2021). The English verb has clearly made its way into Dutch where it has been adapted, presumedly inevitably (cf. Winter-Froemel, 2008), to Dutch verbal inflection. In Excerpt (14), the English verb to bottle flip underwent the same morphological adaptation to Dutch. Sarah is talking about the boys, including her brother, and informs the researcher about their activities on the playground: most often they bottle flip. Remarkably, this gossipy statement is immediately followed by a detailed explanation of the verb. The same happened in the conversation with Charlotte (Excerpt 13) who even stopped in the middle of the word dissen and converted to an explanation “saying swear words to each other”. The English insertions in the excerpts thus interestingly get flagged¹⁵, i.e. marked as being foreign material in the receptor language (Levendis and Calude, 2019). Setting this aside, Excerpts (13) and (14) seem to confirm an additional topic effect of girl-oriented activities, where high girlishness leads to more English use.

4 Discussion and Conclusion

In this paper we investigated preadolescents’ developing use of English lexical resources in Belgian Dutch. In our effort to chart the largely undocumented sociolinguistic transition process from childhood to adolescence, we addressed three research questions: RQ1. How many English insertions do we find, taking into account the type of English used?; RQ2. To what extent is the use of these English insertions stratified by age and gender?; and RQ3. Can the patterns found in RQ1 and RQ2 be explained through an in-depth analysis of Kobe (8;11), Max (11;6) and Camille (12;5)?

The corpus of 15,465 preadolescent utterances was mined for English insertions according to our phased identification protocol, focusing on recognizable unavoidable English (RUE) and recognizable avoidable English (RAE). Results showed an overall utterance-based English frequency of 9.7% (RQ1) This percentage fits in nicely within the expected trajectory, as it is situated in between the previous finding of less than 1% of English insertions in the family home (Zenner and Van De Mieroop, 2021) and corresponding 13.3% in teenagers’ social media interactions (De Decker and Vandekerckhove, 2012). As for stratification (RQ2), we found no age effect, some indication of a gender effect for RAE with girls using it significantly more often than boys, particularly when leaving out outlier Kobe; and high levels of idiosyncrasies. These idiosyncrasies reflect the importance of the gaming topic (Kobe) which, at the same time, is nuanced by the possibility to use only Dutch when talking about gaming (Max) and the presumably additional importance of activities stereotypically oriented toward preadolescent girls (Camille, RQ3). This also points to the distinction between English-prone semantic fields and English-prone topics, a tension we have not explored significantly in our contribution. For instance, the English word bro can occur in a Dutch conversation on gaming (an English-prone topic), without itself belonging to the semantic field of gaming. In any case, the importance to distinguish between different types of English is confirmed, since we found that RAE (3.0%) occurred much less frequently than RUE (6.7%).

What can these results reveal about the overall development of the position of English in the lexicon of preadolescents? We advance the hypothesis that English is starting to become a youth language identity marker for our respondents (Leppänen, 2007; Leppänen and Nikula, 2007), with an emerging and varying socio-indexical potential, a process in which one preadolescent seems to be faster than the other. Four arguments derived from our study can support this hypothesis.

Firstly, we did not see an age-related development in cohorts but rather saw a fairly stable use of English across preadolescents disrupted by the presence of several outliers. Secondly, the socio-indexical potential of English is observed in the varying results for the use of English when discussing gaming, which seems to be linked to incipient identity construction. Whether or not a preadolescent uses English when talking about gaming, seems in part to depend on the degree to which the preadolescent aims to assume the “gamer identity”. Compare Kobe, an ardent gamer and a clear member of the ingroup (high gaming-related English use), to Camille, also a gamer, but not taking it “too seriously”, thus not aiming to become an ingroup member (lower gaming-related English use), and to Max, a clear outsider to the gaming community (near-absence of gaming-related English use). Thirdly, a gender effect seems to be at play, with different socio-indexical attributes associated with English in both groups. This is firstly supported by the significance of RAE frequencies in these two groups (when outlier Kobe is excluded), and further in the stereotypically girl-oriented topics addressed and “soon-to-be teenage girl” identity assumed by the girls. Our gender-related conclusions tie in with the mixed gender results in previous (developmental) sociolinguistic studies (Sections 1.1 and 1.2, cf. Nardy et al., 2013; Zenner and Van De Mieroop, 2017) and indicate that gender may be created very “locally” in discourse. This certainly requires further research. Lastly, the instances of flagging we found in Excerpts (13) and (14) can be interpreted as a developing awareness of the ingroup and outgroup marking potential of English insertions. In providing additional explanation (flagging) to their use of English terms (dissen, bottle flippen), the girls indicate that they believe the researcher is likely not familiar with these terms. This further indicates that they might consider the researcher as an outgroup member. These observations of flagging and outgroup marking seem to be a sign of emerging sociolinguistic competence.

In order to solidify our claim about an emerging and variable socio-indexical potential of using English in Dutch, seeping in at various speeds, further research is needed that additionally allows us to overcome some of the shortcomings of this study. First of all, although we factored in the possibility of lexical competition through our distinction between RUE and RAE, follow-up studies might wish to take the onomasiological perspective further (Geeraerts, 2010; Zenner et al., 2014). Particularly our topic analysis would benefit from a comparison of English insertions with the Dutch words or counterparts they are encountered with, for instance benchmarking the use of game against its Dutch counterpart (video)spel. For one thing, this would allow us to arrive at a more local operationalization of relative markedness in our English identification protocol. Our current operationalization is indeed not airtight as it does not account for the individual perspective of the language user. Consider for example stick (as in hockey stick), the third-most frequent RAE insertion in the corpus (see Section 3.1). Stick received the RAE label because of its Dutch alternative stok, which was frequent enough to be considered in the study (4,000+ hits on Dutch Google). A more fine-grained analysis of the Google hits suggests nonetheless that stok was only used by hockey novices or amateurs. Our respondents are frequent hockey players, however, for whom “stick” is probably the only possible lexicalization. The RAE status of stick and possibly other insertions as well is thus questionable since it does not account for the respondents’ individual worlds and the existence of Dutch alternatives therein. This further supports the fact that specificity might be an important motivation for using English insertions (see Backus, 1996 on specificity as driving factor in Turkish-Dutch interactions).

Another prospect for further studies is aiming to find out what our respondents themselves consider to be “English insertions”. Given the instances of flagging we observed (cf. explaining an English insertion to the researcher e.g. dissen, gebottleflipt), it is likely that respondents are aware of having produced that insertion, moreover, that they have a perception of what characterizes it as being “English”. This points towards the importance of including the respondents’ perception in the analysis and is what currently lacks in the recognizability level of our identification protocol. That recognizability level is now based on two proxies, namely “grapheme-phoneme mapping” and “non-Dutch chargrams”, which can both be criticized for relying too much on spelling conventions our young respondent group has little affinity with. Besides, following the diverging sociolinguistic competence of our respondents, some of whom are not yet presenting flagging practices, we could expect that at least part of them are not aware of producing English at all and think they are using Dutch. Again, the identification protocol does not yet include this individual metalinguistic feature. We therefore argue for a center role for perception and metalinguistic awareness (cf. Drager and Kirtley, 2016) in future studies, also factoring in the evolution in children’s use of English and the input they rely on in their acquisition (see Leona et al., 2021). These concepts, when thoroughly methodologically developed and implemented, can shed more light on (preadolescent) children’s developing production of and socio-indexical attributes awarded to English and, in that sense, on their motivations for using English. Likely candidates of such social-attributes for English are ‘coolness’ (Garley, 2019) or modernity (Piller, 2001) and motivations could include emotion regulation (Eekhof, 2017) or increasing semantic specificity (Backus, 1996).

As a final avenue for future work, we highlight the importance of including a variety of data types. Our one-sided method to conduct sociolinguistic interviews is indeed not without problems since it can be responsible for the above-mentioned outgroup effect: as the participants in this interview are typically only the researcher and the interviewee, the research setting is characterized as “distanced”. Apart from the age difference between the participants (approximately 15 years), the interviewee has never met the researcher who does not classify as a peer group member. In sum, it may be that the overall fairly low proportion of utterances containing English (specifically RAE) in our preadolescent corpus is a result of the one-on-one research context, with the children realizing that the researcher is not part of the ingroup and will not understand the ingroup English-inspired code. This conclusion must, however, be tempered by the fact that, in spite of the distanced research setting, and precisely because of careful preparation and child-friendly changes to the format, respondents clearly felt at ease as they spontaneously started to talk about a range of fairly private subjects: getting into trouble, keeping secrets from their parents and having a little sweetheart. In future research, it would nonetheless still be interesting to include data collection in peer group settings (without the researcher being present) in order to counterbalance the possible English inhibitory effect of the one-on-one research setting.

Although the suggested future work is needed to get a full picture, this study already allows us to draw important conclusions about the development of English use in preadolescents, revealing part of the understudied trajectory from childhood to adolescence and arguing for a so-called socio-pragmatic developmental turn in anglicism research.

Data Availability Statement

The data sets presented in this article are not readily available because of ethical restrictions that could comprise the privacy of the research participants.

Ethics Statement

This study involving human participants was reviewed and approved by the Social and Societal Ethics Committee (SMEC) at KU Leuven, approval number G-2020-1998-R5. Written informed consent to participate in this study was provided by the participants’ legal guardian/next of kin.

Author Contributions

MS performed the data collection, analyzed the data and is first author of the paper. EZ is the overseer of the research project. She provided critical feedback and helped shape the data collection, analysis and manuscript.

Funding

This study is part of larger PhD project that is funded by C1420043, KU Leuven.

Conflict of Interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Publisher’s Note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Acknowledgments

We would like to thank Sophie Holmes-Elliott for her inspiring advice in drafting a child-friendly sociolinguistic interview protocol.

Supplementary Material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fcomm.2021.788768/full#supplementary-material

Footnotes

¹For secondary school students in other countries, studies have shown similar high English proficiency rates, yet equally reveal L1 transfer effects (see Lorenz et al., 2021).

²In this paper, age is displayed in “years(;)months”, following the CHAT conventions of the CHILDES project (MacWhinney, 2000, see Section 2.3).

³Two participants, brother and sister, have been raised bilingually in Dutch and French. Where necessary, this is taken into account in the analysis of the data.

⁴Except for the two oldest girls in our sample (Girl[13] and Girl[14], cf. Table 1) whose schools organize English courses as of the first year of secondary school. We factored this into the analysis of the data.

⁵More specifically, respondents, after being asked for “assent”, were told they were participating in “linguistic research” leaving our goal to monitor English insertions unclear. Parents had full disclosure from the start and provided informed consent, but were repeatedly asked not to share the research purpose with their children. All personal information in this paper has been pseudonymized following the ethics application for this study and the larger ongoing research project, approved by the Social and Societal Ethics Committee “SMEC” at KU Leuven, approval number G-2020-1998-R5.

⁶After extensive pretesting of both the online and the traditional face-to-face format, we decided to work with online data collection. Aside from the fact that the online format was more practical to organize during the pandemic, a recorded meeting was considered less intrusive than a face-to-face interview. The main reason for this is the discreetness of the camera in the online setting, resulting in a reduced “Observer’s paradox” (Labov, 1972).

⁷Needless to say, priming effects have been checked for the entire corpus.

⁸The number of points in between the brackets indicates the duration of the pause with (.) corresponding to a short pause and (…) corresponding to a very long pause.

⁹We use “Dutch alternative” to refer to all lexicalizations behaving as near-synonyms, thus following Edmonds and Hirst (2002) who state that the existence of true synonymy is arguable. This implies that the Dutch alternatives inevitably differ from the English borrowings in at least some ways, “varying in their shades of denotation, connotation, implicature, emphasis, or register” (ibid: 107) and also, as is the case for most loanwords, in their degree of specificity (see Backus, 1996).

¹⁰“Fortnite” is a free and popular online video game with the purpose of being the last survivor on an island.

¹¹We followed the gender distinction as it is applied in the community of practice under scrutiny. The hockey club works with the traditional and mandatory gender division in boys’ and girls’ (competition) teams.

¹²In this study, we are performing multiple tests on the same respondent variable, which leads to higher probability of observing significant effects by chance. Bonferroni correction accounts for this by adjusting the significance level: the traditional significance level (p < 0.05) is divided by the number of tests performed.

¹³Note that the effect here remains significant upon Bonferroni correction.

¹⁴The division in girl- and boy-oriented activities, we are aware, is in need of more nuance and research and would benefit from being connected to the ongoing societal debate on gender stereotyping.

¹⁵As Levendis and Calude (2019 : 1) note, flagging can be done in multiple ways including translating/explaining a loan, or using bold face font or italics to demarcate it from surrounding discourse.

References

Andersen, G. (2005). “Assessing Algorithms for Automatic Extraction of Anglicisms in Norwegian Texts,” in Paper presented at the Corpus Linguistics Conference, Birmingham, July 2005.

Google Scholar

Andersen, G. (2017). A Corpus Study of Pragmatic Adaptation: The Case of the Anglicism [ Jobb ] in Norwegian. J. Pragmatics 113, 127–143. doi:10.1016/j.pragma.2016.12.015