Lexical transfer as a resource in pedagogical translanguaging

ABSTRACT

‘Translanguaging’ has become the most popular term in discussions about how to use learners’ languages as resources for target language teaching/learning, but it has also become ambiguous because it is being developed in proposals adopting different perspectives on multilingualism. The first aim of this article is to offer an overview of the two main current proposals of translanguaging together: García and colleagues’ ‘spontaneous translanguaging’ and Cenoz and Gorter’s ‘pedagogical translanguaging’. ‘Spontaneous translanguaging’ argues that languages in the mind are represented as one single system and that the notion of transfer, consequently, must be rejected. This article argues that this view may not be practical because it makes it difficult for teachers to conceptualise what they are doing when raising awareness of similarities between learners’ languages. ‘Pedagogical translanguaging’, by contrast, centres around transfer. Whereas transfer is traditionally seen as unintentional ‘interference’, pedagogical translanguaging considers it a phenomenon that learners can use intentionally and creatively, and which teachers could promote by raising learners’ awareness of similarities between their languages. However, very little is known about how learners use transfer unintentionally versus intentionally. This article discusses key findings and implications from Fuster’s (2022) initial study on intentionality in lexical transfer within pedagogical translanguaging.

KEYWORDS:

• Pedagogical translanguaging
• spontaneous translanguaging
• intentional transfer
• transfer strategies
• teaching for transfer
• multilingual education

1. Introduction

This article discusses the different meanings of ‘translanguaging’ and elaborates on Cenoz and Gorter’s (e.g. 2014, 2020) theoretical proposal of pedagogical translanguaging on basis of a study on intentionality in lexical transfer. The article is based on key ideas and findings in Fuster’s (2022) doctoral monograph, which advocated the proposal of pedagogical translanguaging and investigated how multilingual school learners of Spanish employ lexical transfer with or without intention in written production, as well as their teachers’ attitudes towards a multilingual instructional approach. As will be shown, studying the difference between how learners use transfer unintentionally (i.e. not on purpose, by accident) versus intentionally (i.e. on purpose, as a conscious strategy to solve gaps of knowledge) can advance ideas about how teachers may work with learners’ multilingualism as a resource in the classroom so as to enhance target language (TL)¹ acquisition.

In recent years, ‘translanguaging’ has become the most popular term in discussions about how to move towards a multilingual instructional approach in which learners’ background languages (BLs)² would be harnessed as resources for TL acquisition, rather than considered risks for linguistic ‘interferences’. However, the term ‘translanguaging’ has also become ambiguous because its original semantic scope of a planned alternation of languages in the classroom is being developed in different theoretical proposals that adopt different perspectives on individual multilingualism and related central notions, particularly those of code-switching and transfer. To date, there have been very few publications that discuss the different proposals of translanguaging together (but see Cenoz & Gorter, 2022), and the first aim of this article, therefore, is to offer a critical overview of the two main current ones.

I adhere to the terminology of ‘spontaneous translanguaging’ and ‘pedagogical translanguaging’ as used in Cenoz and Gorter (2017a) to distinguish between the two proposals, but other terms have been used elsewhere. García and colleagues (e.g. García & Kleyn, 2016) refer to their own proposal (spontaneous translanguaging) as ‘the strong version of translanguaging’ and mention that there is also a ‘weak version of translanguaging’ (pedagogical translanguaging). Cummins (2021), meanwhile, refers the former as ‘the unitary translanguaging theory’ and to the latter as the ‘crosslinguistic translanguaging theory’.

A central argument in García and colleagues’ proposal of spontaneous translanguaging is that languages are represented as one unitary system in the mind and that the notion of transfer, consequently, must be rejected. The present article questions the practicality of rejecting the notion of transfer when wanting to move towards an instructional approach that takes advantage of learners’ BLs to enhance TL acquisition. Instead, the article advocates Cenoz and Gorter’s proposal of pedagogical translanguaging for moving towards such an approach. In contrast to spontaneous translanguaging, the notion of transfer is core to pedagogical translanguaging. Pedagogical translanguaging argues against the negative way in which transfer has traditionally been viewed and offers a conceptualisation of transfer as a phenomenon that learners can use intentionally, as a creative strategy to solve gaps of knowledge in the TL. Cenoz and Gorter propose that teachers raise learners’ awareness of linguistic similarities between their languages so as to help them find more opportunities to make use of (or transfer) aspects from their BLs creatively, as a resource. This conceptualisation of transfer as a phenomenon that learners can use intentionally has great potential for moving towards a multilingual perspective in education.

Some publications, in line with pedagogical translanguaging, have noted that transfer may not only occur unintentionally, but also intentionally as a strategy (e.g. Jarvis & Pavlenko, 2008; Jessner, 2006; Williams & Hammarberg, 1998). However, as Fuster and Neuser (2021) discuss in their overview of theoretical models of multilingual language organisation and activation, the mainstream view on transfer is still that it is an unintentional phenomenon that occurs as a by-product of the co-activation of languages in the mental lexicon and, impliedly, which learners should try to avoid. Therefore, from an empirical perspective very little has been explored about (un)intentionality in transfer and it is not clear how learners employ transfer with or without intention and how linguistic awareness or other potentially important variables may be related to (un)intentionality in transfer. Gaining insight into these questions contributes to advance the idea in pedagogical translanguaging of raising learners’ crosslinguistic awareness so as to help them control their use of transfer and enhance, thereby, their TL performance. The second aim of this article, therefore, is to discuss key findings from Fuster’s (2022) initial study on (un)intentionality in lexical transfer within the theory of pedagogical translanguaging.

The next section turns to the relatively young area of research on multilingualism in Second Language Acquisition (SLA), within which the developments of the notion of translanguaging, as well as research on transfer in multilingual learners, can be understood.

2. Research on multilingualism

Research on multilingualism in SLA emerged in the 2000s as a result of various socio-political developments during the 1980s and 1990s that were making multilingualism more visible and common. These developments include movements of language revitalisation in Europe (e.g. with Catalan and Basque in Spain), the commitment of the European Commission to promote multilingualism, the spread of English as the main global lingua franca and the rapid growth of globalisation, which has led to increasingly heterogeneous societies in both cultural and linguistic terms (e.g. Jessner, 2008; Vertovec, 2019). In the turn to the twenty-first century, all these developments led to a surge of scholar interest in multilingualism that resulted in the emergence of a broad area of studies within SLA that are concerned with multilingual language learning and use. These studies adopt different, more or less overlapping, psycho- or sociolinguistically oriented perspectives, but in a general way they can be referred to as ‘multilingualism research’ (e.g. Berthele & Udry, 2021) or ‘research on multilingualism’ (e.g. Cenoz, 2013) (see Fuster, 2022, for an overview).

While SLA has traditionally focused on monolinguals learning a chronologically first non-native (or ‘second’ or ‘foreign’) language (L2), research on multilingualism typically focuses on learners and users who already know several languages, often learners of non-first non-native languages – so-called ‘third languages’ (L3s) (Bardel & Falk, 2020; Hammarberg, 2018). Research on multilingualism represents, moreover, a paradigm shift – the so-called ‘multilingual turn’ in SLA (Meier, 2017) – as it criticises the traditional use of the competence of an idealised monolingual native speaker as the yardstick to study whether L2 learners ‘succeed’ or ‘fail’ in the new language (Cook, 1997, 2019). Traditionally, the goal of L2 acquisition in SLA has been to use the L2 like a monolingual native speaker, without influences or ‘interferences’ (i.e. transfer³) traceable to other languages (Ortega, 2019). Research on multilingualism, by contrast, aims to take multilingualism as the point of reference and study what multilinguals can do with their languages without comparing them to monolingual native speakers.

Multilingualism has long been considered a risk for linguistic ‘confusions’, but today a common notion is that the more languages you know, the easier it is to learn a new one. Influential research in the psycholinguistically oriented subarea of Third Language Acquisition (TLA) has indeed indicated that multilingual individuals seem to have a cognitive advantage for foreign language learning as compared to monolinguals or people who know and use few languages. From their greater experience with language learning, multilingual learners have been found to develop an enhanced linguistic awareness and general cognitive executive functioning (e.g. Adesope, Lavin, Thompson, & Ungerleider, 2010; Bialystok, 2007; Cenoz, 2013). Linguistic awareness can enhance language learning because it helps learners analyse language, strategically look for crosslinguistic similarities and make more linguistic guesses (e.g. Kemp, 2007; Psaltou-Joycey & Kantaridou, 2009). Additionally, more multilingual individuals can have a cognitive advantage for language acquisition because they have a wider language repertoire from which to transfer knowledge, thus getting ‘many “free rides’’’ (Cenoz & Todeva, 2009, p. 278), especially if their languages are typologically close to each other. In an extensive monograph on linguistic awareness, Jessner (2006) shows that linguistic awareness seems to help learners to consciously transfer detected similarities as a strategy for communicative purposes or to solve gaps of TL knowledge.

Nevertheless, more recently an increasing number of scholars are finding contradictory evidence in the literature and pointing out that multilinguals vary with regards to many sociolinguistic variables which may determine the development of cognitive benefits deriving from the experience with multiple languages (e.g. De Bot, 2017; Lehtonen et al., 2018). In two of the few large-scale studies on foreign language learning in the school context that have controlled for sociolinguistic variables, Berthele and Udry (2021) found that immigrant students in Switzerland who were more multilingual than non-immigrant students actually had no advantage for foreign language learning, while Pfenninger (2020), also in Switzerland, found that L1 literacy and socio-economic status predicted success in L2/L3 English better than whether students had started studying English at a younger age. Most scholars argue that the cognitive advantages multilingualism can offer may primarily be mediated by educational and language biographical variables. As Berthele (2021a) points out, it may be depending on their sociolinguistic and educational situation that multilingual language learners do or do not unfold their potential.

The role of education in maximising the benefits that multilingualism can have for language learning is thus crucial. There is therefore a rapidly increasing interest in moving towards an educational approach that would make the most of learner’s multilingualism and, since the 2010s, ‘translanguaging’ has become the most popular term to refer to such an approach (Poza, 2017).

3. Translanguaging

In his thesis, Williams (1994) originally visited Welsh revitalisation school programmes with the goal of identifying successful bilingual pedagogical practices. He identified as a successful such practice a planned alternation between two languages in the classroom for input and output, which he labelled ‘trawsieithu’ and Baker (2001) later translated as ‘translanguaging’. In this practice, sometimes the teacher would teach in Welsh and the learners would respond largely in English, while sometimes the learners would read in Welsh and the teacher would give explanations largely in English. This was seen as a scaffolding strategy to promote proficiency in the revitalisation language, Welsh. As Wei (2017) notes, this crosslinguistic practice was not unique to Wales, but what was novel is that Williams considered it a scaffolding strategy, unlike traditional monolingual ideologies that consider the use of multiple languages in the classroom a pedagogical failure.

As mentioned earlier, however, the meaning of translanguaging is now being extended in different directions and the concept is often understood rather vaguely, as a practice wherein several languages are in some way used within the same lesson. Cenoz and Gorter’s (e.g. 2014, 2020) proposal of pedagogical translanguaging adheres to Williams (1994) but includes a greater number of planned instructional crosslinguistic strategies with the aim of promoting proficiency. By contrast, García and colleagues’ (e.g. García & Otheguy, 2020) proposal of spontaneous translanguaging aims to legitimise in the classroom the use of crosslinguistic practices in which multilinguals (sometimes) engage during informal interactions ‘in the street’ (García, 2009, p. 47).

3.1. Spontaneous translanguaging

García and colleagues’ (e.g. García & Lin, 2017; Vogel & García, 2017) proposal of spontaneous translanguaging can be linked to a socio-linguistically oriented body of studies within research on multilingualism (e.g. Blommaert, 2016; Pennycook, 2010) that focuses on the fluidity with which multilinguals sometimes move across their languages during social interaction, often referred to as ‘languaging’. It can be described as a sociolinguistic educational theory focused on empowering multilingual students who speak both a minority and a majority language. These students are often students from low-income households with an immigrant background who (sometimes) include elements from both languages in their informal speech practices. García and colleagues argue that they can be empowered by legitimising in the classroom the use of such informal multilingual speech practices. Doing this is seen as a matter of social justice and a ‘powerful mechanism to construct understanding, to include [minorities], and to mediate understandings across language groups’ (García, 2009, p. 9). In order to strengthen this argument, García and colleagues claim that languages do not have a cognitive reality as distinct systems, and that people thus do not speak ‘languages’, but engage in ‘languaging’. Therefore, they argue, students’ ‘named languages’ should not be separated or distinguished in the classroom and students should be allowed to use them together as they (sometimes) do in their spontaneous interactions. Whereas in Williams (1994) thesis the prefix ‘trans’ refers to moving across languages in the classroom, in García and colleagues’ proposal it thus refers to going beyond the notion of languages itself.

Adapting the concept of translanguaging in the context of bilingualism in the USA, García (2009) focuses on the example of ‘Spanglish’, describing it as a term used to belittle the speech practices in which Latino Spanish-English bilingual speakers sometimes engage. García discusses the insight that only certain language varieties or sociolects – those associated with the privileged class – are considered legitimate, while sociolects of those outside the empowered centre are stigmatised. Sociolects considered legitimate are socio-politically defined as standard languages and called by names, for instance ‘English’, while all other sociolects, such as Spanglish or Black English, are often considered ‘impure’ ways of speaking. In a seminal article, Labov (1972) points out for instance how policy-makers and linguists in the USA were long convinced that Black English was a form of slang incapable of expressing complex thought and that Black English speakers therefore ‘cannot speak complete sentences, do not know the names of common objects, and cannot form concepts or convey logical thoughts’ (no page number). While Spanglish is usually described in terms of code-switching, García and colleagues argue that Spanglish speakers are neither failing to keep their languages apart nor switching between them. Instead, they argue, for these speakers English and Spanish represent ‘a unitary linguistic system [or ‘linguistic repertoire’] that is not compartmentalised into boundaries corresponding to those of the named languages’ (García & Otheguy, 2020, p. 25). García and colleagues reason that since named languages (‘English’, ‘Spanish’) are socio-politically defined, languages are not real, but constructs, and therefore do not represent distinct systems in the mind, either. Consequently, people ‘do not speak languages, but rather use their repertoire of linguistic features’ (García & Kleyn, 2016, p. 19) ‘without adherence to the socially and politically defined boundaries of […] languages’ (Otheguy et al., 2015, p. 283). This is a ‘discursive practice in which they engage […] to make sense of their bilingual worlds’ (García, 2009, p. 45).

Traditionally, a TL-only approach has been adopted in language education based on the monolingual perspective described in the previous section. García and colleagues’ translanguaging wants to dismount this ideology and proposes that students be allowed to use all their linguistic features (i.e. all their languages/language varieties) in the classroom. They describe their proposal of translanguaging as a ‘transformative pedagogy’ because legitimising multilingual speeches in the classroom ‘giv[es] authenticity to the bilingual practices of many (García, 2009, p. 12) as well as calls ‘forth bilingual subjectivities [… in a way that] gives back the voice that had been taken away by ideologies of monoglot standards’ (García & Wei, 2014, pp. 92–105) and that ‘disrupt[s] the hegemony of the named national languages and of the power of the political state’ (García & Kleyn, 2016, p. 21). They see allowing students to use all their languages in the classroom as a matter of social justice (García & Wei, 2014, p. 116) and as the only approach to teach all children in the twenty-first century, ‘the rich and the poor, […] the Indigenous […] and immigrants, speakers of official and/or national languages […] and regional languages […] for hearing children, as well as Deaf children’ (García, 2009, p. 11). As Jaspers (2018, p. 3) discusses in a critique, spontaneous translanguaging can be described as a transformative pedagogy also because García and colleagues associate it as causally related to school success. Jaspers illustrates with the following quotation from García and Wei (2014, p. 55) how spontaneous translanguaging intimates that it will increase educational success by ascribing educational failure to a lack of a translanguaging approach:

Today, the punishment is not corporal, but relies on instruction and assessments that follow monolingual language standards, ensuring that bilingual students get lower grades, are made to feel inadequate, and fail in schools […] Clearly, the educational consequence of the socio-political inability to authenticate a multilingual heteroglossic reality is responsible for educational failure.

Although spontaneous translanguaging is an ideological rather than instructional proposal, it emphasises interpersonal and collaboration between students in the classroom, suggesting that they could often be grouped according to the named language(s) they use at home in order to facilitate the use of their whole linguistic repertoire (García & Kleyn, 2016, p. 22). Allowing students to collaborate using their whole linguistic repertoire in this way is expected to increase their understanding and generation of abstract content, to increase participation and to enhance their self-image as competent language users in today’s superdiverse world (ibid.). To date, some explorative qualitative studies have supported these effects (e.g. García et al., 2015; Wei, 2015), but translanguaging is still mostly being theorised and more robust evidence of effects is needed (Berthele, 2021b; Jaspers, 2018).

Undoubtedly, García and colleagues’ spontaneous translanguaging has had an enormous impact and contributed to raising critical voices against the traditional monolingual ideology, remarking that crosslinguistic practices are not a sign of deficiency but a reality of language use in many contexts where boundaries between named languages are fluid. Yet, while its central idea that languages are represented as one single system may be attractive from a social or even philosophical point of view, its theoretical and empirical grounds, as well as its pedagogical implications, have been questioned in various respects.

Theoretically, many terms have been used to distinguish between the social and individual dimensions of language. The distinction is highlighted, for instance, in Chomsky’s famous terms of ‘E-language’ (external language) and ‘I-language’ (internal language) from the 1960s or in the usage by the Council of Europe and some languages of (geographical) multilingualism and (individual) plurilingualism. As MacSwan (2017) discusses, researchers interested in the nature of language as a cognitive phenomenon investigate aspects related to the internal representation and processing of language, such as transfer, whereas those interested in the nature of language as a social phenomenon investigate how language use varies as a function of variables such as sex or social class. MacSwan (2017) criticises the way García and colleagues arrive at conclusions about the internal representation of languages based directly on insights about the way languages are socio-politically defined. The reasoning in spontaneous translanguaging is that since the boundaries between named languages are socio-political constructs, boundaries between languages in the mind must also be constructs. The implication of this reasoning is that, for García and colleagues, multilingualism is not psycholinguistically real: speakers have one linguistic repertoire, not languages, irrespective of whether they speak one or several named languages. In turn, García and colleagues therefore reject the central notions of code-switching (e.g. García & Otheguy, 2014, p. 282) and transfer (e.g. García & Wei, 2014, p. 80), asserting that these notions endorse a dangerous view of languages as monolithic systems (e.g. Otheguy et al., 2015, p. 282). As MacSwan (2017) points out, however, these claims are unfortunately formulated as self-evident claims not tied to an empirical analysis. In fact, scholars investigating code-switching do not advocate for a view of languages as monolithic systems, but as partly separate and partly integrated systems. Already several decades ago, Sankoff and Poplack (1981) noted that studies on code-switching have found code-switches adhere to the syntactic rules of the language with which each phrase starts (the ‘lexifier language’), leading to phrases such as ‘the white casa’, where the phrase starts with English (‘the’) and follows an English word order (adjective + noun), but not to phrases such as ‘the casa white’, where the phrase follows a Spanish order (noun + adjective). Since code-switching thus follows the rules of the lexifier language, code-switching scholars have long agreed that it is not plausible that in bilingualism the two systems unite to form an undifferentiated system, as this would yield phrases like ‘the casa white’, which have yet to be observed. Neither is it plausible that they are represented as fully separate systems, as this would not allow for the combination of words from different languages within phrases (MacSwan, 2017). Transfer scholars, too, consider languages to be represented as partly separate and partly integrated systems, not as monolithic systems. The models of the multilingual mental lexicon, which offer a theoretical framework for transfer, describe how words in the psycholinguistic system are connected to each other based on shared characteristics (e.g. words belonging to ‘English’) and factors such as morphological similarity (e.g. the English ‘wind’, ‘wild’ and ‘kind’ will be connected to and co-activated with the Swedish ‘vind’ [‘wind’], ‘vild’ [‘wild’] and ‘kind’ [‘cheek’]), thus forming intra- and interlinguistic networks (for an overview, see Fuster & Neuser, 2021).

The idea of the unitary linguistic repertoire in spontaneous translanguaging has also been questioned in relation to its practicality. As Berthele (2021b, p. 100) notes, ‘[s]hifting away from countable languages […] also means shifting away from testable hypotheses on language development in the categories relevant to policy makers, teachers, parents and learners’. Cenoz and Gorter (2017) warn from policies that do not sufficiently differentiate between languages in contexts where the aim is to revitalise and ensure the maintenance of minority languages. Many minority languages such as Spanish or Mandarin in the USA, where spontaneous translanguaging has been formulated, are demographically strong and enjoy a high status in other countries. The case is very different, however, for minority languages that are only spoken in certain regions and have been historically oppressed, such as Basque or Catalan in Spain. The future for these languages is never secured, but specific space for them needs to be created constantly. There is a fear that if languages are not differentiated enough in school, the minority language may become a pidgin and eventually transform into lexical reminiscences in the high-status language, which is bound to survive given its dominant status. Focusing on the dimension of teaching, Cummins (2017) additionally questions the usefulness of the notion of the unitary linguistic repertoire for making use of learners’ BLs as resources in TL teaching/learning. He describes how an important precursor of translanguaging as formulated in Williams (1994) is his own seminal Interdependence Hypothesis from the 1980s, which, based on the notion of a ‘common underlying proficiency’, argues that ‘instruction of multilingual students should aim explicitly to promote crosslinguistic transfer’ (Cummins, 2017, p. 1). As Cummins (2019) notes, the usefulness of rejecting the notion of transfer in spontaneous translanguaging can be questioned because ‘[i]f not teaching for transfer, how should teachers in Spanish/English bilingual programs conceptualize what they are doing when they draw students’ attention to similarities between encontrar and encounter?’ (p. 10). Fuster (2022), too, agrees that if teachers point out such crosslinguistic similarities, they are essentially treating languages as distinct systems. If they, instead, take García and colleagues’ notion of the unitary linguistic repertoire to its logical conclusion and do not discuss linguistic features as belonging to distinct languages, they cannot engage in what Cummins (2008, 2017, 2019, 2008) calls ‘teaching for transfer’, i.e. in raising awareness of similarities between learners’ BLs and the TL in order to facilitate transfer as a resource.

In Neuser thesis, Fuster (2022) explains how Cenoz and Gorter’s (e.g. 2014, 2020) proposal of pedagogical translanguaging, in the same vein as Cummins’ notion of ‘teaching for transfer’, argues that raising students’ awareness of similarities between their languages will help them make use of transfer in an intentional and creative way, as a strategy to enhance TL performance.

3.2. Pedagogical translanguaging

Cenoz and Gorter’s proposal of pedagogical translanguaging connects the findings from the aforementioned TLA body of studies on linguistic awareness, learner strategies and transfer. It connects these findings by suggesting that teachers should ‘highlight the relationships between languages […] so as to enhance linguistic awareness’ and in this way promote the transfer of BL features as a learner strategy to enhance TL performance (Cenoz & Gorter, 2011, p. 360). Cenoz and Gorter (2022) describe pedagogical translanguaging as ‘a theoretical and instructional approach’ (p. 1) focused on ‘activat[ing] […] the pre-existing knowledge that students have in their multilingual repertoire’ (p. 43), ‘softening boundaries between languages so that [they] make the most of their own multilingualism’ (p. 4).

In response to García and colleagues’ spontaneous translanguaging, Cenoz and Gorter (2017) note that the boundaries defining named languages are indeed socio-politically constructed, but that speakers treat languages as distinct and use them in isolation or in alternation, depending on the situation. An important reason to refer to them as distinct systems, they argue, is the need students have to develop an ability to use specific languages according to the rules that have been socially constructed (Cenoz & Gorter, 2022, p. 13). Moreover, Cenoz and Gorter (ibid.) agree with Canagarajah (2013) that although there is crosslinguistic transfer and languages are partly integrated, languages have a reality as distinct systems for social groups and are an important form of identity. Cenoz and Gorter (2014, p. 241) also note that spontaneous and pedagogical translanguaging share their heteroglossic ideology and highlight learners’ agency in language use, but whereas spontaneous translanguaging ‘places emphasis on spontaneous multilingual speech’, pedagogical translanguaging ‘has its roots in multilingual education understood as the use of two or more languages in education provided that schools aim at multilingualism’.

While Williams (1994) translanguaging referred to the planned alternation between two languages for input and output as a strategy to promote proficiency in a minority language, Cenoz and Gorter’s (2020, p. 1) pedagogical translanguaging proposes even ‘more intentional instructional strategies that integrate two or more languages’. The most emphasised dimension of pedagogical translanguaging is raising linguistic awareness so as to teach for transfer and enhance TL performance, but raising language awareness is also proposed as a dimension to foster critical thinking on multilingualism. Here, ‘language awareness’ refers to an awareness of languages in society, whereas ‘(meta)linguistic awareness’ refers to an awareness of aspects about a language system (e.g. lexical or phonological aspects).

The dimension of language awareness raising concerns raising students’ awareness on issues such as the social status of languages, power differences between languages or the ways in which multilinguals can use their languages depending on the situation (Cenoz & Gorter, 2017b). Such language awareness is expected to improve learners’ self-image as multilingual users and to promote critical thinking on linguistic diversity. Based in the context of the Basque Country, where Basque cohabits with either Spanish in Spain or French in France and has historically been oppressed, an aspect emphasised by Cenoz and Gorter (2017) is that language awareness can contribute to a better understanding of the role of minority languages in society. In a study, Cenoz and Gorter (2006) photographed 975 signs in shopping areas in Donostia-San Sebastián (in the Basque Autonomous Community, Spain) and Leeuwarden-Ljouwert (in Friesland, the Netherlands). They found that 55% of the signs in the former city and 44% in the latter included words both in the majority language (Spanish/Dutch) and in the minority language (Basque/Frisian), whereas the remaining signs consisted of words in the majority languages alone. Due to a more protective policy towards Basque in the Basque Autonomous Community, Basque has a relatively strong presence on the signage, whereas the less protective policy in Friesland results in less visibility for Frisian. Cenoz and Gorter (2018) say that having students discuss such work on the linguistic landscape can be an example of a language awareness-raising activity.

In the dimension of linguistic awareness-raising, Cenoz and Gorter (2014) consider raising learners’ linguistic awareness a way of helping them use ‘all [their] languages as connected growers’ (p. 246). With this metaphor from dynamic systems theory, Cenoz and Gorter mean that languages have features that can be used as resources that strengthen – not interfere with – each other. Although Cenoz and Gorter do not explicitly discuss the use of resources resulting from linguistic awareness with the term ‘transfer’, they contrast their idea with the statement that ‘transfer […] has [traditionally] been regarded as negative’ (2017b, p. 242). They (ibid., p. 313) write that pedagogical translanguaging, instead, considers transfer in a more positive way, as ‘a way [in which students] can be creative’. Cenoz and Gorter (2011, p. 12), after providing examples of lexical transfer, conclude that ‘in some cases learners use a term from another language […] because they find the term matches their communicative intent [… They] use their resources in the repertoire in different ways for communicative purposes’. Whereas transfer is traditionally conceptualised as a negative phenomenon of interference, which per definition occurs unintentionally, pedagogical trans-languaging thus offers a conceptualisation of transfer as a phenomenon that can be used intentionally, as a creative strategy or resource.

A reason why learners do not always use transfer in such an intentional way, however, is that ‘the monolingual focus on the TL can prevent [them] from using their own resources’ (Cenoz & Gorter, 2020, p. 5). As discussed previously, multilingualism has been found to increase linguistic awareness, but recent studies also show that multilingual learners’ linguistic awareness is not as advanced as it could be and the role of education in maximising the benefits that multilingualism can have is crucial. Leonet et al. (2020), for instance, found that multilingual learners often do not recognise many helpful similarities between their languages and, consequently, cannot take advantage of those similarities through transfer. Pedagogical translanguaging argues that raising learners’ awareness of crosslinguistic similarities ‘can help making associations with prior knowledge’ and thus help to find more opportunities to use transfer intentionally as a strategy (Cenoz and Gorter, 2022, p. 28; see also Cenoz & Gorter, 2020, p. 5).

To date, most empirical studies that can be discussed in relation to pedagogical translanguaging have focused on morphological awareness-raising activities and indicated that such activities seem to have a positive effect on vocabulary size as well as learners’ motivation to proactively reflect on unfamiliar words and make lexical guesses (e.g. Bowers et al., 2010; Leonet et al., 2020; Lyster et al., 2013; Otwinowska et al., 2020; White & Horst, 2012; Zhang & Li, 2016). As Fuster (2022) argues, these studies support the idea in pedagogical translanguaging that linguistic awareness is beneficial for enhancing TL performance. However, there is a need to also investigate how students employ transfer as an intentional strategy based on linguistic awareness or other potentially important variables. As noted in the introduction, although some theoretical accounts note that transfer can be used as a strategy and some empirical studies have also considered transfer as such (cf., e.g. Herdina & Jessner, 2002; Jarvis & Pavlenko, 2008; Jessner, 2006; Williams & Hammarberg, 1998), the mainstream view is still that transfer is an automatic or unintentional by-product of the co-activation of several languages in the mind (see, e.g. De Bot, 2004; Ecke & Hall, 2014) (see also Fuster & Neuser, 2021, for an overview). Therefore, from an empirical perspective very little has been explored about the difference between how learners employ transfer intentionally as opposed to unintentionally and what variables may be related to (un)intentionality in the use of transfer. In the next section, I outline key findings from Fuster (2022), which explored precisely this question.

3.2.1. Key findings about learners’ unintentional vs. intentional transfer

Fuster (2022) focused on lexical transfer in written production, which is less affected by time constraints than oral production and where transfer as a strategy, therefore, may be employed more easily. The study was guided by the following research questions:

1. In what proportions do learners use unintentional versus intentional lexical transfer?

2. What potentially important transfer variables affect learners’ amounts of unintentional versus intentional lexical transfer?

3. How are the different types of lexical transfer (borrowing, foreignizing, spelling transfer, lexeme matching, semantic extension, direct translation) manifested depending on whether they are produced unintentionally or intentionally?

The participants were 78 upper secondary Spanish students (16–18 y.o.) in Sweden, who were familiar with an average of 4.29 typologically related and unrelated languages (Standard deviation [SD] 0.80) and who had a low or low-intermediate level of TL proficiency. All of them were familiar with Swedish and English, and 12 other languages were also represented in the sample (e.g. Aramaic, Chechen, Persian, Polish, Italian, Portuguese). They attended a ‘heterogeneously multilingual school’, where most students have a foreign background and speak many different L1s. 29% reported to speak Swedish as an L1, while 71% reported to speak it as an L2 (i.e. not learned at home with their caretakers).

In the empirical task, 64 learners were asked to describe Mayer’s (1969) ‘Frog Where Are You’ picture story (20 min) and try to convey the story to a speaker of Spanish in any possible way, without worrying too much about making mistakes. They wrote the description simultaneously during their regular Spanish class and, once they had finished writing, they were instructed to conduct a self-reflection and underline the words that they had been unsure of (10 min) (a kind of retrospective recall interview but without providing explanations). They were told to underline the words that they felt unsure of at the time of writing the text, and not in retrospect. The transfer words that were underlined were considered to represent intentional use, while those that were not underlined but were identified as negative transfer by the researcher were considered to represent unintentional use. The first 14 learners, however, underwent the same procedure but also produced a think-aloud protocol while writing the picture-story description. The procedure for these first 14 students was the one used in a pilot study with four adult learners of Catalan, published in Fuster and Neuser (2020). Verbal reports from think-aloud protocols provide deep insights into the decision-making processes underlying intentional transfer (e.g. why a learner decides to adapt a word by changing certain letter combinations in one way and not in another way). However, the think-aloud protocol is a very time-consuming method and, given the larger sample of 78 learners in Fuster’s (2022) study, the faster procedure without the think-aloud was tested with the next 14 learners. A t-test found no significant differences between these two groups of 14 learners with regards to the total amount of transfer, the amount of unintentional transfer and the amount of intentional transfer. Therefore, the remaining 50 students underwent the faster procedure.

There were 299 transfer words in the dataset. If a student had transferred the same word several times, only one token was included in the dataset. On average, 8% of the total words produced by the learners were transferred words (standard deviation [SD] 6), a similar percentage to those in previous studies – e.g. 5.3% in Lindqvist (2006) or 7.7% in Neuser (2017). The data showed that learners produced lexical transfer both intentionally and unintentionally. 32% of the transferred words were transferred unintentionally and 68% were transferred intentionally. The upper rows in Tables 1 and 2 show the learners’ counts of unintentional and intentional transfer observed in the data, with the lower rows giving the percentage of learners in each count.

Table 1. Counts and percentages of unintentional transfer.

counts	0	1	2	3	4	5	6	10	16	total
% learners	46%	27%	18%	3%	3%	1%	1%	1%	1%	100%

Table 2. Counts and percentages of intentional transfer.

counts	0	1	2	3	4	5	6	7	8	10	total
% learners	10%	24%	22%	17%	13%	5%	4%	3%	1%	1%	100%

As illustrated, most learners (91%) produced between 0 and 2 instances of unintentional transfer and the remaining 10% were dispersed between 3 and 16 counts, contributing to a considerable degree of inter-subject variation (mean 1.24, SD 2.31, variance 5.49). When it comes to intentional transfer, 86% of the learners produced between 0 and 4 instances, while the remaining were dispersed between 5 and 10 counts (mean 2.58, SD 2.00, variance 4.06).

Given this inter-subject variation, a regression analysis with R was conducted to test the association of potentially important variables with learners’ counts of unintentional vs. intentional lexical transfer. For count variables like these, which will ideally exhibit a Poisson distribution, Poisson regression analysis is used (Roback & Legler, 2021). Ideal Poisson data distributions have a variance equal to the mean, but in practice this characteristic often does not hold and this was also the case in the present study, where the variance was higher than the mean. Given these over-dispersed count variables, a quasi-Poisson regression analysis was used. Quasi-Poisson regression, like Poisson regression, uses the mean regression function and the variance function, but estimates the dispersion parameter from the data instead of assuming it to be fixed at 1 (Zeileis, Kleiber, & Jackman, 2008). Quasi-Poisson regression yields the same coefficient estimates as Poisson regression, but the precision in the inference is adjusted for over-dispersion. The independent variables were TL proficiency, the average of transfer from L1 vs. L2 source languages (SLs)⁴ of transfer, the average of proficiency in the SLs of transfer, the average of frequency of use of the SLs of transfer, the average of typological closeness of the SLs to TL Spanish, the average of perceived typological closeness of the SLs to TL Spanish by the learner (i.e. ‘psychotypology’) and the number of languages known by the learner. The first six variables have been found to be main factors of the predominance of a SL over another SL of transfer (e.g. more transfer from L2s than from L1s) (see Falk & Bardel, 2010; Ecke, 2015; Neuser, 2017, for overviews). In this study, the variable of number of languages was used as a proxy for linguistic awareness. Throughout the literature, it has been argued that more multilingual individuals develop a heightened level of linguistic awareness, which appears to result in a greater number and more frequent use of learner strategies. While number of languages has generally not been studied as a factor of the SL of transfer, in theoretical models of multilingual language organisation and activation, Ecke and Hall (2014) posit that linguistic awareness may also be an important factor modulating cross-lexical connections and activation and affecting, thereby, the probability that a word is transferred.

The clearest finding with regards to unintentional lexical transfer was that it most often occurred due to crosslinguistic morphological similarity. In Table 1, the first column shows the estimated regression coefficients, with their standard errors in the second column. The results are discussed in terms of exponentiated regression coefficients (or odds ratios), as usually done in generalised linear models. The confidence intervals for these exponentiated regression coefficients are shown in the fifth column. As illustrated in Table 1, the regression analysis found no significant effect at the conventional level of p < 0.05 for any variable except for typological closeness of the SLs in learners’ amounts of unintentional transfer. All the other variables held constant, a learner with one additional score in typological closeness of the SLs of transfer was estimated to produce almost 9 times the amount of unintentional transfer of a learner with one score less. Thus, the amount of unintentional transfer from typologically very close SLs (Portuguese and Italian) was much higher in speakers of these languages than in speakers of typologically less close SLs (Swedish and English) and in speakers of typologically very distant SLs (Arabic). A learner who transferred only from Romance languages, for example, is estimated to produce around 9 times the amount of unintentional transfer of a learner who transferred only from Germanic ones (which were coded as one value lower in typological closeness). Although this effect was very strongly significant (p < 0.001), it should be noted that the 95% confidence interval was wide (4.26, 18.14), however, indicating that there is a degree of uncertainty and that future studies should further explore this variable (Table 3).

Table 3. Results of the quasi-Poisson regression analysis for the learners’ amounts of unintentional transfer.

	Estimated β	Standard error for β	Exp(β)	p values and codes	95% CI for exp(β)
(Intercept)	–3.61	2.13	0.03	0.085	(0.00, 1.76)
Nr. languages known	–0.31	0.16	0.74	0.064	(0.54, 1.00)
SL typological closeness	2.17	0.37	8.79	0.000***	(4.26, 18.14)
SL psychotypology	0.16	0.18	1.17	0.372	(0.65, 1.82)
SL proficiency	0.09	0.26	1.09	0.688	(0.55, 2.01)
SL frequency	0.05	0.33	1.05	0.910	(0.37, 1.29)
L2 status	–0.37	0.32	0.69	0.232	(0.37, 1.29)
TL proficiency	0.09	0.12	1.09	0.357	(0.87, 1.37)

Codes for conventional levels of significance: ‘***’ p < 0.001, ‘**’ p < 0.01, ‘*’ p < 0.05, ‘·’ p < 0.1

Fuster and Neuser (2021) discussed how this result provides support for the assumption in theoretical models of the multilingual mental lexicon that morphologically similar BL words (which will be more frequent in typologically close BLs) will be highly co-activated during TL use and, consequently, more likely to ‘slip in’ unintentionally (e.g. De Bot, 2004; Ecke & Hall, 2014). This result is also in line with the pilot study in Fuster and Neuser (2020) with adult learners of Catalan, which, employing descriptive statistics, found that 90% of transfer from typologically distant SLs (English, Swedish, German) was produced intentionally, whereas 63% of transfer from typologically close SLs (Spanish, French, Italian, Portuguese) occurred unintentionally. Given that words from typologically distant languages will generally not fit well within the morphological constraints of the TL, Fuster and Neuser (2020) reasoned that learners use these words very selectively or consciously. By contrast, words from typologically close languages will often fit relatively well within the morphological constraints of the TL and, therefore, be strongly connected to morphologically similar words in the TL. Due to such strong cross-lexical connections, these words will be highly activated during TL use and, consequently, be more prone to be transferred unintentionally, without requiring conscious effort for them to be selected.

The association between unintentional transfer and morphological similarity was also observed on a word-specific level when exploring the different types of lexical transfer. Unintentional transfer predominantly occurred on SL words that already fit TL morphology, such as in the case of spelling transfer in cognates – e.g. occupado, influenced by the English ‘occupied’ (Spanish: ‘ocupado’) – or in the case of lexeme matching (which usually occurs with so-called ‘false friends’) – e.g. carpeta (‘folder’), used with the meaning of the English ‘carpet’ (Spanish: ‘alfombra’). Intentional transfer, by contrast, predominantly occurred on SL words that required more advanced adjustments in order to fit TL morphology, usually in the form of foreignisings. The fact that intentional transfer usually occurred on SL words that did not fit TL morphology and therefore required more difficult adjustments was clearly observed in the differences between unintentional foreignisings and intentional foreignisings. The changes made in unintentional foreignisings consisted of exchanging SL suffixes for their corresponding TL suffixes – e.g. English ‘exits’ → exita [Spanish: ‘sale’]), Swedish ‘ropar’ → ropando (Spanish: ‘llamando’, English: ‘calling’) – whereas intentional foreignisings also consisted of more complicated changes made to the root of the word – e.g. Swedish ‘skrämd’ → escremado (English: ‘afraid’, Spanish: ‘asustado’), English ‘stressed’ → estresado (Spanish: ‘estresado’). The way learners adapt such BL words reflects their knowledge of the morphological constraints of in the TL. In the examples above, learners demonstrate knowledge of rules such as /s/ in Spanish being able to be initial only if followed by a vowel, otherwise requiring an /e/ preceding it, or /ä/ in Swedish being represented by /e/ in Spanish.

Another interesting finding with regards to unintentional lexical transfer was that learners who knew more languages exhibited considerably smaller amounts of unintentional transfer than learners who knew fewer languages. While not a statistically significant result, the p value for number of languages known was close to 0.05 (p = 0.06), while the p values for the other non-significant variables were all above 0.1 (Table 1). Fuster (2022) reasoned that this trend is in line with previous studies indicating that more multilingual individuals perform more strategically due to an increased level of linguistic awareness (see, e.g. Cenoz, 2013). More multilingual learners may notice more differences and similarities between the TL and their BLs and develop, thereby, a strengthened control of their transfer use, being more capable of inhibiting the transfer of words that are highly activated but which they do not intend to use. This can be aligned to the notion of ‘the enhanced multilingual monitor’ in the Dynamic Model of Multilingualism (Herdina & Jessner, 2002), which is described as a function of linguistic awareness that is directly proportional to the number of languages known by an individual and responsible for controlling potential errors due to cross-lexical co-activation (i.e. unintentional transfer).

When it comes to learners’ amounts of intentional lexical transfer, a significant association (p < 0.01) was found with the variable of L2 status, while the other variables were non-significant (Table 2). The higher a learner’s average of transfer from L2s, the higher the learner’s count of intentional transfer tended to be. All the other variables held constant, a learner with one additional score in amount of transfer from L2s was estimated to produce almost 2 times the amount of intentional transfer of a learner with one score less in the amount of transfer from L2s. Most often, intentional transfer came from English, which was an L2 for all learners except one (Table 4).

Table 4. Results of the quasi-Poisson regression analysis for the learners’ amounts of intentional transfer.

	Estimated β	Standard error for β	Exp(β)	p values and codes	95% CI for exp(β)
(Intercept)	1.08	1.01	2.93	0.290	(0.40, 21.21)
nr.languages	–0.15	0.11	0.86	0.173	(0.69, 1.07)
SL.morphology	0.21	0.35	1.23	0.551	(0.62, 2.43)
SL.psychotypology	–0.05	0.13	0.95	0.684	(0.74, 1.22)
SL.proficiency	–0.12	0.14	0.88	0.396	(0.67, 1.17)
SL.frequency	–0.12	0.18	0.89	0.502	(0.63, 1.25)
L2.status	0.64	0.23	1.90	0.005**	(1.22, 2.97)
TL.proficiency	–0.02	0.12	0.98	0.879	(0.77, 1.24)

Codes for conventional levels of significance: ‘***’ p < 0.001, ‘**’ p < 0.01, ‘*’ p < 0.05, ‘·’ p < 0.1

As Fuster (2022) discusses, this result may be aligned with the idea in the L2 Status Hypothesis (Bardel & Falk, 2012) that L2s usually rely on the conscious system of declarative memory, where metalinguistic knowledge is processed, to a greater extent than L1s. Since English (the most common SL of transfer and the most common SL of intentional transfer, in particular) was most often an L2 in the current sample, being taught explicitly in school and used for academic purposes, the learners might have developed more metalinguistic knowledge in this L2 than in their L1s. With an increased level of metalinguistic knowledge in English, the learners may more easily access and manipulate words from this language in order to solve gaps of knowledge in the TL by intentionally transferring those words.

4. Conclusions

‘Translanguaging’ has become the most popular term in discussions about how to move away from monolingual ideologies in education and make use of students’ BLs as resources for TL teaching/learning. However, the term ‘translanguaging’ has also become ambiguous because it is being developed in different theoretical proposals that adopt different perspectives on individual multilingualism. This article has offered one of the few overviews of the two main current proposals of translanguaging together.

The main aim of García and colleagues’ proposal of spontaneous translanguaging is to legitimise in the classroom the use of crosslinguistic practices in which multilinguals sometimes engage during informal spontaneous interactions outside school, which is seen as a matter of social justice. As a way to motivate such linguistic practices in the classroom, spontaneous translanguaging argues that since the boundaries defining named languages are not real, but socio-politically constructed, languages in the mind cannot be represented as distinct systems, either, but as one single unitary linguistic repertoire. Thus, students do not speak different ‘languages’, but engage in ‘languaging’, and they should be allowed to make use of all their linguistic resources (from their different named languages) in the classroom. An implication of this view, García and colleagues argue, is that the notion of transfer must be rejected. This article has offered an overview of the criticisms that have been made in relation to the unitary view of spontaneous trans-languaging. Aligning with Cummins (2019), it was argued here that rejecting the notion of transfer may not be useful when wanting to make use of learners’ BLs as resources, because it makes it difficult for teachers to conceptualise what they are doing when raising awareness of linguistic similarities between learners’ languages – a practice that Cummins (2008, 2017, 2019, 2021) calls ‘teaching for transfer’.

By contrast, the notion of transfer lies at the heart of Cenoz and Gorter’s proposal of pedagogical translanguaging. Pedagogical translanguaging recognises that multilingual speakers use their languages in isolation or in alternation depending on the situation and aims to enhance learners’ TL acquisition through teaching for transfer. One of the central ideas in pedagogical translanguaging is that teachers could raise awareness of linguistic similarities between the TL and learners’ BLs so as to facilitate for learners to find opportunities to use transfer as creative, intentional strategy to solve gaps of knowledge in the TL. This view of transfer stands in contrast to the traditional and still mainstream view of transfer as an unintentional phenomenon that learners should try to avoid. This article has argued that the view of transfer as a creative strategy has great potential for moving towards a multilingual perspective in language education, but that research is needed on how learners employ transfer intentionally versus unintentionally.

Key findings from Fuster’s (2022) doctoral monograph, which explored precisely this question, were therefore outlined. The study showed that lexical transfer occurs both unintentionally and intentionally. An important implication of this finding is that lexical transfer should not be conceptualised in a dichotomous way, as either an unintentional mechanism to be avoided or an intentional and creative one to be promoted. Any theoretical account concerned with learners’ agency in transfer should consider that learners produce lexical transfer both unintentionally and intentionally. Often, unintentional transfer occurs in relation to crosslinguistic morphological similarity (e.g. in the case of false friends) and does not enhance learners’ TL performance (e.g. using the Spanish ‘carpeta’ [English: ‘folder’] as though it had the meaning of the English ‘carpet’ [Spanish: ‘alfombra’]). Intentional lexical transfer, by contrast, is often produced with BL words that do not fit within the morphological constraints of the TL and which learners foreignise according to their metalinguistic knowledge of morphological rules in the TL (e.g. Swedish ‘skrämd’ → escremado [Spanish: ‘asustado’, English: ‘afraid’; English ‘stressed’ → estresado [Spanish: ‘estresado’]). That intentional lexical transfer usually comes from L2s, in which learners are supposed to have developed more metalinguistic knowledge, rather than from L1s, also indicates that linguistic awareness in the form of metalinguistic knowledge helps learners find opportunities to use transfer strategically and creatively. A trend was also observed, moreover, for learners who know more languages to exhibit smaller amounts of unintentional transfer than learners who know fewer languages. If multilingualism increases linguistic awareness, this may indicate that more multilingual learners have developed an enhanced level of linguistic awareness in the form of conscious control of language use. Having developed an enhanced conscious control of language, more multilingual learners may be able to inhibit, to a greater extent than less multilingual learners, the transfer of highly co-activated BL words that they do not intend to transfer.

The finding that intentional transfer seems to be based mainly on linguistic awareness provides support for the idea in pedagogical translanguaging of raising learners’ awareness of crosslinguistic similarities so as to help them make use of BL features in a creative strategic way. However, the finding that learners also use transfer unintentionally, and mainly as an effect of crosslinguistic morphological similarity, suggests that helping learners with the use of their BLs in the TL cannot only involve raising awareness of similarities so as to promote intentional transfer. The results in Fuster’s (2022) study suggest that making use of learners' BLs as a resource for TL teaching/learners should involve helping learners find opportunities to use lexical transfer as a strategy to solve gaps of knowledge in the TL as well as helping them to avoid potential typical instances of unintentional lexical transfer that do not lead to grammatically correct TL words. Moreover, such pedagogical crosslinguistic awareness-raising activities should focus not only on similarities, such as the existence of cognates or corres-pondences of verb and noun inflections, but also on differences within similarities, especially when it comes to semantic and spelling differences in morphologically similar words, which often lead to unintentional transfer.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Correction Statement

This article has been corrected with minor changes. These changes do not impact the academic content of the article.

Notes

1 The term ‘target language’ (TL) refers to a specific language being taught and learnt.

2 The term ‘background language’ (BL) is used adhering to Williams and Hammarberg (1998) to refer to any non-target language (whether L1 or L2), which might be a source of influence2017.

3 Kellerman and Sharwood Smith (1986) introduced the term ‘crosslinguistic influence’ to depart from both the traditional term ‘interference’ and connotations of behaviourism that ‘transfer’ had come to have. Today, however, ‘transfer’ and ‘crosslinguistic influence’ are generally used interchangeably (e.g., Bardel, 2019; Jarvis, 2017). In this article, ‘transfer’ is preferred over ‘crosslinguistic influence’ because the focus is on learners’ agency in the use of BLs and ‘transfer’ can be a verb denoting an intentional action.

4 A ‘source language’ refers to a language being relied upon for transfer.