Extramural English research

The focus of this paper is specifically on English as a target language, and the concept of extramural English (EE) will be used. EE encompasses both intentional and incidental informal learning of English through learner-initiated activities that can take place either online or in real life (Sundqvist, Reference Sundqvist2009; Sundqvist & Sylvén, Reference Sundqvist and Sylvén2016). Engagement in activities occurs outside (from the Latin word “extra”) the walls (from the Latin word “mural”) of the classroom and is not connected with school(ing). Some typical extramural activities are listening to music in English, watching English-medium television/films, reading books in English, and playing digital games using L2 English. There is a great variety of types of EE exposure or input, and some can be pedagogical in nature, although technically, they are examples of informal learning (e.g., Instagram accounts such as @englishgeekz and @englishlifewithbenjamin). Despite being a contradiction in terms, some individuals may also engage in EE inside the classroom – at least in some settings. For example, during English lessons, students may view English-language YouTube videos on their own laptops when they should be doing something else.

Engagement is a fundamental concept in EE research. It encompasses (at least) three interconnected dimensions: behavioral, cognitive, and affective/emotional engagement (Fredricks et al., Reference Fredricks, Filsecker and Lawson2016). Schmitt (Reference Schmitt2008) has specifically applied the concept of engagement to L2 vocabulary learning. He emphasizes the importance of learners’ self-regulation in the learning process, asserting that increased exposure, attention, manipulation, and time dedicated to lexical items contribute to learning. Schmitt’s reasoning can be extended to EE engagement, which is also self-regulated. Increased exposure, attention, and time invested in specific EE activities have been shown to enhance the acquisition of abilities and content knowledge associated with them. The frequency at which a learner engages in EE activities reflects behavioral engagement, and the voluntary nature of participation reflects emotional engagement. Learners tend to engage in activities they enjoy but discontinue their involvement once their emotional connection diminishes (Sundqvist, Reference Sundqvist2019). Moreover, any form of EE engagement implies cognitive engagement, as learners utilize their L2 skills. Certain EE activities necessitate higher levels of cognitive demand, particularly when interaction with others is involved (the interaction hypothesis, see, e.g., Gass & Mackey, Reference Gass and Mackey2006).

EE research is a relatively young area of applied linguistics research and finds its home in the field of SLA. The relationship between EE and vocabulary knowledge has been most extensively researched, with positive findings (e.g., De Wilde et al., Reference De Wilde, Brysbaert and Eyckmans2020a; De Wilde & Eyckmans, Reference De Wilde and Eyckmans2017; Peters, Reference Peters2018; Peters et al., Reference Peters, Noreillie, Heylen, Bulté and Desmet2019; Puimège & Peters, Reference Puimège and Peters2019; Schwarz, Reference Schwarz2020; Sundqvist, Reference Sundqvist2009). Some studies have focused on vocabulary related to a specific EE activity, such as playing digital games (e.g., Hannibal Jensen, Reference Hannibal Jensen2017; Sundqvist, Reference Sundqvist2019; Sylvén & Sundqvist, Reference Sylvén and Sundqvist2012). Further, empirical studies have consistently demonstrated the advantageous impact of audiovisual material consumption on L2 learning (e.g., Lindgren & Muñoz, Reference Lindgren and Muñoz2013; Muñoz et al., Reference Muñoz, Cadierno and Casas2018). Additionally, EE research has explored connections with writing skills (e.g., Olsson & Sylvén, Reference Olsson and Sylvén2015; Sundqvist & Wikström, Reference Sundqvist and Wikström2015) as well as with listening and reading comprehension (e.g., De Wilde et al., Reference De Wilde, Brysbaert and Eyckmans2021; Muñoz, Reference Muñoz2020; Sylvén & Sundqvist, Reference Sylvén and Sundqvist2012). However, the relationship between EE and speaking skills has received considerably less attention, with a few exceptions (see De Wilde et al., Reference De Wilde, Brysbaert and Eyckmans2021; Lyrigkou, Reference Lyrigkou2019; and Sundqvist, Reference Sundqvist2009). Furthermore, engagement in EE activities has been shown to have positive effects on cognitive and affective domains, including increased confidence (Lai et al., Reference Lai, Zhu and Gong2015) and enhanced willingness to communicate (Lee & Drajati, Reference Lee and Drajati2020). This suggests that EE can benefit learners in ways that extend beyond the actual acquisition of English language skills.

In sum, in a relatively short period of time, EE research has yielded important findings related to L2 learning and overall L2 development and shown potential to contribute to learners’ psychological attitudes toward language learning. Thus, the importance of EE as an ID variable should not be overlooked. In addition, quite recently, there has been a structural change, which will be discussed next, that puts new demands on EE research, as this change is bound to influence the methodology employed in future EE research.

A structural change

While EE research has only been around for about two decades, the phenomenon itself is not new. People have always learned languages informally. That said, access to English is abundant thanks both to technology and the fact that English is a global language (e.g., Graddol, Reference Graddol2006). To move EE research forward at this point in time, current theoretical perspectives, conceptual frameworks, and methodological approaches need to be expanded and improved (discussed further below) because of the structural change, which has to do with EE replacing lessons in school as a beginning of and basis for L2 English learning.

Supposedly, the change is a result of a widespread general increase in learners’ engagement in EE – at least as observed in the European context. For example, while an early EE study by Sundqvist (Reference Sundqvist2009) reported that 15- and 16-year-old Swedish learners, on average, spent 18.4 hours per week on EE, a decade later, Schwarz (Reference Schwarz2020) reported that learners of the same age in Austria averaged 28.9 hours per week (i.e., 10 more hours per week). This increase can largely be explained by young people’s expanded use of English-mediated media that is easily accessible via smartphones, including the emergence of numerous streaming services and more social media during the period. YouTube, launched already in 2005, has become extremely popular all over the world. For instance, as of January 2023, the YouTube audience (all ages) in the Asia–Pacific region reached approximately 467 million users (Hughes, Reference Hughes2023). Media councils/agencies in several countries report high and fairly similar frequencies of media habits among young people, such as in Austria (where German is the majority language, Saferinternet.at, 2022), Norway (where Norwegian is the majority language, Norwegian Media Authority, 2022), and Sweden (where Swedish is the majority language, Swedish Media Council, 2023). While some media consumption occurs through the majority language, a vast proportion occurs through English. As an example, in Norway, a majority of 9–18-year-olds answered that they use mainly English for viewing films/series/television (62%), gaming (63%), and YouTube (64%), whereas mainly Norwegian was used for reading, watching, or listening to news (68%) (Norwegian Media Authority, 2022).

The structural change can be illustrated with the help of the L2 English language learning pyramid that was originally introduced in Sundqvist and Sylvén (Reference Sundqvist and Sylvén2016, see Figure 1), here revised and split into two L2 English learning pyramids, one for high EE users and one for low EE users (see Figure 2).

Figure 1.

The L2 English learning pyramid (based on Sundqvist & Sylvén, Reference Sundqvist and Sylvén2016, p. 222).

Figure 2.

The revised L2 English learning pyramid split into two: high EE users (left) and low EE users (right).

In the original L2 English learning pyramid (Figure 1), EE is floating on the top, detached from the rest of the pyramid and its base. EE is described as flexible: “for some learners EE is minute, for others EE constitutes by far the largest part of their L2 English” (Sundqvist & Sylvén, Reference Sundqvist and Sylvén2016, pp. 222–223). The detachment is important as it symbolizes that EE does not have any connection with school or schooling, that is, no connection with formal (or “intramural”) learning. It should be stressed that in the original pyramid, the base consists of classroom activities. Individuals may be involved in other learning activities, too, such as evening school or extracurricular tutoring, which is very common in some countries, not least in East Asia (see, e.g., Butler, Reference Butler, Murray and Scarino2014, for China). In the revised version(s) of the L2 English learning pyramid (Figure 2), the base consists of EE instead of classroom activities. This is the actual structural change: in a very short period, EE has gone from being viewed as something extra (i.e., the floating on the top) to being viewed as something fundamental. In essence, the revised pyramids illustrate that EE is an ID variable that, for many (but not all), has replaced the classroom as the starting point and foundation for learning English. For learners who use a lot of English outside of the school context (high EE users), the base of the learning pyramid is thick (the left pyramid in Figure 2), whereas the base is very thin for low EE users (the right pyramid in Figure 2). Altogether, this means that in settings where English is easily accessible to anyone, including children, EE is an ID variable that will play a role in learning even from an early age, and as a consequence, something that both researchers and teachers will need to acknowledge and be aware of (cf. Schurz & Sundqvist, Reference Schurz and Sundqvist2022; Schwarz, Reference Schwarz2020; and Sundqvist & Sylvén, Reference Sundqvist and Sylvén2016).

The structural change proposed here, where EE replaces classroom activities in school as the starting point and foundation for L2 English learning, finds strong empirical support in studies from Flanders, the Dutch-speaking part of Belgium, where formal English instruction is not introduced until in school years 7 or 8. Studies among Flemish primary school learners have shown that children know more than 3,000 English words without having had a single English lesson in school (Puimège & Peters, Reference Puimège and Peters2019), and some already score at level A2 according to the Common European Framework of Reference (Council of Europe, 2020) for listening comprehension, writing, and speaking before even beginning formal English L2 learning in school (De Wilde et al., Reference De Wilde, Brysbaert and Eyckmans2020b). Not only does this change in exposure to English have pedagogical implications but it also affects how we research EE. Clearly, the situation calls for new tools and methods to capture and measure learners’ engagement in different EE activities. To quote Bob Dylan (Reference Dylan1964): “[t]he times they are a-changin’.”

Methodology matters

Interestingly, Skehan (Reference Skehan1991) uses formal versus informal learning as an example of how learning contexts tend to change, so in a way, he predicted the structural change discussed above. From a methodological perspective, he rightly claims that it is essential to “probe how consistently a particular relationship is found” (Skehan, Reference Skehan1991, p. 290) between different ID variables (e.g., EE) and learning outcomes (e.g., measurements of English proficiency) when contexts do not necessarily stay the same – and contexts seem to be changing rapidly. The present situation with English as a prestigious lingua franca combined with digitization on a global scale certainly makes EE engagement fertile ground for informal learning. However, a well-known problem in this line of research is how to capture learners’ contact with EE validly and reliably. Some studies, especially the early ones (including some of my own), lack sufficient accounts of validity and reliability. However, such problems have diminished somewhat as EE research has made progress, most likely thanks to general methodological advancements in the broad field of applied linguistics, including recommendations provided in publications that describe how to report and interpret quantitative and qualitative findings, as well as improved author guidelines announced by high-impact journals and publishing houses (see, e.g., Chapelle & Duff, Reference Chapelle and Duff2003; Mackey & Gass, Reference Mackey and Gass2023; Norris et al., Reference Norris, Plonsky, Ross and Schoonen2015; Paltridge & Phakiti, Reference Paltridge and Phakiti2015; Plonsky & Oswald, Reference Plonsky and Oswald2014; and Twining et al., Reference Twining, Heller, Nussbaum and Tsai2017).

As EE engagement occurs outside educational institutions, it obviously cannot be “controlled” and, thus, poses challenges in terms of methods, not least ethical ones. Compared to the 2000s, current ethical demands are much stricter due to new data protection legislation, such as the General Data Protection Regulation in Europe, implemented in 2018 (European Union, 2018). For reasons such as these, the time is ripe to address methodological challenges in studying EE and to provide suggestions for how EE research may be conducted in the future. Some guidance for the future can be found in previous work on ID variables, to which we turn next.

Hierarchical and concatenative approaches in naturalistic and confirmatory research

In his description of types of ID research, following Skehan (Reference Skehan1989), Ellis (Reference Ellis1994, pp. 474–475) distinguishes two approaches that concern the relation between theory and research: the hierarchical approach (“a theory that affords predictions about how particular IDs affect learning”; research typically involving hypotheses that are tested empirically) and the concatenative approach (“a research-then-theory approach”; research typically involving data collection followed by correlational analyses between independent variables – i.e., the IDs – and dependent variables, such as measures of L2 learning). Further, Ellis brings up two general traditions in ID research: naturalistic and confirmatory research. Research on EE started out in the naturalistic tradition, adopting a concatenative approach. It was common with exploratory studies that mapped how young people engage in various EE activities in their spare time, often in the home, and analyzed the relationship between EE data and measures of different aspects of L2 English proficiency (see, e.g., De Wilde & Eyckmans, Reference De Wilde and Eyckmans2017; Hannibal Jensen, Reference Hannibal Jensen2017; and Sundqvist, Reference Sundqvist2009). This research-then-theory approach, in comparison with the hierarchical approach, adopts a less prescriptive stance and encourages extensive data collection and the formulation of broad generalizations. However, a drawback of the approach is that the researcher risks being overwhelmed by an abundance of data, with little guidance about how to move forward. While the endpoint for an EE research design would be a theory-then-research design, current EE research appears to be in what Skehan (Reference Skehan1991) refers to as a “ground-clearing phase” (p. 296). He argues that this phase is necessary in order to reach the endpoint, and at present, exploring new ways forward in EE research is, thus, welcomed.

Methods, research instruments, and ways forward

Lee’s (Reference Lee, Reinders, Lai and Sundqvist2022) systematic review of articles published between 2010 and 2020 examines instruments utilized in L2 learning beyond the classroom. An analysis of 76 documents (covering 144 research tools) revealed that questionnaires were the most frequently used instrument, but that interviews and observations had also been used often and consistently. Less commonly used instruments or methods were language logs/diaries, group interviews, reflective journals, computer tracking, stimulated recall, and language learning history. A key reason for using questionnaires is, of course, that data can be collected from large samples, and depending on the sampling method and size, generalizations can be made about the statistical population. In EE research, many researchers have developed their own questionnaires to capture EE. This can be expected in a relatively new area of research, but the differing formats and (sometimes) lack of reports about validity and reliability make study comparison and replication difficult (Sundqvist & Uztosun, Reference Sundqvist and Uztosun2023). Since it is important to yield research results “that are robust, credible, and reproducible” (Gass et al., Reference Gass, Loewen and Plonsky2021, p. 249), this multitude of instruments is problematic for research on informal L2 learning.

In response to the problem, two new instruments have recently been developed: the Informal Second Language Engagement (ISLE) questionnaire among L2 English learners in Germany (N = 382) (Arndt, Reference Arndt2023) and the Extramural English Scale among L2 English learners in Denmark, Norway, Sweden, and Turkey (N = 907) (Sundqvist & Uztosun, Reference Sundqvist and Uztosun2023). Both instruments build on a multi-dimensional conceptualization of engagement and were developed in stages (including piloting, exploratory factor analysis, confirmatory factor analysis, and different validation procedures). High reliability is reported for both.

The ISLE captures affective, cognitive, behavioral, and linguistic aspects of learner engagement with informal L2 practices and is intended to be implemented as an online survey as part of an event-contingent experience sampling method (ESM) approach (Arndt, Reference Arndt2023). In ESM, participants are contacted several times per day to provide data right then in the moment, and this happens every day for a period of time (e.g., for a week). One benefit of this approach is that real-time experiences are reported, which means high ecological validity (Arndt et al., Reference Arndt, Granfeldt and Gullberg2023). The method minimizes the impact of recall bias and can lead to more reliable data, allowing researchers to examine the dynamics of activities and experiences over time. Further, Arndt et al. (Reference Arndt, Granfeldt and Gullberg2023) propose that ESM allows for new types of analyses because of a three-level “nested” structure of data: several responses/moments per day and several days per participant. The assumption is that “measurements are likely to be more similar if they stem from the same individual than across different participants, and if they are collected on the same day versus different days” (Arndt et al., Reference Arndt, Granfeldt and Gullberg2023, p. 43). However, as discussed by the authors, there are drawbacks too, such as the burden put on participants to frequently respond to survey questions, which risks leading to attrition and disrupting participants’ natural behavior. There is also a risk of selection bias, and there can be habitual effects or reactivity, which means that participants might become more attentive to their experiences and can potentially start changing their behaviors, consciously or subconsciously.

The second instrument, the EE Scale, captures the total frequency of learners’ EE engagement per eight factors: EE Digital Creativity, EE Gaming, EE Internalized, EE Music, EE Niche Activities, EE Reading and Listening, EE Social Interaction, and EE Viewing (Sundqvist & Uztosun, Reference Sundqvist and Uztosun2023). The 7-grade EE Scale is intended as a one-off retrospective survey and includes 32 items. Some benefits of this instrument are that it is easily administered (10 minutes, online or on paper), flexible (e.g., all items, or a selection based on factors, can be used), and sustainable. As regards its sustainability, unlike the ISLE and many other surveys tapping into informal learning, the EE Scale does not use any brand names in any questions; instead, all items are created based on learner agency and language skills connected with engagement in different activities, and whether activities are carried out alone or with others. However, a known problem with this type of questionnaire is that the answers are recall-based, and it is not possible to know how participants conceptualize the scale.

As regards the development of innovative research instruments for capturing EE, including questionnaires, new software, and apps, they appear at such a speed that any comments on them here would risk becoming outdated overnight; still, a safe recommendation would be to consider data protection carefully.

While relationships between EE on the one hand and L2 proficiency/skills/abilities or cognitive and affective domains on the other have been investigated, some relationships are less explored than others. There is a clear need for more studies on links between EE and language production—that is, speaking, including interactional competence (Salaberry & Kunitz, Reference Salaberry and Kunitz2019), and writing. Moreover, very little is known about the relationship between EE and grammar knowledge (cf. Cadierno et al., Reference Cadierno, Hansen, Lauridsen, Eskildsen, Fenyvesi, Hannibal Jensen and Aus der Wieschen2020). In addition, the role of specific EE activities in learning (e.g., gaming) and the context-specific nature of EE need further exploration, preferably in transnational studies, which are still quite rare. EE research considering cognitive and affective domains has recently grown more common, not least in a number of studies from Asia (see, e.g., Lai & Zheng, Reference Lai and Zheng2018; Lee & Dressman, Reference Lee and Dressman2018; and Lee & Taylor, Reference Lee and Taylor2022), but more such studies from other contexts would be welcome. Furthermore, some parts of the world are clearly under-represented in EE research, namely South America and Africa (for an exception, see Dressman, Reference Dressman, Dressman and Sadler2020). In addition to the need to explore new regions and regional differences, it would be valuable to consider socioeconomic or class differences as well, and the question of access to various forms of EE (including digital resources and the internet) within particular regions or countries, as EE exposure may be viewed as a global form of social and cultural capital (cf. Bourdieu, Reference Bourdieu and Brown1973).

To date, quantitative methods are frequently used in EE research, but combining quantitative and qualitative methods by adopting a mixed-methods approach also occurs (e.g., Lai & Zheng, Reference Lai and Zheng2018; Schwarz, Reference Schwarz2020; Sundqvist, Reference Sundqvist2019). Some studies are qualitative interview studies (e.g., Soyoof, Reference Soyoof2023; Sundqvist, Reference Sundqvist2015), at times ethnographically oriented (Hannibal Jensen, Reference Hannibal Jensen2019; Rothoni, Reference Rothoni2017). An advantage of interviews is that they yield in-depth data, and such data are needed to shed light on learners characterized by high, mid, and low EE engagement and why they choose (not) to be involved in EE. Popular science descriptions of such different types of learners would be very helpful in communicating results from EE research to important stakeholders. Moreover, there is room for improvement in terms of interview data elicitation techniques, such as well-structured EE interview guides. In addition, since many learners are exposed to EE from an early age, more interviews with very young learners (pre- and primary school level) would be useful, possibly even home interviews with guardians and siblings participating, too. Overall, young learners are an under-researched group, as are old learners (see problems with sampling bias, Andringa & Godfroid, Reference Andringa and Godfroid2020). It would be valuable to develop the conceptualization of EE in relation to adult and old learners since many may not spend time in educational institutions, begging the question of whether “extramural” is applicable and, if so, what it would mean.

Many quantitative EE studies are correlational, but recent work may encompass more advanced methods, such as mixed effects modeling and structural equation modeling (SEM; e.g., Zhang & Liu, Reference Zhang and Liu2022). While related, mixed effects modeling is predominantly employed for the analysis of hierarchical data structures, whereas SEM is frequently used to assess complex theoretical models and investigates the connections between observed and latent variables, which makes it possible to test cause and effect based on correlational data (Dörnyei, Reference Dörnyei2007; Field, Reference Field2013; for recent reviews of EE/informal L2 learning research, see Soyoof et al., Reference Soyoof, Reynolds, Shadiev and Vazquez-Calvo2024 and Zhang et al., Reference Zhang, Zou, Cheng, Xie, Wang and Au2021). Considering the ground-clearing phase mentioned above, EE research seems well underway in the theory-then-research design stage.