Highlights
-
• Examined the potential nonlinear relationship between bilingual engagement and cognition across adulthood.
-
• Found a limited relationship between bilingual engagement and cognition in regional minority language speakers.
-
• Provided suggestions for research into language control and cognition in diaglossia.
1. Introduction
Cognitive decline associated with typical aging usually begins around ages 50–60 – though subtle changes can be found even in young adulthood (Salthouse, Reference Salthouse2004) – and it tends to escalate with advancing age (Salthouse, Reference Salthouse2019; Ghisletta et al., Reference Ghisletta, Mason, Von Oertzen, Hertzog, Nilsson and Lindenberger2020). Cognitive decline trajectories differ in onset and rate, however, which suggests that some individuals are more resilient to the decline in cognitive functioning associated with neural deterioration (Stern, Reference Stern2012). Several lifestyle factors and experiences have been associated with the preservation of cognitive functioning in the context of typical or pathological brain changes with increasing age. These include but are not limited to higher educational attainment (Meng & D’Arcy, Reference Meng and D’Arcy2012), occupational complexity (Hyun et al., Reference Hyun, Katz, Lipton and Sliwinski2019), higher levels of physical activity (Blondell et al., Reference Blondell, Hammersley-Mather and Veerman2014), and life-long engagement in mentally stimulating leisure activities, such as playing a musical instrument (Amer et al., Reference Amer, Kalender, Hasher, Trehub and Wong2013).
Another suggested cognitive resilience factor that has received ample attention in the past 15 years is speaking multiple languages (Bialystok, Reference Bialystok2021a). It has been posited that neurocognitive adaptations related to bilingualism emerge through the continuous engagement of attentional control mechanisms to regulate the activation of multiple languages in the mind in order to speak the language(s) appropriate for a given context (Kroll et al., Reference Kroll, Dussias, Bice and Perrotti2015; Bialystok & Craik, Reference Bialystok and Craik2022). According to this view, the (lifelong) engagement with multiple languages increases efficiency in using neurocognitive resources associated with attentional control, leading to better cognitive aging outcomes. In clinical settings, for instance, bilinguals have been shown to present with dementia symptoms several years later than their monolingual peers, suggesting that bilingualism contributes to a compensatory mechanism for dementia-related brain pathology (see recent meta-analyses by Anderson et al., Reference Anderson, Hawrylewicz and Grundy2020; Brini et al., Reference Brini, Sohrabi, Hebert, Forrest, Laine, Hämäläinen, Karrasch, Peiffer, Martins and Fairchild2020; Paulavicius et al., Reference Paulavicius, Mizzaci, Tavares, Rocha, Civile, Schultz, Pinto and Trevisani2020). Furthermore, speaking multiple languages seems to interact with typical cognitive aging as well, as several studies have found better cognitive performance by cognitively healthy older adult bilinguals in comparison to monolinguals (as reviewed in Dash et al., Reference Dash, Joanette and Ansaldo2022; Gallo et al., Reference Gallo, Myachykov, Shtyrov and Abutalebi2020, Reference Gallo, DeLuca, Prystauka, Voits, Rothman and Abutalebi2022).
Onset of symptoms and rate of cognitive decline have been well documented for bilinguals diagnosed with dementia and its preclinical stages, but investigations into the relationship between bilingual experience and typical (neuro)cognitive aging trajectories, taking an adult lifespan perspective and including data from young adulthood all the way into older adulthood, are limited (but see Bialystok et al., Reference Bialystok, Craik, Klein and Viswanathan2004; DeLuca & Voits, Reference DeLuca and Voits2022; Nichols et al., Reference Nichols, Wild, Stojanoski, Battista and Owen2020; Vega-Mendoza et al., Reference Vega-Mendoza, Eriksson Sörman, Josefsson and Ljungberg2022). In part, this may be because substantial resources are needed for the recruitment and testing of a sufficient number of participants across adulthood to detect the generally small effects that are reported for cognitive consequences of bilingualism (Bialystok, Reference Bialystok2021b). Given that age-related cognitive decline typically commences in middle age, but middle-aged adults have rarely been included in bilingualism research, the question of how bilingualism could mitigate cognitive decline from younger adulthood into older adulthood requires further scrutiny.
The specific neurocognitive effects of bilingual experience are still poorly understood, partly because operationalizing and quantifying bilingual experience is far from straightforward. Several recent models have attempted to describe the dynamic nature of functional, structural and behavioral adaptations ensuing from bilingual experiences, such as age of onset of L2 acquisition/bilingualism, (relative) language proficiency and patterns of language exposure and use (Green & Abutalebi, Reference Green and Abutalebi2013; Grundy et al., Reference Grundy, Anderson and Bialystok2017; DeLuca et al., Reference DeLuca, Segaert, Mazaheri and Krott2020; Pliatsikas, Reference Pliatsikas2020). These models recognize that neurocognitive adaptations may differ as a function of the demands the sociolinguistic environment incurs on bilinguals’ attentional control. For example, the type of interactional context in which bilinguals (primarily) engage has been hypothesized to play a crucial role in cognitive adaptations (e.g., Green & Abutalebi, Reference Green and Abutalebi2013). In dual-language contexts, the use of two languages is generally accepted, but restrictions are placed on which language to use with whom. Bilinguals engaging in such contexts need to cognitively manage the uncertainty of which language to use in a particular situation with a certain conversational partner by continuously attending to environmental cues (Gullifer & Titone, Reference Gullifer and Titone2021a). This need may be argued to be less strong for bilinguals whose language use is more compartmentalized.
Gullifer and Titone (Reference Gullifer and Titone2020) have recently attempted to capture this language-related uncertainty and diversity in a single, continuous measure labeled language entropy. Language entropy reflects a spectrum of engagement with multiple languages, ranging from compartmentalized language use (i.e., using a single language in a certain context, leading to low uncertainty) to fully integrated language use (i.e., perfectly balanced engagement with multiple languages, leading to high uncertainty).Footnote 1 Language entropy has already been shown to relate to (neuro)cognition in bilinguals (Gullifer et al., Reference Gullifer, Chai, Whitford, Pivneva, Baum, Klein and Titone2018; Sulpizio et al., Reference Sulpizio, Del Maschio, Del Mauro, Fedeli and Abutalebi2020; Li et al., Reference Li, Ng, Wong, Lee, Zhou and Yow2021; Van den Berg et al., Reference van den Berg, Brouwer, Tienkamp, Verhagen and Keijzer2022). However, effects of higher engagement with multiple languages on cognition have not consistently been found (Kałamała et al., Reference Kałamała, Szewczyk, Chuderski, Senderecka and Wodniecka2020; Wagner et al., Reference Wagner, Bekas and Bialystok2023), are only found on structural brain measures and not on behavioral indices (Voits et al., Reference Voits, Rothman, Calabria, Robson, Aguirre, Cattaneo, Costumero, Hernández, Puig, Marín-Marín, Suades, Costa and Pliatsikas2023) or are small and not highly predictive of cognitive performance (Gullifer & Titone, Reference Gullifer and Titone2021b). To further consolidate the language entropy measure and its underlying theoretical perspective, generating data from different geographic locations and diverse bilingual populations is crucial (Gullifer & Titone, Reference Gullifer and Titone2021b; Wagner et al., Reference Wagner, Bekas and Bialystok2023).
1.1. Regional minority language bilingualism
Under-researched bilingual contexts that can be informative for the investigation into the relationship between bilingual engagement and cognition are environments concerning regional minority languages in dynamic interaction with a dominant majority language. What constitutes a language, minority language or dialect is very much a sociopolitical decision (e.g., Auer, Reference Auer, Delbecque, van der Auwera and Geeraerts2005; Kloss, Reference Kloss1967; but see Melinger, Reference Melinger2018), but here, we are specifically interested in the association between cognition and the proficient engagement with two typologically close languages, one of which is the standard language in a given country and the other a regional language spoken by a minority (sometimes also referred to as “bidialectalism”; Hazen, Reference Hazen2001; Melinger, Reference Melinger2018). In previous work on bilingualism, speakers of regional minority varieties may have been labeled as monolinguals, presumably also because some speakers themselves perceive their regional language as a variety of the standard language rather than a separate language (Declerck & Kirk, Reference Declerck and Kirk2023). Still, there is emerging evidence that language control mechanisms involved in speaking a regional minority language and a majority language are quite similar to those found in societal bilingual groups speaking more typologically distant languages, but such effects are allegedly sensitive to factors related to the sociolinguistic context (Kirk et al., Reference Kirk, Kempe, Scott-Brown, Philipp and Declerck2018, Reference Kirk, Declerck, Kemp and Kempe2022; Vorwerg et al., Reference Vorwerg, Suntharam and Morand2019). In addition, much remains unknown about the implications of regional minority and majority language usage dynamics for adaptations of domain-general attentional control.
The relatively few studies on cognitive adaptations in regional minority language speakers have collectively investigated several language pairs and age groups, but so far, findings are mixed (for a recent review, see Alrwaita et al., Reference Alrwaita, Houston-Price and Pliatsikas2023). The discrepancy in findings could be attributed to the fact that these bilingual groups often operate in different interactional contexts. For example, several studies have emphasized the possible differential effects of low versus high bilingual engagement in educational settings in speakers’ childhood and adolescence (e.g., Lauchlan et al., Reference Lauchlan, Parisi and Fadda2013; Kirk et al., Reference Kirk, Fiala, Scott-Brown and Kempe2014; Scaltritti et al., Reference Scaltritti, Peressotti and Miozzo2017; Garraffa et al., Reference Garraffa, Obregon, O’Rourke and Sorace2020). However, earlier work has mostly conducted group comparisons and has not yet directly related individual levels of engagement with the minority and majority languages to cognitive functioning (apart from Houtzager et al., Reference Houtzager, Lowie, de Bot, Filipović and Pütz2014; Poarch et al., Reference Poarch, Vanhove and Berthele2019). Considering that relatively few studies have investigated cognitive adaptations associated with bilingual experience through the lens of regional minority language contexts, and especially against the backdrop of cognitive aging, our study aims to examine the effect of individual bilingual engagement levels across the adult lifespan in such environments.
1.1.1. The cases of Frisian and Low Saxon
The regional minority language contexts in the north of the Netherlands offer interesting case studies for this purpose. The northern region comprises regional minority language communities in which Frisian and Low Saxon are spoken, in addition to the majority language (Dutch), by part of the population in the provinces of Fryslân, and Groningen and Drenthe, respectively. Crucially, and as will be further explained below, Frisian and Low Saxon differ in their sociopolitical status, vitality and their presence in the respective sociolinguistic landscapes. These differences necessitate contrasting these two regional language speaker groups in our study, in addition to assessing bilingual engagement at the individual level (cf. Wagner et al., Reference Wagner, Bekas and Bialystok2023).
Frisian is a West Germanic language spoken as a first language by approximately 57.3% of people living in the province of Fryslân, but higher percentages (64.1%–75%) are reported for people speaking the language well to very well (Klinkenberg et al., Reference Klinkenberg, Stefan and Jonkman2018; Provinsje Fryslân, 2020). This suggests that a portion of the Frisian population has acquired Frisian as “new speakers,” which denotes people who have acquired a minority language through education or immersion, as opposed to home exposure (O’Rourke et al., Reference O’Rourke, Pujolar and Ramallo2015; Kircher & Vellinga, Reference Kircher and Vellinga2023). Thus, although Frisian is a minority language within the wider context of the Netherlands, Frisian is spoken widely in Fryslân. Frisian obtained the highest degree of recognition under Part III of the European Charter for Regional or Minority Languages in 1998, and since 2014, Frisian and Dutch have been co-official languages of Fryslân. This means that individuals who speak Frisian are entitled to use their preferred language within the Frisian language region. This right extends to various scenarios, including communication with municipal or provincial employees, participation in court hearings, and the administration of official oaths. Furthermore, primary schools in Fryslân have been required to offer Frisian as a subject since 1980, and schools are allowed to teach in the Frisian language in addition to Dutch (but see Bayat et al., Reference Bayat, Kircher and Van de Velde2023, on the limited implementation of existing regulations). It is therefore unsurprising that Frisian speakers are reported to have equally positive attitudes toward speakers of the minority and majority language (Hilton & Gooskens, Reference Hilton, Gooskens, Gooskens and van Bezooijen2013), consider speaking Frisian being a contributing factor to their shared sociocultural identity (Fries Sociaal Planbureau, 2021) and, anecdotally, generally view Frisian and Dutch as two separate languages (Bosma et al., Reference Bosma, Knooihuizen, Blom and van Koppen2023).
Low Saxon is another West Germanic language, and it is spoken in regions in the north-east of the Netherlands, Germany and Denmark. Our study revolves around the Low Saxon varieties spoken in the Dutch provinces of Groningen and Drenthe. It has been estimated that approximately one-third of the population in Groningen and Drenthe speaks exclusively Low Saxon at home, but higher percentages have been reported for people speaking both Low Saxon and Dutch at home (Schmeets & Cornips, Reference Schmeets and Cornips2022; Buurke et al., Reference Buurke, Bartelds, Knooihuizen and Wieling2023) and for people being able to speak a variety of Low Saxon but do not do so at home (Bloemhoff, Reference Bloemhoff2005). Low Saxon received official recognition in the Netherlands in 1998 under Part II of the European Charter for Regional or Minority Languages. Although the use and acknowledgment of the language is encouraged, there are no specific laws addressing, for instance, the right to receive education in Low Saxon, and, as such, it lacks the same legal protection that Frisian has. Despite its official recognition, Low Saxon speakers tend to perceive Low Saxon as a variety of Dutch. This is in contrast with the attitudes of Frisian speakers in Fryslân toward their regional minority language, who view Frisian as a separate language from Dutch (Bosma et al., Reference Bosma, Knooihuizen, Blom and van Koppen2023).
When we further compare these two regional minority language areas, it becomes clear that Frisian is more strongly present in Fryslân than Low Saxon is in Groningen and Drenthe. First and foremost, there are substantial differences in the intergenerational transfer of Frisian and Low Saxon. The proportion of Frisian speakers is relatively stable across generations, whereas the speaker population of Low Saxon has been starkly declining, effectively resulting in fewer speakers among younger age groups (Klinkenberg et al., Reference Klinkenberg, Stefan and Jonkman2018; Provinsje Fryslân, 2020; Schmeets & Cornips, Reference Schmeets and Cornips2022; Buurke et al., Reference Buurke, Bartelds, Knooihuizen and Wieling2023). Furthermore, home users of Frisian and Low Saxon also use these regional languages in direct social circles, such as with friends, neighbors, and in town. However, Low Saxon speakers tend to primarily use Dutch in broader public domains (e.g., in shops, at school, or at work), whereas Frisian speakers are more likely to use Frisian in these contexts (Schmeets & Cornips, Reference Schmeets and Cornips2022). In sum, variation in individual bilingual engagement is likely to characterize speakers within and between the two regional minority language contexts. We may observe compartmentalized use as well as high bilingual engagement with the regional language and Dutch in both areas.
Auer (Reference Auer, Delbecque, van der Auwera and Geeraerts2005) categorizes the type of bilingualism observed in most European regional minority language contexts today, including the Frisian and Low Saxon ones examined in this study, as diaglossia. Diaglossia refers to a spectrum ranging from a prototypical variant of the regional minority language, through intermediate forms containing a variable number of regional features, to the standard majority language. As such, in the context of Frisian and Low Saxon, high variability is expected for the linguistic varieties referred to by these labels in our study, such that they range from typical minority language forms to varieties with fewer distinct features from the varieties classified as the standard language. It is also known that speakers can flexibly adjust the number of regional features in their speech, depending on the situation and/or their interlocutor (Ghyselen & Van Keymeulen, Reference Ghyselen and van Keymeulen2016).
Only a handful of studies have examined the cognitive effects specifically associated with regional minority and majority language dynamics in the Netherlands. Looking at early life stages, Blom et al. (Reference Blom, Boerma, Bosma, Cornips and Everaert2017) demonstrated that Frisian–Dutch children and Polish–Dutch children with high proficiency in Polish outperformed Dutch monolinguals on selective attention, whereas no robust difference was found between Limburgish–Dutch children and monolingual children. The authors attributed this finding to the possible difference in the degree of language separation characterizing the different sociolinguistic contexts. Considering that children in the Limburgish context may more frequently mix the regional language with Dutch than the Frisian–Dutch children do, the authors argued that Limburgish–Dutch children may therefore need to engage selective attention to a lesser extent than Frisian–Dutch children (cf. Green & Abutalebi, Reference Green and Abutalebi2013). Furthermore, a longitudinal study by Bosma et al. (Reference Bosma, Hoekstra, Versloot and Blom2017a) found that Frisian–Dutch children who were more balanced in terms of proficiency outperformed Dutch-dominant children on a measure of selective attention at ages 5–6 (see also Bosma et al., Reference Bosma, Blom, Versloot, Lauchlan and Couto2017b), but this effect did not persist over time. This suggests that a more balanced proficiency in the minority and majority languages initially places higher demands on attentional control and fosters adaptations in this domain, but other factors, such as increased literacy in the majority language, may become more predictive of selective attention over time. The enhanced attentional control found for more balanced Frisian–Dutch bilingual children also seems to extend into older adulthood. Houtzager et al. (Reference Houtzager, Lowie, de Bot, Filipović and Pütz2014) found that a greater balance in Frisian and Dutch (comprising past and current usage and language preference) was associated with smaller switching costs for early Frisian–Dutch bilinguals in middle and older adulthood. When comparing these groups to German monolinguals, Houtzager et al. (Reference Houtzager, Lowie, Sprenger and Bot2017) found superior switching performance for the Frisian–Dutch bilinguals, with less steep declines in performance found when comparing the bilingual middle-aged and older adult groups relative to their monolingual peers. In sum, more balanced engagement with multiple languages was related to better cognitive performance in Frisian–Dutch bilinguals. Moreover, bilingualism in Frisian and Dutch (when compared to monolingualism) could potentially attenuate cognitive aging effects, at least in the switching domain. To the best of our knowledge, no previous studies investigating cognitive adaptations specifically examined Low Saxon–Dutch bilinguals (but see Pot et al., Reference Pot, Keijzer and de Bot2018, who included Low Saxon speakers in a broader multilingual sample).
1.2. The current study
The present study aims to address the questions raised in our review of the literature by investigating the relationship between bilingual engagement and cognitive functioning across the adult lifespan in Frisian and Low Saxon minority language contexts. We do so on the basis of a large sample of bilinguals from the Lifelines Cohort Study (see Section 2). Even within these specific regional minority language contexts, the degree of engagement with the minority and majority languages is expected to vary greatly between individuals. Therefore, in line with recent developments in the literature viewing bilingualism as a gradient experience, we operationalize bilingual engagement using language entropy (Gullifer & Titone, Reference Gullifer and Titone2020). We relate the potential interaction between age and bilingual engagement to performance on a range of cognitive tasks available in the context of Lifelines, measuring processing speed, attention, working memory and recognition memory. Our study, therefore, aims to address the following research question and subquestions:
-
1. To what extent is bilingual engagement associated with processing speed, attention, working memory and recognition memory in regional minority language speakers?
-
a) Does the effect of bilingual engagement on cognitive functioning interact with age?
-
b) Does the effect of bilingual engagement on cognition differ between Frisian–Dutch and Low Saxon–Dutch bilinguals?
Although it is possible that a higher degree of bilingual engagement relates to better cognitive functioning across the adult lifespan, based on the presumption that higher engagement requires more adaptation of the attentional control system, it is equally likely that effects of bilingual engagement only emerge when individual differences in cognitive functioning become more apparent, such as in older adulthood. As such, based on the hypothesis that bilingual engagement mitigates the cognitive aging process, we expect to observe a less steep and/or delayed onset of decline in processing speed, attention, working memory and recognition memory across our age range for higher bilingual engagement relative to lower bilingual engagement.
Our second subquestion is more explorative, considering that previous research has not directly compared bilingual engagement effects in two distinct bilingual contexts. In the Frisian context, the use of Dutch and Frisian is accepted in the public domain, but not all residents of Fryslân speak Frisian, and, as such, attentional control is needed to some degree in Frisian–Dutch bilinguals to manage this language-related uncertainty. The situation for the Low Saxon areas is less clear, however. Based on Schmeets and Cornips (Reference Schmeets and Cornips2022), these areas seem to approximate a compartmentalized context in which Low Saxon (sometimes together with Dutch) is mostly reserved for direct social circles, and Dutch is spoken most of the time in other, including more formal, contexts. We may therefore observe a different effect of bilingual engagement on cognitive aging in Frisian–Dutch speakers compared with Low Saxon–Dutch speakers, simply because there are fewer opportunities to use the minority and majority languages within the same interactional context for the latter group.
2. Method
2.1. Setting
Our study was based on data from Lifelines, a longitudinal cohort study in the northern Netherlands (i.e., the provinces of Fryslân, Groningen and Drenthe). Lifelines is a multi-disciplinary, prospective, population-based cohort study examining the health and health-related behaviors of 167,729 persons living in the north of the Netherlands, using a unique three-generation design. It employs a broad range of investigative procedures to assess biomedical, sociodemographic, behavioral, physical and psychological factors that contribute to the health and disease in the general population, with a special focus on multi-morbidity and complex genetics (Scholtens et al., Reference Scholtens, Smidt, Swertz, Bakker, Dotinga, Vonk, van Dijk, van Zon, Wijmenga, Wolffenbuttel and Stolk2015; Sijtsma et al., Reference Sijtsma, Rienks, van der Harst, Navis, Rosmalen and Dotinga2022). The data are taken to be generally representative of the population. However, it should be noted that the Lifelines study population is biased toward participants who are women, highly educated, middle-aged, married and living in a semi-urban environment (Klijs et al., Reference Klijs, Scholtens, Mandemakers, Snieder, Stolk and Smidt2015). Participants were recruited in waves between 2006 and 2013, with the first wave being contacted through general practitioners and the rest of the participants through already included participants and online self-registration.
The Lifelines study adheres to the principles of the Declaration of Helsinki and is in accordance with the research code of the University Medical Center Groningen (UMCG), the Netherlands. The Lifelines study was approved by the medical ethical committee of the UMCG (Ref. No. 2007/152). All participants provided written informed consent.
2.2. Participants and regional language experience
A questionnaire was designed as part of a Lifelines add-on study to investigate intergenerational transfer of the regional language and the regional language use patterns of people in the north of the Netherlands (Buurke et al., Reference Buurke, Bartelds, Knooihuizen and Wieling2023). An overview of the regional language experience factors assessed in the questionnaire is available under Materials at our entry in the Open Science Framework (OSF) repository (https://doi.org/10.17605/OSF.IO/423WR). The questionnaire was sent out through email in November 2021 by the Lifelines team to all active Lifelines participants aged 18 years or older (n = 131,627) and was completed by approximately 38,500 participants (i.e., a response rate of 29%).
A flowchart detailing the sample selection is displayed in Figure 1. Based on the inspection of the histogram of the sample’s age distribution, participants were excluded if they were younger than 18 or older than 80 at the second in-person assessment (between 2014 and 2018) to prevent the few observations outside this age range from skewing the results (0.8% of the data). Following Fredrickson et al. (Reference Fredrickson, Maruff, Woodward, Moore, Fredrickson, Sach and Darby2010) and recommendations by Lifelines, we excluded observations for the cognitive outcomes based on the criteria set for completion and response integrity (see Supplementary information on OSF).
Flowchart of the sample selection procedure.
The final sample comprised 17,567 participants aged 20–80 (M age = 51.80, SDage = 11), of whom 7,448 indicated to be able to speak Frisian and 10,114 indicated to be able to speak Low Saxon. Dutch is the standard language in the Netherlands, and high proficiency in Dutch is a prerequisite for participation in the Lifelines study. It can therefore be assumed that all speakers of a regional minority language in the northern region also fluently speak Dutch. There was variation in participants’ self-reported use of, exposure to, and proficiency in the regional language, but all participants in the final sample indicated that they could have at least simple conversations in their regional language variety. Although the use of Frisian and Low Saxon is not strictly bound to province borders, in our sample, all Frisian–Dutch bilinguals lived in Fryslân, and all Low Saxon–Dutch bilinguals lived in Groningen or Drenthe at the time of administration of the regional language questionnaire.
For this study, we used the concept of language entropy (Gullifer & Titone, Reference Gullifer and Titone2020), a continuous measure of linguistic diversity and uncertainty, to calculate bilingual engagement, based on regional language use per interactional context. Participants were prompted to rate their frequency of regional language use in a range of situations (e.g., at home, at work, on the street; 12 contexts in total) on a 5-point Likert scale (1 = never; 2 = from time to time; 3 = approximately half of the time; 4 = most of the time; 5 = always). When a participant selected “never,” this was taken as the participant predominantly using Dutch in this context, assigning a proportion score of 1 to the use of Dutch and a score of 0 to regional language use. Likewise, when the answer option “approximately half of the time” was selected, we assigned a score of 0.5 to both use of Dutch and use of the regional language (for a similar procedure, see Wagner et al., Reference Wagner, Bekas and Bialystok2023). Language use scores for Dutch and the regional language per context thus reflected a proportion ranging from 0 to 1, where 0 indicated no use of the language in question and 1 exclusive use of the respective language.
We used the proportion of language use to calculate language entropy in the home, at school, at work, on the street, in shops, at the sports club, and in bars/cafés/restaurants, on the premise that individuals engage in these contexts at least somewhat regularly. Language entropy is calculated using the following formula (Shannon, Reference Shannon1948; Gullifer & Titone, Reference Gullifer and Titone2020):
$$ Entropy=-{\sum}_{i=1}^n{P}_i{\log}_2\left({P}_i\right). $$
In this formula, the number of possible languages within the interactional context is represented by n, and Pi is the proportion of the use of language i in that context. We arrived at a bilingual engagement score by averaging the individual language entropy scores for the aforementioned contexts for each participant, taking into account whether the contexts applied to them (e.g., not all participants went to school or worked). In a bilingual sample, entropy values range from 0 to 1, where 0 represents low overall bilingual engagement (i.e., using a single language across contexts) and low uncertainty of which language is going to be used, and 1 high overall bilingual engagement (i.e., perfectly balanced engagement with two languages) and high uncertainty of which language is going to be used. The calculation of bilingual engagement is available in our data preprocessing script published in Data preparation in our OSF repository.
Participants were also asked to rate their proficiency in the regional language in comparison to Dutch on a 7-point Likert scale (1 = much worse; 2 = worse; 3 = somewhat worse; 4 = equally good; 5 = somewhat better; 6 = better; 7 = much better). This Likert scale was converted to a proportion score ranging from 0 to 1, a value of 0 indicating most proficient in Dutch, 0.5 equal proficiency in both languages, and 1 most proficient in the regional language.
All participants also indicated their current overall frequency of regional language use and exposure on a Likert scale (1 = never; 2 = a few times per year; 3 = (approximately) monthly; 4 = (approximately) weekly; 5 = daily). As can be seen in Figure 1, we excluded participants who indicated that they were never exposed to, nor spoke, their regional language. Age of acquisition was included in the questionnaire for the regional language, but not for Dutch, so participants’ onset of bilingualism was unclear. The order and timing of acquisition of the regional language and Dutch is known to vary considerably in these communities (for Fryslân: Dijkstra et al., Reference Dijkstra, Kuiken, Jorna and Klinkenberg2016), but it can be assumed that children growing up with the regional language in the home started acquiring Dutch around the ages of 4–6 at the latest, when starting primary education.
The descriptive statistics of relative language proficiency, overall regional language use and age of onset of regional language acquisition are reported in Table 1. The majority of our sample reported having started learning the regional language before the start of primary school (84%), and approximately half of the sample claimed to be equally proficient in Dutch and the regional language. In addition, because relative language proficiency was comparable between Frisian–Dutch and Low Saxon–Dutch bilinguals, and because we were mainly interested in language use patterns as measured by bilingual engagement, we did not include relative language proficiency and overall frequency of use in our analysis.
Descriptive statistics of demographic characteristics and regional language experience, stratified by bilingual group
Note: Percentages are rounded. Lower educational attainment: no education, primary education, lower or preparatory secondary vocational education, and junior general secondary education. Middle educational attainment: secondary vocational education or work-based learning pathway, senior general secondary education, and pre-university secondary education. Higher educational attainment: higher vocational education and university education. Bilingual engagement reflects the average language entropy in the home, at school, at work, at the sports club, on the street, in shops, and in bars/cafés/restaurants. Bilingual engagement values occurred across the entire possible range (0–1).
a t-test unless otherwise specified.
b Chi-square test.
2.3. Cognitive measures
In the second assessment period of the Lifelines study (2014–2018; the assessment closest to administration of the questionnaire), cognitive functioning was assessed with the Cogstate Brief Battery. It comprises four computerized tasks assessing the cognitive domains of processing speed, attention, working memory and recognition memory. In all four tasks, participants are presented with playing cards that appear, one by one, in the center of the screen. The tasks are briefly described below, in the order in which they were administered (for more details, see Fredrickson et al., Reference Fredrickson, Maruff, Woodward, Moore, Fredrickson, Sach and Darby2010; Kuiper et al., Reference Kuiper, Oude Voshaar, Verhoeven, Zuidema and Smidt2017).
2.3.1. Detection task (processing speed)
The Detection task served to measure psychomotor functioning and processing speed. In this task, participants were instructed to press a key on the keyboard corresponding to “Yes” as soon as the playing card in the center of the screen turned face up. The task ended after the participant had completed 35 correct trials. The primary outcome for the Detection task was reaction time in milliseconds for correct responses, and a faster reaction time indicates better performance.
2.3.2. Identification task (attention)
In the Identification task, which measured visual attention, participants were asked to press a key corresponding to “Yes” if the playing card in the center of the screen was red, and a key corresponding to “No” if it was not. In order for the task to end, 30 correct responses were required. Reaction time in milliseconds for correct responsesFootnote 2 was the main outcome, and a faster reaction time indicates better performance.
2.3.3. One Back task (working memory)
Working memory and attention were assessed by means of the One Back task, in which participants were instructed to press the “Yes” key if the playing card in the center of the screen was the same as the one from the previous trial, and the “No” key if this was not the case. The task ended after 30 correct trials. The outcome that was used in the analyses was reaction time in milliseconds for correct responses, and a faster reaction time indicates better performance.
2.3.4. One Card Learning task (recognition memory)
The final task, the One Card Learning task, measured visual learning and recognition memory. Six cards were randomly drawn from the deck and reappeared throughout the task, alternating with non-repeating cards. Participants were required to press the “Yes” key if they had seen the playing card in the center of the screen before, and press “No” if this was not the case. Participants had to complete 42 trials before the task ended; unlike for the other tasks, there was no criterion for the number of correct answers. The proportion of correct answers (i.e., hit rate) was the main outcome resulting from this task. A higher hit rate indicates better performance.
2.4. Statistical analysis
Generalized additive mixed models (GAMMs) were built in RStudio (R version 4.2.2.; R Core Team, 2022) using the bam function from the “mgcv” package (Wood, Reference Wood2017) to assess the potential nonlinear relationship between bilingual engagement, age and cognitive functioning in Frisian–Dutch and Low Saxon–Dutch bilinguals. GAMMs are an extension of linear mixed models, and they add the possibility of modeling nonlinear relationships between a response variable and predictors using smoothed functions (i.e., smooths). Smooths are able to more closely resemble the underlying data than linear regression splines, while overfitting is automatically penalized. GAMM summaries report on the estimated degrees of freedom (edf), where an edf of around 1 indicates a mostly linear pattern and an edf of greater than 1 denotes a nonlinear effect (Wieling, Reference Wieling2018).
A forward stepwise model building procedure was employed to assess whether bilingual engagement and its interactions with age, task and bilingual group improved the model fit and explained additional deviance in the data over and above demographic predictors such as age, gender, educational attainment and location-related variability (modeled through a random intercept to account for variation introduced by living in a certain area; 713 locations in total). As a measure of effect size, we looked at how much additional variation in the data is explained by including a new model term. “Explained deviance” is used here, which is a generalization for both Gaussian and non-Gaussian regression models of the more widely known “explained variance,” or R 2 (Wood, Reference Wood2017). We compared the (f) restricted maximum likelihood scores of simpler models with more complicated models using the compareML function from the “itsadug” package (Van Rij et al., Reference van Rij, Wieling, Baayen and van Rijn2022). When no compareML result was available (due to the difference in degrees of freedom between models being too large), we relied on the change in Akaike’s Information Criterion (AIC; Cavanaugh & Neath, Reference Cavanaugh and Neath2019). If the more complex model had an AIC decrease of more than 2 in comparison to the simpler model, the additional predictor was retained (following Wieling, Reference Wieling2018). In addition, we set the criterion for significance level of predictors and interaction terms at α < .001, considering our very large sample size (Lantz, Reference Lantz2013; Faber & Fonseca, Reference Faber and Fonseca2014). The model summary in Section 3.2 presents the results from the best-fitting model, including our predictor of interest, bilingual engagement. Additional information about the statistical analysis is presented in Supplementary information on OSF.
3. Results
3.1. Sample characteristics
Descriptive statistics of demographic and regional language experience variables, stratified by bilingual group, are available in Table 1. Compared with the Frisian–Dutch group, the Low Saxon–Dutch group was older, comprised a lower proportion of women, and had lower levels of educational attainment. This is largely in line with sociodemographic factors that have previously been associated with speaking Low Saxon, as discussed in Schmeets and Cornips (Reference Schmeets and Cornips2022). Bilingual engagement was weakly positively correlated with relative proficiency in the regional language (Spearman’s ρ = 0.16), and there was a negligible correlation with the age of acquisition of the regional language (Spearman’s ρ = −0.03), demonstrating that there was no strong relationship between the knowledge and diversity in use of the regional minority and majority languages in our sample (cf. Gullifer & Titone, Reference Gullifer and Titone2020). Bilingual engagement scores were moderately correlated with the overall use of the regional language (Spearman’s ρ = 0.28), suggesting that higher engagement with the two languages reflects more frequent speaking of the regional language overall. Lastly, we found a remarkable difference between the bilingual groups in terms of how they perceive their bilingual status, substantiating the anecdotal evidence presented in Bosma et al. (Reference Bosma, Knooihuizen, Blom and van Koppen2023): the majority of Frisian–Dutch bilinguals (85%) saw their variety of Frisian as a separate language from Dutch, whereas 86% of Low Saxon–Dutch bilinguals considered their variety of Low Saxon to be a variety of Dutch. This difference may be related to the different degree of political recognition of the two regional languages.
3.2. Model results
The fixed-effect and random-effect coefficients for the optimally specified GAMM, including bilingual engagement, are reported in Table 2.
Model output for the optimally specified GAMM, including bilingual engagement
Note: GAMM formula: bam(value ~ task × gender + task × educational attainment + s(age, by = task) + s(bilingual engagement) + bilingual group + s(location, bs = “re”), data = dat, family = scat()). The values in bold indicate a significance level of α < .001. Detection = processing speed, Identification = attention, One Back = working memory, One Card Learning = recognition memory. Reference level task: Detection. Reference-level gender: woman. Reference-level bilingual group: Frisian–Dutch bilinguals.
The model summary shows that, in comparison with the performance on the Detection task, measuring processing speed, participants perform significantly worse on the Identification (attention), One Back (working memory) and One Card Learning (recognition memory) tasks. The addition of the smooth for age explained 6.40% additional deviance. Age also significantly interacted with the task, such that on each of the four tasks, performance decreased with increasing age but with slightly different trajectories (see Figure 2). The fact that the edfs of the smooth interactions are higher than 1 indicates that the effect of age on each task is nonlinear.
Visualization of the interaction smooth between age and task. Higher values indicate better performance on the cognitive tasks. The shaded areas reflect the 95% confidence intervals.
Our other demographic control predictors, gender and educational attainment, also explained additional deviance in the underlying data (0.06% and 0.69%, respectively). They also significantly interacted with the task, suggesting that the effects of gender and educational attainment differed in magnitude and/or direction per cognitive task. Men had slightly faster processing speed than women, but attention performance was comparable between men and women. Furthermore, men outperformed women on working memory, but women performed better than men on recognition memory. Higher educational attainment was associated with better performance on each cognitive task. The effect of educational attainment was comparable for processing speed and attention, but it had a stronger effect on working memory and recognition memory performance. Visualization of the effects for these demographic variables is available under Data analysis on OSF.
Adding our predictor of interest, bilingual engagement, to the model specified so far (i.e., containing age, gender, educational attainment and their interactions with task) explained 0.01% additional deviance. The effect did not meet our significance criterion (p = .002), and model comparison also showed that adding bilingual engagement did not significantly improve the simpler models’ fit. The interaction between bilingual engagement and task was tested, but it did not explain additional deviance in the model. As such, the minimal effect of bilingual engagement was similar for each cognitive function that was measured, such that cognitive performance slightly decreased with higher levels of bilingual engagement. Furthermore, the interaction between age and bilingual engagement also did not meet our significance criterion (p = .009) and also did not explain additional variance in comparison to the simpler model. This suggests that the effect of bilingual engagement, which was already minimal, did not significantly differ in magnitude across the adult lifespan.
The predictor bilingual group explained 0.16% additional deviance. The main effect of the bilingual group was such that Low Saxon–Dutch bilinguals outperformed Frisian–Dutch bilinguals on each of the cognitive tasks to a similar degree (see Data analysis on OSF). In line with the second subquestion of our study, we tested whether the effect of bilingual engagement differed for the two bilingual groups by adding an interaction between bilingual engagement and bilingual group to the model. This interaction explained 0.01% additional deviance, but model comparison showed that the addition of this interaction did not significantly improve the simpler models’ fit. Therefore, the interaction was not included in subsequent models.
As a final step, we added the random intercept for location-related variability, which explained 0.92% additional deviance. This suggests that some deviance in the data could be explained by variables that we did not account for, related to participants’ specific town or city of residence.Footnote 3 The addition of this random intercept rendered the p-value of bilingual engagement to >.05. This shows again that the effect of bilingual engagement explained only very little additional variation in the data on top of the deviance explained by the random intercept for location-related variability and the main effect of the bilingual group.
Further exploration of the significant difference in cognitive performance between Low Saxon–Dutch bilinguals and Frisian–Dutch bilinguals showed that residents in Groningen and Drenthe (the Low Saxon language areas) outperformed residents of Fryslân, regardless of knowing a regional language or not (see Data analysis on OSF). As such, we cannot univocally conclude that the difference we found between the two speaker groups is related to factors associated with bilingual experience, including but not limited to attitudes toward their bilingual status, the typological distance between the regional language and Dutch or the overall presence of the regional language in the two language contexts.
4. Discussion
Our study investigated the relationship between the degree of engagement with two typologically close languages and cognitive aging, using a large sample of Frisian–Dutch and Low Saxon–Dutch bilinguals drawn from the Lifelines Cohort Study. For this purpose, we relied on the Cogstate Brief Battery data available in Lifelines, comprising four tasks measuring processing speed, attention, working memory and recognition memory, respectively. We expected greater bilingual engagement to mitigate the decline in cognitive functions observed with advancing age, based on the premise that greater engagement with two languages, and thereby greater experience with managing language-related uncertainty, should contribute to a protective mechanism for cognitive decline (Green & Abutalebi, Reference Green and Abutalebi2013; Bialystok, Reference Bialystok2021a; Gullifer & Titone, Reference Gullifer and Titone2021a). In addition, we explored the possibility that the effects of bilingual engagement differed for Frisian–Dutch and Low Saxon–Dutch bilinguals, considering differences in sociopolitical status and general presence of the regional language between the respective sociolinguistic landscapes.
Model comparison demonstrated that bilingual engagement had no additive predictive value in our models. Our findings, therefore, suggest that bilingual engagement may not exert much influence on processing speed, attention, working memory and recognition memory performance, especially when also taking into account major factors associated with cognition such as biological age and educational attainment. Note that we found no effect for our specific sample of Frisian–Dutch and Low Saxon–Dutch bilinguals, so our findings may not necessarily generalize to other regional minority language contexts or bilingual groups.
A strength of our study is that, because of its large sample size, we could reliably determine the effect size of bilingual engagement in regional minority language speakers. Our results corroborate the findings by Gullifer and Titone (Reference Gullifer and Titone2021b), who conducted a machine-learning study examining the relationship between proactive and reactive control and bilingual engagement as measured by language entropy in a large sample of your adult bilinguals. The authors observed a similar predictive accuracy between a model containing language entropy and other bilingual experience factors and the most parsimonious model not containing these predictors, suggesting that the effect on proactive control may not generalize to other datasets of bilingual participants. We replicate this finding by showing no additive predictive value of bilingual engagement on cognition when accounting for age, educational attainment and location-related variability. If the effect of bilingual engagement is indeed small in proficient (regional minority language) bilinguals, then studies with smaller samples are unlikely to reliably detect statistical significance of such effects. The small size of the effect could partly explain why the evidence for behavioral effects in the literature has been mixed so far (Leivada et al., Reference Leivada, Westergaard, Duñabeitia and Rothman2021). We concur with Gullifer and Titone that, in order to be able to consistently and reliably find cognitive effects of bilingual experience, investigations with sufficiently large sample sizes and validated measures of bilingual experience are needed.
Furthermore, because middle-aged adults were – uniquely for the bilingualism literature – well represented in our dataset, we were able to assess the potential nonlinear relationship between bilingual engagement and cognition across the adult lifespan. Even though it has been previously suggested that bilingual experience can mitigate age-related cognitive decline (Gallo et al., Reference Gallo, Myachykov, Shtyrov and Abutalebi2020, Reference Gallo, DeLuca, Prystauka, Voits, Rothman and Abutalebi2022; Bialystok, Reference Bialystok2021a; Dash et al., Reference Dash, Joanette and Ansaldo2022), our results demonstrate that the effect of bilingual engagement, however minimal, was similar across the entire age range of our sample. A higher degree of bilingual engagement thus did not make a significant contribution to better-preserved behavioral performance in the face of more expected neural deterioration with advancing age in our regional minority language speakers. Considering that our study is one of the few investigations into the effects of bilingual engagement on cognitive functioning across the adult lifespan, we recommend examining cognitive aging trajectories also in future work on bilingualism and neurocognitive aging.
Embedding our results in the small literature on cognition in regional minority language bilingualism, our finding that the degree of bilingual engagement was hardly predictive of cognitive performance is also in line with previous work finding no differences in cognitive functioning between groups of regional minority language bilinguals with varying levels of bilingual engagement or monolinguals (e.g., Kirk et al., Reference Kirk, Fiala, Scott-Brown and Kempe2014; Scaltritti et al., Reference Scaltritti, Peressotti and Miozzo2017). However, it is not consistent with other studies that have found superior performance for bidialectals (with higher bilingual engagement) in younger adults (Poarch et al., Reference Poarch, Vanhove and Berthele2019; Garraffa et al., Reference Garraffa, Obregon, O’Rourke and Sorace2020; Antoniou & Spanoudis, Reference Antoniou and Spanoudis2025) or middle-aged and older Frisian–Dutch speakers (Houtzager et al., Reference Houtzager, Lowie, de Bot, Filipović and Pütz2014, Reference Houtzager, Lowie, Sprenger and Bot2017). This implies that the absence of an effect may be specific to (the way in which we examined) the regional language speaker groups in our study (for further discussion, see below). It is well known that this field is characterized by the intricate interactions between the multifactorial phenomena of bilingual experience, cognition and aging (Dash et al., Reference Dash, Joanette and Ansaldo2022). As such, we acknowledge that the fact that we found no robust relationship between bilingual engagement, cognitive functioning and advancing age in regional minority language bilinguals could be due to several other (methodological) factors, apart from effect size and the bilingual engagement measure alone.
The first potential explanation for the absence of a bilingual engagement effect on cognition in our study is that the cognitive tasks from the Cogstate Brief Battery, measuring processing speed, attention, working memory and recognition memory, are perhaps less suitable to capture effects related to bilingual engagement. This could be because (1) the cognitive functions measured by these specific tasks do not reflect the cognitive processes involved in managing language-related uncertainty (Gullifer & Titone, Reference Gullifer and Titone2021a) or (2) these tasks were not challenging enough to reliably show differences in behavioral performance as a function of bilingual engagement (DeLuca et al., Reference DeLuca, Segaert, Mazaheri and Krott2020; Anderson et al., Reference Anderson, Grundy, Grady, Craik and Bialystok2021; Dash et al., Reference Dash, Joanette and Ansaldo2022). The Cogstate Brief Battery can be said to be responsive to broader measures of bi-/multilingual experience (Van den Berg et al., Reference van den Berg, Brouwer, Loerts, Knooihuizen and Keijzer2025), but it may lack the sensitivity to capture finer-grained individual variation in attentional control potentially related to bilingual engagement in regional minority language speakers. Engagement in greater contextual language diversity is theorized to rely more heavily on proactive control processes, such as goal maintenance and conflict monitoring, relative to more reactive processes (Gullifer & Titone, Reference Gullifer and Titone2021a). Tasks like the AX-CPT or cued switching tasks are well suited to measure these dynamics. Such tasks may be more suitable to capture the nuanced cognitive effects of bilingual engagement – including those related to language entropy – than the Cogstate Brief Battery, which seems to rely on more automated cognitive processes. Alternatively, systematically varying the task difficulty – such as increasing the n in the n-back task (Anderson et al., Reference Anderson, Grundy, Grady, Craik and Bialystok2021; Bialystok & Craik, Reference Bialystok and Craik2022) – could also reveal individual differences in attentional control. Bilinguals with higher bilingual engagement and potentially greater attentional control are expected to continue to perform well on more challenging versions of the task. Incorporating such tasks and manipulating task difficulty in future research could provide a more comprehensive understanding of the relationship between attentional control and bilingual engagement in bilinguals more broadly and regional minority language speakers more specifically.
However, previous work using language entropy to gauge bilingual engagement in adults has shown that adaptations are sometimes found in structural and functional brain indices but not in behavioral correlates of cognitive functioning (Van den Berg et al., Reference van den Berg, Brouwer, Tienkamp, Verhagen and Keijzer2022; Voits et al., Reference Voits, Rothman, Calabria, Robson, Aguirre, Cattaneo, Costumero, Hernández, Puig, Marín-Marín, Suades, Costa and Pliatsikas2023). As such, a second possibility is that higher bilingual engagement by cognitively healthy regional minority language speakers primarily affects neural organization in the long term, which may not be directly detectable in behavior. For example, higher bilingual engagement individuals could perform on par with lower engagement bilinguals, but recruit fewer neural resources to achieve this level of performance, and/or display better-preserved structural integrity with advancing age (see discussion in Anderson et al., Reference Anderson, Grundy, Grady, Craik and Bialystok2021; DeLuca et al., Reference DeLuca, Segaert, Mazaheri and Krott2020). We can only speculate on this matter, however, considering the absence of structural and functional brain measurements in our study and the nature of the Lifelines data we relied on more generally.
Lastly, we turn to discussing the implications for the field of our observations from investigating Frisian–Dutch and Low Saxon–Dutch bilinguals specifically. Crucially, we not only directly compared bilingual groups, but also captured individual variation within these groups using language entropy, which is a validated and continuous measure of bilingual engagement informed by recent theories and developments in the field treating bilingualism as a spectrum of experiences (Green & Abutalebi, Reference Green and Abutalebi2013; DeLuca et al., Reference DeLuca, Segaert, Mazaheri and Krott2020; Gullifer & Titone, Reference Gullifer and Titone2020). We found that the mean bilingual engagement values were relatively comparable between the two groups, with slightly lower levels for Low Saxon–Dutch bilinguals. This shows that bilingual language use in Low Saxon–Dutch bilinguals was somewhat more compartmentalized, which was to be expected based on previous work (Schmeets & Cornips, Reference Schmeets and Cornips2022). To reiterate, no significant association between bilingual engagement and cognition was found across our two groups of regional minority language speakers. Although our study examined cognitive functions from a different perspective than other studies investigating the neurocognitive effects of language entropy have done (i.e., proactive and reactive control), this result does suggest that the effects ensuing from language entropy may not be universal across sociolinguistic contexts (cf. Wagner et al., Reference Wagner, Bekas and Bialystok2023). This further emphasizes the need to detail sociolinguistic and sociopolitical factors that may influence language entropy effects in future studies, in order to enhance our understanding and further develop the theoretical aspects underlying this measure. Another possibility is, however, that language entropy may not reliably measure the same aspect of bilingual language use for different groups on the bilingualism spectrum. The question arises, then, whether language entropy as a quantification of bilingual engagement is an appropriate metric to use for assessing the effects of language use patterns specifically in the wide range of diglossic and diaglossic contexts we see today (Auer, Reference Auer, Delbecque, van der Auwera and Geeraerts2005).
4.1. Language control and attentional control in bidialectism/diaglossia
A standardized methodology to measure bilingual experience in diaglossic speakers is currently lacking, and, as such, we relied on language entropy as a bilingual engagement measure validated in a societal bilingual sample (Gullifer & Titone, Reference Gullifer and Titone2020). It is possible that the Low Saxon–Dutch bilinguals, of whom 86% perceived Low Saxon as a variant of the standard language, interpreted the questions related to regional language use across contexts differently than the Frisian–Dutch bilinguals did, of whom 85% perceived Frisian and Dutch as two separate languages. Although we did not directly assess specific language use aspects related to diaglossia, it is likely that the degree of diaglossia is reflected in precisely this distinction between the two groups, in conjunction with a smaller linguistic distance between Low Saxon and Dutch than Frisian and Dutch (Buurke et al., Reference Buurke, Sekeres, Heeringa, Knooihuizen and Wieling2022). Because the language entropy measure highly depends on the proportion of use of two “separate” languages, bilingual engagement did perhaps not capture the bilingual language use patterns that we intended to measure, especially in the Low Saxon–Dutch bilinguals. Estimating how much the regional language is spoken may be particularly challenging for speakers who see their regional language as a variety of the standard language, because there is no clear boundary for them where the use of their regional language (labeled by Low Saxon and Frisian) starts and ends. Possibly, the distinction between Dutch and the regional minority language is clearer for speakers when they primarily use Dutch across contexts (proxied by lower bilingual engagement values). When the use of the regional language increases, this boundary becomes more blurred.
Therefore, it could be the case that higher values of bilingual engagement do not reflect a higher degree of switching between languages with interlocutors, but a higher degree of mixing the regional language with Dutch. This would also match the description of diaglossia in Auer (Reference Auer, Delbecque, van der Auwera and Geeraerts2005), who states that diaglossia is characterized by code-switching within turns and frequent code-mixing of the regional minority language and majority language, with fewer regional language features mixed in more formal conversational contexts. If it is indeed the case that higher bilingual engagement values reflect a higher degree of code-mixing, rather than switching between languages, then our null result for bilingual engagement would, in fact, be more in line with predictions for attentional control (Green & Abutalebi, Reference Green and Abutalebi2013). Code-mixing contexts offer speakers the possibility of using both language varieties in a flexible and cooperative way, whereas in compartmentalized contexts, it is more likely that a single language is used. As such, code-mixing and more compartmentalized language use incur fewer attentional control demands than dual language use, simply because there is less uncertainty about when to use which language with whom in these situations. This would explain the negligible difference we observed in cognitive performance measured by the Cogstate tasks for lower versus higher bilingual engagement individuals. Still, regional language speakers need to determine with whom they can or want to use certain regional language features (Ghyselen & Van Keymeulen, Reference Ghyselen and van Keymeulen2016), implying that there is a degree of cue detection and inhibition of regional language features involved in their language control. However, because no specific items in our questionnaire targeted language mixing patterns, and because of the a priori difference we found between residents of Fryslân versus Groningen and Drenthe, we cannot ascertain whether (perceived) language separation or language mixing patterns play a role in the attested effects. Future research should therefore assess bilingual language use on different dimensions, taking into account overarching interactional contexts, individual degree of engagement with two languages in the same conversational context, as well as to what degree these languages are then mixed.
Our study also has implications for the larger discussion on how to assess language experience in diaglossia, a type of bilingualism on the continuum that has received little attention in the field so far, in future studies. A recent paper found that the available standardized questionnaires to assess language experience across the bilingualism spectrum only overlap for 15% or less in terms of content (Dass et al., Reference Dass, Smirnova-Godoy, McColl, Grundy, Luk and Anderson2024). This suggests that the language experiences captured across questionnaires – and therefore across studies – differ considerably. The authors recommend carefully selecting questionnaires depending on the purpose or bilingual population of the intended study, rather than devising a questionnaire that captures all constructs identified across the questionnaires. Importantly, however, we do not currently know if these questionnaires sufficiently fit bilingual experience in diaglossia (Poarch et al., Reference Poarch, Vanhove and Berthele2019), where the boundary between the regional language variety and the majority language is often less clear. Furthermore, Kirk et al. (Reference Kirk, Declerck, Kemp and Kempe2022) argue that regional minority language speakers would likely not list the regional minority language variety as a separate language in a questionnaire when they perceive this language as a variant of the standard language, even though these language varieties may also compete for activation, perhaps similarly to two more typologically distant languages. This is problematic because regional language speakers might be classified as monolinguals in some studies, even though they might also need to engage in bilingual language control to some extent.
Although developing measurements to accurately describe language experience in diaglossia is important, it is perhaps best to first take a step back and continue with further explorations of how bilinguals in bidialectal and diaglossia contexts control their available languages or varieties through psycholinguistic experiments. The current evidence for language control in bidialectism shows that the regional variety is not fully processed or accessed as part of the standard language and that the regional and standard variety may compete for activation (Kirk et al., Reference Kirk, Kempe, Scott-Brown, Philipp and Declerck2018, Reference Kirk, Declerck, Kemp and Kempe2022; Vorwerg et al., Reference Vorwerg, Suntharam and Morand2019; Declerck & Kirk, Reference Declerck and Kirk2023). However, sociolinguistic factors related to how these languages are represented in bidialectals’ language repertoire seem to play an important role, such as attitude toward the regional variety and language dominance. Thus, also considering that other work has shown that bidialectal words may be processed like synonyms of the standard language (Melinger, Reference Melinger2018, Reference Melinger2021), the issue of whether bidialectal lexical access and language control rely on similar mechanisms as (societal) bilingualism remains far from settled. Because the ideas underlying possible neurocognitive adaptations associated with bilingual experience heavily rely on the interplay between language control and attentional control, such investigations also have implications for the effects that are observed in regional minority language bilingualism.
4.2. Limitations
Our investigation shows no significant association between bilingual engagement and cognitive functioning as measured by the Cogstate Brief Battery across the adult lifespan in our sample of regional minority language speakers. Despite the lack of an effect, we want to note that our retrospective cohort study could have determined a correlational relationship at most. Although we carefully controlled for individual differences in educational attainment, as well as variance pertaining to the specific place of residence, the results we observed could also be due to generational or age cohort differences in our sample (Salthouse, Reference Salthouse2019) in using regional languages and their associated prestige by the speakers themselves. It is also possible, especially in the older participants in our sample, that current patterns of regional language use and exposure (as assessed in our questionnaire) do not match language use patterns in earlier life (Dash et al., Reference Dash, Joanette and Ansaldo2022). For example, it is more than likely that language use patterns change over the lifespan as a result of significant life events, such as retirement (e.g., an increased proportion of regional language usage after retirement, because the majority language was predominantly used at work). Furthermore, many older participants in our sample may have grown up with the regional language in the home, but did not transfer the regional language to their children, leading to different regional language usage patterns in adulthood (Buurke et al., Reference Buurke, Bartelds, Knooihuizen and Wieling2023). Therefore, future studies would do well to take a lifespan approach in relating usage experience to neurocognitive adaptations in aging. Although our results could be used to yield hypotheses about bilingual engagement and typical cognitive aging trajectories in regional minority language speakers, longitudinal investigations are needed (DeLuca & Voits, Reference DeLuca and Voits2022; Voits et al., Reference Voits, DeLuca and Abutalebi2022).
4.3. Conclusion
To conclude, our study’s findings contribute to the rapidly expanding literature on bilingual experience and cognitive aging. Specifically, by using a large sample of regional minority language speakers in the north of the Netherlands, we have shown little support for a relationship between bilingual engagement and processing speed, attention, working memory and recognition memory in Frisian–Dutch and Low Saxon–Dutch bilinguals. We posit that any cognitive effects associated with bilingual engagement are likely modest when juxtaposed with more influential factors such as age, educational attainment and location-related variability. Still, our novel investigation into the cognitive aging effects of bilingual engagement in Frisian–Dutch and Low Saxon–Dutch bilinguals has raised many more questions that deserve further attention. Our study highlights the need for a careful characterization of language control and (neuro)cognitive adaptations associated with that in regional minority language speakers across different geographic locations and sociolinguistic contexts, especially in diaglossia.
Supplementary material
The supplementary material accompanying this paper, including materials, data preparation and analysis code, additional information on the methodology, and additional figures, is openly available at our Open Science Framework entry: https://doi.org/10.17605/OSF.IO/423WR.
Data availability statement
The data catalog of Lifelines is publicly accessible on https://data-catalogue.lifelines.nl. These data may be obtained from a third party and are not publicly available. Researchers can apply for the data used in this study at the Lifelines research office (research@lifelines.nl). More information about how to request Lifelines’ data and the conditions of use can be found on their website (https://www.lifelines-biobank.com/researchers/working-with-us/step-1-prepare-and-submit-your-application).
Acknowledgements
The authors wish to acknowledge the services of the Lifelines Cohort Study, the contributing research centers delivering data to Lifelines, and all the study participants.
Funding statement
The Lifelines initiative has been made possible by subsidy from the Dutch Ministry of Health, Welfare and Sport, the Dutch Ministry of Economic Affairs, the University Medical Center Groningen (UMCG), Groningen University and the Provinces in the north of the Netherlands (Drenthe, Fryslân, Groningen). Open access funding provided by University of Groningen. This work was supported by a personal grant awarded to Merel Keijzer (NWO Vidigrant: 016.Vidi.185.1906208). The Groninger Centrum voor Taal en Cultuur supported the development of the regional language experience questionnaire.
Competing interests
The authors declare none.
Open access
