1. Introduction
The construct of possible selves, or ‘future self-guides’ (Markus & Nurius, Reference Markus and Nurius1986) originates in psychology and represents the ways individuals conceptualize and imagine their current and future self-identities. Zoltán Dörnyei adapted this notion and core principles in self-discrepancy theory (Higgins, Reference Higgins1987) to the context of learning second or foreign languages, which is encapsulated in his sociodynamic framework of second language (L2) motivation, or the L2 Motivational Self System (Dörnyei, Reference Dörnyei2005, Reference Dörnyei, Dörnyei and Ushioda2009; Dörnyei & Ushioda, Reference Dörnyei, Ushioda, Dörnyei and Ushioda2009), which proposes that L2 motivation is driven by a desire to reduce the discrepancy between current and future self-states.
Dörnyei's model has generated significant research productivity for over a decade and has been primarily motivated by questions related to three areas. The first strand of inquiry has aimed to investigate the theoretical construct validity of possible selves with a focus on the relationship between and relative explanatory power of influential motivational constructs of the ideal L2 self (Dörnyei, Reference Dörnyei2005) and integrativeness (Gardner, Reference Gardner1985) in different contexts (e.g., Ryan, Reference Ryan, Dörnyei and Ushioda2009; Taguchi, Magid, & Papi, Reference Taguchi, Magid, Papi, Dörnyei and Ushioda2009; Dörnyei, Reference Dörnyei and Macaro2010). This large-scale, psychometric research tends to be conducted with large samples of learners at the secondary and tertiary levels of education who complete Likert-scale, self-report measures and whose data are analyzed using sophisticated statistical techniques like structural equation modeling to reveal the internal structure of L2 motivation as proposed in Dörnyei's model. A second, related line of research has focused on assessing the strength of relationships between different components of the L2 Motivational Self System and subjective and objective measures of L2 learning (e.g., Kormos & Csizér, Reference Kormos and Csizér2008; Ryan, Reference Ryan, Dörnyei and Ushioda2009; Serafini, Reference Serafini2013; Al-Hoorie, Reference Al-Hoorie2017).
While the first two lines of quantitative research were crucial to establish the credibility of the self as a relevant and robust learner individual difference (ID) construct in L2 learning (see Dörnyei & Ryan, Reference Dörnyei and Ryan2015 for further discussion), other researchers argue for a more contextually situated and temporally dynamic view of motivational phenomena with a focus on the individual learner (Ushioda, Reference Ushioda, Hickey and Williams1996, Reference Ushioda, Dörnyei and Schmidt2001, Reference Ushioda, Dörnyei and Ushioda2009, Reference Ushioda, Dörnyei, MacIntyre and Henry2015; Dörnyei, MacIntyre, & Henry, Reference Dörnyei, MacIntyre and Henry2015; Mercer, Reference Mercer, Dörnyei, MacIntyre and Henry2015, Reference Mercer and King2016), reflecting a broader conceptual shift in L2 research as a whole (de Bot, Reference de Bot2015a) and motivating a third strand of mixed methods research that has sought to explore the dynamic nature of motivation. For example, several studies have looked at how motivation emerges over time from the mutually influential interaction between the external (i.e., contexts) and the internal (i.e., individual learner characteristics) within a wide range of social and educational settings such as Spain (Lasagabaster, Doiz, & Sierra, Reference Lasagabaster, Doiz and Sierra2014; Lasagabaster, Reference Lasagabaster2017), Sweden (Henry & Apelgren, Reference Henry and Apelgren2008; Henry, Reference Henry2009, Reference Henry2011), Japan (Sampson, Reference Sampson2012), Indonesia (Lamb, Reference Lamb2007, Reference Lamb2012), and Canada (Kim, Reference Kim, Dörneyi and Ushioda2009). Researchers have also explored how learner selves fluctuate over different timescales in formal instructed settings (Waninge, Dörnyei, & de Bot, Reference Waninge, Dörnyei and de Bot2014) and how the perception and construal of different aspects of the self evolve in naturalistic immersion settings (e.g., Du, Reference Du2015; Müller & Schmenk, Reference Müller and Schmenk2017), including the influence of study-abroad interlocutors on the way learners view and construct different facets of self (Pellegrino, Reference Pellegrino2005; Serafini, Reference Serafini and Gurzynski-Weissforthcoming).
Though different in their theoretical and methodological orientation, a common gap across all three strands of research is that the overwhelming majority have been carried out in instructed English as a second/foreign language (ESL/EFL) contexts (for reviews, see Sugita McEown, Noels, & Chaffee, Reference Sugita McEown, Noels, Chaffee, Csizér and Magid2014; Boo, Dörnyei, & Ryan, Reference Boo, Dörnyei and Ryan2015; Al-Hoorie, Reference Al-Hoorie2017; Ushioda, Reference Ushioda2017). As a result, our understanding remains limited in terms of how learners’ perception and construal of their possible selves interact with cultural, historical, pedagogical, political, and social factors that characterize different learning environments (e.g., heritage, immersion, multilingual) and that may be unique to learning target languages other than English. This gap highlights a broader question in the field, which is the extent to which the self is situated, or dependent on context, versus being a more global construct (e.g., Mercer, Reference Mercer2011a, Reference Mercer2011b, Reference Mercer, Mercer and Williams2014, Reference Mercer, Dörnyei, MacIntyre and Henry2015, Reference Mercer and King2016).
The current paper argues for the need to replicate empirical research that investigates self- and identity-related phenomena with different types of learners in diverse language learning contexts. Specifically, I argue that the robustness and generalizability of learner possible selves can be gradually and systematically assessed through external replication, which aims to collect new data from new subjects within a new context (Porte, Reference Porte2012) and investigate ‘whether the same findings are obtained by another researcher in another context, and whether the outcome appears to reflect knowledge which can therefore be separated from the context in which it was originally found’ (p. 3). As Porte (Reference Porte2012) explains, whereas an approximate replication study follows the original study's experimental procedures closely but modifies, for example, non-target variables, conceptual replication might aim to test previous results using a different research methodology (e.g., different data collection methods and instruments) or altering the independent and dependent variables under focus as well as changing the way they are measured. Whenever possible, to enable sound comparison between the original studies and replicated studies, Porte and McManus (Reference Porte and McManus2019) recommend first conducting close replication (only one major variable change), followed by approximate, and supplemented by conceptual replication.
In the current proposal, two original studies, Henry and Cliffordson (Reference Henry and Cliffordson2013) and Lasagabaster (Reference Lasagabaster2016), were identified as ideal for replication based on perceived potential importance in the field and for providing a rigorous methodological detail necessary for replication (Mackey, Reference Mackey and Porte2012). While the studies differ in terms of the learner population studied, the statistical approach used to analyze the data, and the pedagogical context, they share a similar focus on the role of learner characteristics (i.e., gender and language-related differences) in the perception and construal of learner possible selves within a multilingual context and also take a similar quantitative approach to assess the constructs of interest via self-report measures. Furthermore, both studies address key questions in the selves strand of L2 research though the relative impact of their findings might be broadened through replication.
Specifically, Henry and Cliffordson (Reference Henry and Cliffordson2013) investigated gender-related differences in the ideal L2 self (English) and third language (L3) selves (Spanish, German, or French) of secondary adolescent learners studying L2 English and their L3s as compulsory subjects in Sweden. On the other hand, Lasagabaster (Reference Lasagabaster2016) asked whether the relationship between English-medium instruction (EMI), motivation, and possible selves in L3 English is mediated by gender and students’ first language (L1) (Basque, Spanish, or both) in university students who were voluntarily enrolled in EMI in Spain. Interestingly, the results showed opposite findings pertaining to the role of gender, which are reported and discussed in detail below and critically considered in relation to key aspects in methodological design.
In the following sections, I define the focal constructs and briefly review key findings from the first two strands of research carried out within a quantitative paradigm, given the higher feasibility for replication (Porte & Richards, Reference Porte and Richards2012). Then, I detail the motivation, method, and main findings of the original studies and suggest ways to approximately and conceptually replicate them (e.g., Porte, Reference Porte2010, Reference Porte2012; Porte & McManus, Reference Porte and McManus2019). The paper concludes by highlighting the potential theoretical, empirical, and practical contributions of further situating the self (Mercer, Reference Mercer2011a) through replication in diverse learning contexts.
2. Background
2.1 Self-constructs and L2 development
What we know about L2 motivation and its relationship to L2 development has been significantly impacted by an increasing focus on the role of various self-constructs that originate in other fields like psychology (e.g., self-concept, self-efficacy, self-esteem, self-regulation, possible selves). Of these, the construct of possible selves as conceived and adapted within the L2 Motivational Self System framework (Dörnyei, Reference Dörnyei2005, Reference Dörnyei, Dörnyei and Ushioda2009) has arguably had the most significant impact both in terms of the theoretical evolution of L2 motivation as well as the types of questions motivating current research. The notion of possible selves (Markus & Nurius, Reference Markus and Nurius1986) is proposed to represent imagined future self-states that are inherently complex, specific to certain domains, situated within particular contexts, and dynamic across contexts and time, similar to the broader construct of self-concept (Mercer, Reference Mercer2011a, Reference Mercer, Mercer and Williams2014). In the domain of second/foreign language learning, Dörnyei proposed that L2 motivation is driven by the degree to which achieving TL proficiency is integral to one's imagined identity as a speaker/user of the TL. One's level of motivation to reduce the perceived gap between current and future self-states will lead to action (or not) and thus, positively or negatively impact L2 development.
The three main components of the L2 Motivational Self System are the ideal L2 self, which concerns the attributes that one desires to possess as a speaker/user of the TL and focuses on promoting one's hopes, aspirations, and accomplishments in relation to language learning; the ought-to L2 self, which represents the attributes that one believes one ought to possess in order to meet the expectations of others and aims to prevent negative outcomes; and the L2 learning experience, which encompasses the impact of different aspects of the learning context like the teacher, teaching style, language learning materials, the curriculum.
A review of quantitative research investigating the L2 Motivational Self System reveals two main strands of inquiry,Footnote 1 which are briefly reviewed in turn before I report in detail on the two focal studies proposed for replication.
2.1.1 Strand 1: Theoretical validation of the L2 self
The first strand of inquiry centers on empirically validating the core theoretical constructs in the L2 Motivational Self System, which has accumulated robust empirical support across several contexts. In particular, the construct validity of the ideal L2 self has been demonstrated based on large-scale questionnaire data provided by ESL/EFL learners studying in Hungary (Dörnyei & Csizér, Reference Dörnyei and Csizer2002; Csizér & Dörnyei, Reference Csizér and Dörnyei2005; Dörnyei, Csizér, & Németh, Reference Dörnyei, Csizér and Németh2006; Kormos & Csizér, Reference Kormos and Csizér2008; Csizér & Kormos, Reference Csizér, Kormos, Dörnyei and Ushioda2009; Csizér & Lukács, Reference Csizér and Lukács2010), Japan (Ryan, Reference Ryan, Dörnyei and Ushioda2009; Taguchi et al., Reference Taguchi, Magid, Papi, Dörnyei and Ushioda2009), China (Gao, Zhao, Cheng, & Zhou, Reference Gao, Zhao, Cheng and Zhou2007; Taguchi et al., Reference Taguchi, Magid, Papi, Dörnyei and Ushioda2009; Liu & Thompson, Reference Liu and Thompson2018), Iran (Taguchi et al., Reference Taguchi, Magid, Papi, Dörnyei and Ushioda2009; Papi, Reference Papi2010; Papi & Teimouri, Reference Papi and Teimouri2012, Reference Papi and Teimouri2014; Teimouri, Reference Teimouri2017), and Chile (Kormos, Kiddle, & Csizér, Reference Kormos, Kiddle and Csizér2011).
A consistent finding of this research is the strong relationship between the ideal L2 self and the dominant social-psychological construct of integrativeness (Gardner & Lambert, Reference Gardner and Lambert1972; Gardner, Reference Gardner1985, Reference Gardner2010), indicating that they are tapping into similar psychological and emotional self-identification processes (Dörnyei, Reference Dörnyei and Dörnyei2003) that hinge on the extent to which the learner identifies with (or desires to identify with) the speakers of the target language (Csizér & Dörnyei, Reference Csizér and Dörnyei2005; Dörnyei, Reference Dörnyei2005, Reference Dörnyei, Dörnyei and Ushioda2009; MacIntyre, Mackinnon, & Clément, Reference MacIntyre, Mackinnon, Clément, Dörnyei and Ushioda2009). However, compared to integrativeness, previous research consistently reports relatively stronger relationships between the ideal L2 self and subjective (e.g., reported learning effort) and objective measures (e.g., performance) of learning (e.g., Kormos & Csizér, Reference Kormos and Csizér2008; Ryan, Reference Ryan, Dörnyei and Ushioda2009; Serafini, Reference Serafini2013), accounting for up to 42% of variance in criterion measures on average (Dörnyei, Reference Dörnyei2005, p. 104). This has led to a growing consensus that ‘integrativeness is simply one local manifestation of a much more complex, powerful construct’ (Ryan, Reference Ryan, Dörnyei and Ushioda2009, p. 137) meaning that whereas the explanatory capacity of the integrativeness construct may be limited to certain learning contexts, the ideal L2 self appears to offer ‘a broader frame of reference with increased capacity for explanatory power’ (Dörnyei, Reference Dörnyei2005, p. 104) across a wide range of second and foreign language learning situations.
2.1.2 Strand 2: Relationships between possible selves and L2 outcomes
A second strand of research has focused on investigating the relative strength of relationships between different components of the L2 Motivational Self System and measures of L2 learning. In the first meta-analysis on this topic, Al-Hoorie (Reference Al-Hoorie2017) meta-analyzed a total of 32 research reports, involving 39 unique samples and 32,078 language learners. The results showed that of the three components of the L2 Motivational Self System (the ideal L2 self, the ought-to L2 self, and the L2 learning experience), the ideal L2 self had stronger relationships with L2 development, though associations with subjective intended effort based on learner self-report were stronger than those obtained with objective measures of achievement such as tests of grammatical knowledge (rs = .61 versus .20, respectively). This is in line with previous meta-analyses reporting that correlations with motivational variables tend to be higher with learner-reported effort to learn the TL compared to objective performance (Masgoret & Gardner, Reference Masgoret and Gardner2003).
In terms of the two core motivational constructs, the ideal L2 self, conceived within Dörnyei's sociodynamic model, and integrativeness, conceived within Gardner's socioeducational model, exhibit different patterns of associations with subjective and objective measures of L2 performance. For example, Ryan (Reference Ryan, Dörnyei and Ushioda2009) reported that for university learners of English who were non-English majors in Japan, the ideal L2 self correlated more strongly with intended learning effort (r = 0.74) than integrativeness (r = 0.61); however, relationships for English majors were similar. Kormos and Csizér (Reference Kormos and Csizér2008) similarly reported strong empirical support for the ideal L2 self relative to integrativeness in a large sample of native Hungarian speakers learning English in secondary, university, and adult foreign language learning contexts. Specifically, they found that language learning attitudes and the ideal L2 self were the main factors affecting L2 motivation, and correlations with integrativeness were substantially lower than those found with the ideal L2 self for each learner population.
Research using objective performance measures has also confirmed the relatively more robust role of the ideal L2 self but found that these relationships varied by level of L2 proficiency and over time for university L2 Spanish learners (Serafini, Reference Serafini2013). That is, while there were few relationships between integrativeness and knowledge of L2 Spanish grammar for beginning, intermediate, or advanced learners, either before, during, or after a semester of instruction, the ideal L2 self exhibited strong associations for learners at beginning and intermediate proficiency at the beginning of instruction. However, these relationships got weaker by the end of instruction. Thus, this study provides novel empirical evidence that the explanatory power of the ideal L2 self construct was not only more robust than integrativeness for this learner population, but the positive effects were found to diminish with increasing proficiency and over time, which elucidates the relative capacity of different motivational constructs to explain differential language learning success over time (MacIntyre et al., Reference MacIntyre, Mackinnon, Clément, Dörnyei and Ushioda2009).
Compared to monolingual L2 and foreign language learning settings, research investigating self-related phenomena is less common in multilingual contexts where learners speak and learn multiple languages, with some key exceptions (Henry & Cliffordson, Reference Henry and Cliffordson2013; Lasagabaster, Reference Lasagabaster2016). This is an important gap to fill given the key role of social factors as well as the domain-specificity of the self, meaning that ideal language-speaking/using selves are likely to vary by language.
3. Original studies and proposal for replication
3.1 Study 1: Henry and Cliffordson (Reference Henry and Cliffordson2013)
Henry and Cliffordson's (Reference Henry and Cliffordson2013) study was carried out in the multilingual context of Sweden, where formal study of at least two foreign languages is required, or compulsory, at the secondary level (Henry, Reference Henry2011). The primary motivation for the study was the overall lack of systematic research on the impact of learner ID factors in L2 motivation, specifically gender. However, previous research indicates that females and males consistently differ in measures of dominant motivational constructs (i.e., integrativeness and possible selves) (e.g., Gardner & Lambert, Reference Gardner and Lambert1972; Sung & Padilla, Reference Sung and Padilla1998; Henry & Apelgren, Reference Henry and Apelgren2008; Henry, Reference Henry2009; Ryan, Reference Ryan, Dörnyei and Ushioda2009), which the authors suggest indicate ‘deep-rooted differences in the ways in which females and males identify with TL speakers and how they envisage future situations involving interpersonal communication’ (Henry & Cliffordson, Reference Henry and Cliffordson2013, p. 272).
With these empirical gaps and context in mind, Henry and Cliffordson (Reference Henry and Cliffordson2013) aimed to test the hypothesis that gender impacts the construal of L2 and L3 selves in a multilingual adolescent population. They specifically asked whether gender-related differences in adolescent ideal L2 and L3 language-speaking/using selves are mediated by an interdependent self-construal (ISC) that emphasizes connectedness with others as well as whether any effects of gender varied by the L3 under study. The study was carried out in Sweden where multilingualism is arguably the norm, given compulsory study of at least two foreign languages in school. However, the researchers argue that assigning a ‘foreign language’ status to English is problematic in modern Sweden because the nature of its use has changed over the last decade from mostly being used in school to now being ‘part of a basic social literacy and a medium of expression used extensively in everyday life, particularly among young people’ (p. 277). Moreover, Sweden is a unique setting in terms of gender dynamics as it was ranked first in gender equality among 187 nations in the United Nations Human Development Index, making any gender differences more difficult to detect if they were to exist.
Participants were 269 high school seniors (140 female; 129 male; 15.5 mean age) enrolled in six different schools. They were all studying English (L2) as well as one additional foreign language (L3), including Spanish (n = 136), German (n = 96), and French (n = 33). To measure the constructs of interest, participants completed three 5-point Likert-scale, self-report questionnaires. Administered in the L1, these had been pilot-tested before being distributed in students’ classrooms or school assembly halls. Measures included the Singelis ISC scale (12 items) (Levine et al., Reference Levine, Bresnahan, Park, Lapinski, Wittenbaum, Shearman and Ohashi2003) and the Ideal L2 (English) Self scale (IL2S-5 items) and the Ideal L3 (Spanish/French/German) Self scale (IL3S-5 items), both adapted from Ryan (Reference Ryan, Dörnyei and Ushioda2009). The ideal self-scales showed acceptable internal consistency based on Cronbach's alpha, though the reliability for the ISC scale was determined to be just below the cutoff. However, based on recommendations in the literature the researchers argue that if model fit is acceptable and latent variables are employed, a low alpha estimate is not problematic.
First, the researchers tested each measure for gender differences and found significant differences for the ISC and the Ideal L3 Self, but not the Ideal L2 Self. Thus, a series of Confirmatory Factor Analyses (CFA) were run to test the hypothesis that gender-related differences are mediated by ISC for the Ideal L3 Self only. Four stages of model construction (see pp. 281–284 for more detail) yielded six models, which were interpreted based on the significance of factor loadings for the three variables (ISC, IL3S, and gender) and standard measures of model fit (i.e., χ2 goodness-of-fit test, the root mean square error of approximation, and the standardized root mean square residual (SRMR)).
The full model with all three variables added showed that a significant proportion of the gender differences in IL3S was explained by the gender difference in ISC, indicating that the significant effect of gender on the ideal L3 self is mediated by the extent to which they imagine the self in relation to others. Importantly, this result held regardless of whether students were studying L3 Spanish or L3 German (due to low numbers of students, CFA procedures were not carried out for L3 French learners).
3.1.1 Approach to replication
Henry and Cliffordson's findings valuably shed light on the inherent complexity and multifaceted nature of self- and identity-related phenomena in a multilingual setting, specifically how ideal learner selves vary based on gender, which they found to be linked to differences in self-construal, or how females and males relate to TL speakers and the elaborateness with which they imagine these interactions. However, as the authors conclude, whether similar results would obtain in different sociocultural contexts and with other learner factors is unknown.
To investigate whether the outcomes of Henry and Cliffordson's study can be generalized to a new context, a conceptual replication approach is recommended. Specifically, the target study should be carried out in a different sociocultural context in which the relevant TL(s) may be associated with different degrees of social and public prestige as a majority or minority language, which can systematically impact the ways in which learners imagine the TL and identify with, or desire to identify with, its users. In contrast to Sweden where multilingualism is the norm and English holds international ‘lingua franca’ status as well as being a common language of everyday use, other TLs in other settings present unique complexities with the potential to impact learners’ imagined language-related self-identities.
For example, in the U.S., societal and governmental support for bi/multilingualism in general, and foreign language study in particular, has a long history of being inconsistent, with a tendency to reflect the geopolitical climate of the time (MLA Report, 2007). Spanish, in particular, is a complex TL that, in spite of deep sociohistorical roots in the U.S. and status as the most spoken and studied minority language, it nonetheless remains socially stigmatized due in part to an association with (illegal) immigration (e.g., Leeman, Reference Leeman and Lacorte2014). However, its perceived economic and social value tends to vary by region (e.g., Lynch, Reference Lynch and Roca2000). Thus, findings in Henry and Cliffordson can be strengthened by investigating gender-related differences in the ideal L2 and L3 language-speaking/using selves in learners studying Spanish in the U.S.
Rather than manipulating context, another way to conceptually replicate the target study would be to alter the independent variables under focus (i.e., gender) and the way they are measured. In Henry and Cliffordson's case, gender tends to be conceptualized in a static, binary manner whereas a more dynamic learner ID that is highly subject to changes in environment is attitudes toward the TL and the TL-speaking community, which would ideally be measured both quantitatively (e.g., via the Attitudes and Motivational Test Battery − Gardner & Lambert, Reference Gardner and Lambert1972) and qualitatively (e.g., semi-structured individual interviews). Modifying the learner ID under focus from gender to learner attitudes would align with a more contextually situated view of learner selves.
Finally, attempting to generalize results and strengthen validity and confidence in the findings of the original study can also be accomplished by modifying different methodological aspects of the study's design (Porte, Reference Porte2012). For example, the instrument intended to tap the ideal L2 self in Henry and Cliffordson elicited responses to only five Likert-scale items. As the authors themselves note (p. 287), more elaborated batteries of items are needed to improve the validity and reliability of the findings. Researchers should also pay attention to the wording of self-report items to avoid being vague, which is an issue in some of the original items (e.g., I often imagine myself as someone who is able to speak English). Specifically, it is not entirely clear ‘who’ exactly learners are assumed to imagine (e.g., are the TL speakers referred to in local communities or abroad? Are they monolingual or bi/multilingual speakers? What dialects of the TL do they readily use in communication?). These questions potentially impact learner self-report, and thus, the interpretation of findings and conclusions that are drawn.
3.2 Study 2: Lasagaster (Reference Lasagabaster2016)
Whereas Henry and Cliffordson (Reference Henry and Cliffordson2013) was carried out in a more traditional EFL setting (e.g., Hungary, Japan, China, Iran) where L2 English is studied as a compulsory subject, Lasagabaster (Reference Lasagabaster2016) aimed to address the lack of motivation research conducted in multilingual contexts where learning English as an L3 is voluntary and used to teach content (e.g., EMI). To contribute to filling this gap, Lasagabaster investigated the nature of the relationship between EMI, motivation, and possible selves in L3 English and whether such a relationship is mediated by gender and students’ first language (L1) (Basque, Spanish, or both).
Participants were 189 university students (88 male; 99 female; 2 unspecified; 20 years 8 months; mean age) voluntarily studying in an EMI context in Spain (Basque country). In terms of native language or L1, 20 students reported Basque, 105 reported Spanish, 39 reported both, and 23 reported different L1s (the latter group was not included in the analyses). Over a third of the sample reported having previously spent at least three months abroad in an English-speaking country; almost half perceived themselves to be at an advanced level of English proficiency while half placed themselves at intermediate proficiency level, based on levels described by the Common European Framework of Reference (CEFR). Of the project participants, 8% rated themselves at lower intermediate proficiency level. No participant was enrolled in a language-related degree (e.g., English Studies) to avoid selection bias.
To measure constructs of interest, students completed a 6-point Likert-scale, self-report questionnaire (67 items) that was adapted from Taguchi et al. (Reference Taguchi, Magid, Papi, Dörnyei and Ushioda2009) and pilot-tested with a similar student sample; these students did not participate in the study. A total of ten factors were measured including criterion measures, or intended learning effort (four items), the ideal L2 self (five items), the ought-to L2 self (four items), family influence (four items), instrumentality-promotion (five items), instrumentality-prevention (four items), attitudes to EMI (four items), cultural interest (four items), attitudes to L2 community (four items), and integrativeness (two items).
A multifaceted statistical approach was taken to pursue the five objectives guiding the study. Similar to what has been previously reported for the ideal L2 English self, the ideal L3 English self and integrativeness were significantly positively correlated, confirming that ‘the two variables were tapping into the same construct domain’ (p. 324). Moreover, the ideal L3 self correlated more strongly with criterion measures, or learners’ intended efforts, than did integrativeness, reflecting previous research. Second, the ideal self correlated more strongly with an instrumentality-promotion focus whereas the ought-to self had a higher correlation with instrumentality-prevention, which supports theoretical arguments and previous empirical findings. Third, of the ten factors measured, the ideal self and attitudes to EMI contexts had the strongest correlations with criterion measures and a multiple regression revealed that four factors (the ideal self, attitudes to EMI, family influence, and instrumentality-promotion) accounted for 46% of the variance in learners’ reported learning effort. Finally, a series of t tests revealed significant gender differences for some factors and a one-way ANOVA yielded a significant difference among L1 groups in only one of the ten scales (cultural interest). However, the effect size was small in both cases (eta squared = 0.02–0.04), leading the authors to conclude that gender differences in motivation are neutralized in EMI contexts and that native language does not influence the imagined ideal L3 self of EMI students.
3.2.1 Approach to replication
Lasagabaster's study offers novel insights regarding motivation to learn English as an L3 in Spain, a multilingual setting where multiple languages are in contact and foreign language study is fostered and supported from the top down through multilingual language policies (see Lasagabaster, Reference Lasagabaster2017). Importantly, findings not only empirically support the relevance of the ideal L3 self and attitudes toward EMI for the formation of imagined future selves in university learners but also indicate that gender and native language are neutralized in this learning context. The latter is attributed to the ‘rapid spread of English and the awareness of the need to learn the current lingua franca’ (p. 328), which indicates an attitudinal shift of L1 Basque speakers who were previously reported to regard English as a threat to the minority language and regional identity (Lasagabaster, Reference Lasagabaster2004). As the author argues, this shift underscores the need for attitudinal and motivational studies to be carried out ‘at different points in time if a clear picture of a particular context is to be obtained, because changes may take place in just a decade’ (p. 328).
To assess whether Lasagabaster's findings hold true at different points in time over different timescales (e.g., days, weeks, months, semesters, years), researchers should modify the cross-sectional design of the study to include repeated points of data collection which would enable a longitudinal perspective of self-related phenomena to emerge as well as the variable impact of key shifts in global and local sociocultural factors (e.g., attitudes toward English). Moreover, similar to Henry and Cliffordson, the low number of Likert-scale items on the ideal and ought-to self scales (n = 5) should be increased to strengthen validity and the wording of self-report items should be clarified in terms of ‘who’ participants are intended, and assumed, to imagine. Specifically, the wording of the four items intended to measure ‘Attitudes to L2 community’ and the two measuring ‘Integrativeness’ is problematic (p. 323) as ‘people from English-speaking countries’ or ‘its speakers’ are portrayed in a monolithic manner.
Similar to the suggestions for conceptually replicating Henry and Cliffordson's study, Lasagabaster's results should be tested in a different sociocultural context and in relation to a TL or language(s) other than English. Though the schooling system differs by region, Spain may be characterized as a multilingual context where children are exposed to several languages in school, learning content through Spanish as well as a regional minority language like Basque or Catalán and English (Lasagabaster, Reference Lasagabaster2017). A key question is whether Lasagabaster's results would be replicated in contexts where students have less access to learning other language(s) in school and where the L2 or L3 is not perceived as a lingua franca. As indicated above, the U.S. provides a distinct contrast to Spain in terms of consistent support and access to foreign language study in the public school system, as well as the social, cultural, economic and political value associated with different minority languages. Thus, researchers should attempt to replicate Lasagabaster's study in the U.S. with L1 English learners of Spanish as an L2 or L3. To go one step further and shed light on the degree to which Lasagabaster's results hinge on factors related to the institutional and pedagogical context, other variables might also be manipulated such as the level of study (e.g., primary or secondary versus a tertiary setting) and language teaching approach employed in the classroom (e.g., a traditional grammar-based approach versus a more communicative task-based approach).
Additionally, future replication studies can better isolate the effects of independent variables by improving efforts to control learner background variables such as time spent abroad and self-reported L3 proficiency. For example, in Lasagabaster's study, one third of the sample reported spending at least 3 months in an English-speaking country and almost all participants perceived themselves to be at an advanced or intermediate level of L3 proficiency. The question of whether similar results would obtain with a sample of learners who had not spent any (or minimal) time abroad is particularly relevant given the implication that less time abroad likely means that learners have less elaborate and vivid references of target language speakers and interactions with them to draw on, with implications for how they construct or envision their future self-guides. Moreover, previous research has reported more variable, less stable self-constructs in less proficient learners compared to their advanced peers (Serafini, Reference Serafini2017), which calls into question whether Lasagabaster's results would be replicated with learners on the lower end of the proficiency spectrum.
Finally, instruments used to measure L2 and L3 proficiency and performance in Lasagabaster's study hinge on learner-reported effort. If feasible, a future replication study should aim to independently assess proficiency and performance (Tremblay, Reference Tremblay2011) in addition to, or in lieu of, self-reported data in order to lessen the potential for learner bias. Importantly, employing objective measures can aid in answering the outstanding question of the relative strength of relationships between motivational constructs and subjective versus objective assessments (e.g., MacIntyre et al., Reference MacIntyre, Mackinnon, Clément, Dörnyei and Ushioda2009).
4. Conclusions and future directions
In pursuing robust research designs and sound methodological criteria, replication studies can advance theory and increase the explanatory power of key variables (Porte, Reference Porte2012; Porte & McManus, Reference Porte and McManus2019). The current paper has suggested several ways to conduct an approximate or conceptual replication of Henry and Cliffordson (Reference Henry and Cliffordson2013) and Lasagabaster (Reference Lasagabaster2016), with the immediate goal of assessing the reliability and validity of their findings with an eye toward gauging the generalizability of the L2/L3 self as conceived within Dörnyei's L2 Motivational Self System framework (Reference Dörnyei2005, Reference Dörnyei, Dörnyei and Ushioda2009).
However, making valid comparisons between original studies and replicated studies is tricky as ‘a replicated study can never really be the same, be it repeated by the same researcher in the same context, or others. Time intervals play their part, as do participant variables, changed contexts, and a host of other variables’ (Porte & McManus, Reference Porte and McManus2019). Nonetheless, the value of replication lies in gradually and methodically building up comparative evidence, which is sorely needed to address the key question of the degree to which learner possible selves are generalizable across diverse social and educational contexts and in relation to learning TLs other than English. Context is key in this particular strand of research, given that the status and availability of the TL outside the classroom likely interacts with the way learners perceive, imagine, and construe their possible selves. Specifically, factors such as the perceived social prestige and sociopolitical status of the TL(s) under consideration likely affect the degree to which learners have access to and will pursue opportunities for interaction outside the classroom as well as their views or attitudes toward the TL and its speakers.
Given the key theoretical role assigned to the TL-speaking community in the development and articulation of learners’ imagined future self-states, the degree to which the self is a contextually situated, dynamic, emergent construct (versus being a more stable, universal construct) remains an essential question for future research to address. ’Quantitative replication of studies like Henry and Cliffordson (Reference Henry and Cliffordson2013) and Lasagabaster (Reference Lasagabaster2016) is a good first step to answering this question but is not likely to be sufficient in the long-term without complementing replication with qualitative and mixed methods approaches to data collection. In particular, as Henry (Reference Henry, Dörnyei, MacIntyre and Henry2015) has previously argued, quantitative methods portray the concept of current and ideal selves as highly cognitive, abstract, and somewhat static in nature, ‘presenting them as photographic stills rather than moving pictures’ (p. 93). Thus, cross-sectional data are inadequate for revealing the degree to which self-representations are stable across situations (Dörnyei, Reference Dörnyei and Macaro2010) versus being continuously modified through social interaction and the concomitant processes of ‘self-perception, social comparison and self-appraisal’ (Henry, Reference Henry, Dörnyei, MacIntyre and Henry2015, p. 89). Therefore, moving forward, a longitudinal, mixed methods approach that incorporates both quantitative and qualitative methods of data collection would be ideal including self-report questionnaires, interviews, observations of learner interaction inside and outside the classroom, and stimulated recalls that occur at repeated intervals over the desired timescales (see de Bot, Reference de Bot, Dörnyei, MacIntyre and Henry2015b). Moreover, in line with the person-in-context approach proposed by Ushioda (Reference Ushioda, Dörnyei and Ushioda2009), researchers should endeavor to conduct small-scale case studies which better align with the current concept of the self as a dynamic, multidimensional construct that is contextually situated across time and space (Mercer Reference Mercer2011a, Reference Mercer and King2016) and can reveal the individual variability and dynamic complexity of learner self-perceptions over time in both instructed and naturalistic learning contexts.
Ellen J. Serafini (Ph.D., Georgetown University) is an assistant professor of Spanish Applied Linguistics in the Department of Modern and Classical Languages at George Mason University. Her research interests primarily center on the dynamic role of learner individual differences in a variety of learning settings, task-based approaches to second and heritage language learning and pedagogy, and the long-term outcomes of different models of bilingual education for minority language speakers. Her work in these areas has been published in peer-reviewed book chapters and journals such as Studies in Second Language Acquisition, Modern Language Journal, Hispania and Foreign Language Annals as well as via popular online outlets like Teacher's Hub. Dr Serafini has 15 years of experience teaching Spanish at the secondary and post-secondary level and currently enjoys designing and teaching a variety of graduate and undergraduate courses at George Mason University.
