The relative importance of peace of mind, grit, and classroom environment in predicting willingness to communicate among learners in multi-ethnic regions: a latent dominance analysis

Abstract

From a holistic perspective of positive psychology, there is a dearth of knowledge regarding how its three underlying factors (i.e., positive subjective experience, positive individual traits, and positive collective institutions) co-shape learners’ willingness to communicate (WTC), a critical facilitator of foreign language learning achievement. This research gap is particularly evident for learners in multi-ethnic regions who have been underrepresented in foreign language education in China. This issue may constrain our understanding of the contributions of positive psychology to the field of applied linguistics, since the role of positive psychological factors in influencing WTC may be distinct across diverse ethnic populations. Besides, the relative importance of these three factors in predicting WTC has yet to be investigated in latent models. Given that different pedagogical approaches may engender disparate perceptions and attitudes among learners, it is of the utmost importance to ascertain which factor should be prioritized in classroom psychological interventions, as well as in teacher training programs. To address these gaps, this study addressed the joint effect of positive subjective experience (foreign language peace of mind, FLPOM), positive individual trait (language-specific grit), and positive collective institution (classroom environment) on WTC using structural modeling analysis. Furthermore, the study employed latent dominance analysis to ascertain the relative importance of these three factors in promoting WTC. The sample consisted of 643 multi-ethnic foreign language students from five provinces in Western China. The findings suggested that FLPOM, grit, and classroom environment collectively stimulate learners’ WTC. Notably, FLPOM, a factor that has not previously been examined in relation to in-class or face-to-face WTC, emerged as the most statistically significant predictor of WTC. Therefore, it is imperative that foreign language practitioners and learners recognize the significance of FLPOM in language learning and teaching in Chinese multi-ethnic regions.

Introduction

Willingness to communicate (WTC) represents the decision of language learners to either communicate in a second/foreign language (L2/FL) or remain silent [1]. It is the immediate determinant of actual communication, which is at the core of L2/FL learning [1]. Evidence suggests that high levels of WTC are beneficial to L2/FL learners’ language development [2] and closely related to more target language utterances under task-based education [3], more L2 flow experience [4], greater motivation [5], and better language achievement [6].

From an interactive perspective, WTC embodies the interconnection among learner psychology, the learning process, and the target language acquisition [7]. In line with the tenets of positive psychology, which emphasize the importance of flourishing for individuals [8, 9], an increasing body of research has demonstrated a correlation between WTC and a range of non-cognitive precursors, including foreign language enjoyment (FLE), language-specific grit, and positive classroom environment (CE) [10,11,12,13,14].

While WTC has been extensively studied, several research gaps persist. First, successful foreign language acquisitions require language learners to incorporate positive emotions, strong motivation, and harmonious rapport (e.g., foreign language peace of mind [FLPOM]; grit; positive CE) in the FL learning and communication process [7]. By definition, grit refers to perseverance and passion for long-run goals [15], and CE refers to learners’ perceptions of teacher support, student cohesiveness, and task orientation [16]. Although grit and CE have been associated with learners’ communicative intention (e.g., [13, 17]), to our best knowledge, no studies have explored the association between FLPOM (proposed by [18]), a Chinese culture-specific positive emotion, and WTC in in-class or face-to-face settings. Furthermore, existing research has not concurrently modeled these three positive psychological precursors and verified their patterns of influence on WTC, particularly from the perspective of the three pillars of positive psychology (i.e., positive experience, character strengths, and positive institutions) [9]. This issue may impede a critical understanding of the WTC framework because such WTC determinants may intricately interact with each other [8], subsequently leading to variation in learners’ communication engagement.

The second research gap pertains to the relative importance of the three psychological factors (i.e., FLPOM, grit, and CE) in predicting WTC, which traditional regression analyses are unable to accurately ascertain [19]. This issue may restrict the capacity of administrators, educators, and learners to prioritize specific psychological aspects in teacher-training initiatives or curriculum design, particularly when considering cost and efficiency. Implementing different strategies for capitalizing on these aspects may yield different learner attitudes and perceptions regarding FLPOM, grit, psychological classroom settings, and subsequent WTC. To validate this issue with quantitative evidence for importance orderings in predicting WTC, latent dominance analysis—an objective, stringent, but less-adopted analysis in applied linguistics—was employed in this study.

The third gap concerns the under-represented population of learners in multi-ethnic regions of China. China has 55 minority ethnic groups, whose population accounts for approximately 9% of China’s total population [20]. Yet, those minority ethnic populations have received much less attention in the context of L2/FL education [21, 22]. This issue may limit the valuable insights of FLPOM, grit, and CE, as their impacts may vary across different cultural, and socioeconomic backgrounds, in addition to other potential variables [13, 23]. Furthermore, from the view of learning outcomes, among limited studies that focused on ethnic minority students in L2/FL acquisition contexts, a primary emphasis has been placed on learners’ academic achievement (e.g., [21, 22, 24]). By contrast, multi-ethnic students’ non-linguistic FL learning outcomes of WTC remain under-researched, especially from the perspective of positive psychology. Moreover, there is a notable discrepancy in the allocation of undergraduate and postgraduate education between ethnic minority and majority students in high-stakes, test-driven learning contexts in China. Specifically, ethnic minority students constitute only 9.1% and 5.6% of the total student population at the undergraduate and postgraduate levels, respectively [25]. Therefore, based on a positive psychological perspective, it is highly beneficial to the advancement of equity and high-quality higher education for English-as-a-Foreign-Language (EFL) learning in ethnic minority groups.

Hence, this study attempts to examine the concurrent association between the less-studied positive emotion of FLPOM, language-specific grit, CE, and WTC among less-represented FL learners in multi-ethnic regions of China. Furthermore, the relative importance of these three potential precursors of WTC is evaluated through latent dominance analysis along with Bayesian testing.

Literature review

Theoretical framework of this study—Positive psychology and its three pillars

Positive psychology in the general domain, as defined by Peterson [26], is the scientific study of “what goes right in life” (p. 4). This approach contrasts with traditional psychology, which has historically focused on deficiencies and diseases. Positive psychology is structured around three fundamental pillars: positive subjective experience (e.g., enjoyment and well-being), positive individual traits (e.g., grit and trait curiosity), and positive collective institutions (e.g., community and classroom) [27].

Within the domain of applied linguistics, underpinned by the well-being theory [9] and humanistic principles, positive psychology can be effectively implemented by language educators to promote not only learners’ FL development but also their emotional well-being [28]. The implementation of positive psychology is potentially associated with a multitude of FL outcomes, both linguistic and non-linguistic, including lesser stress, enhanced rapport, pleasant emotion, and improved well-being (WTC in this study; see an overview of such outcomes in the appendix of [29]). However, during such processes, the Three Pillars of positive psychology—positive experience, character strengths, and empowering institutions—play related, yet distinct roles in motivating improvement in language outcomes [8].

The first pillar of positive psychology is positive experience, which encompasses positive emotions and well-being. Building on Fredrickson’s [30] broaden-and-build theory, positive emotions are viewed as a way to achieve long-term psychological development, intellectual advancement, and enhanced well-being [30, 31]. The second pillar of positive psychology, character strengths (e.g., grit), allows us to exploit individual differences [8]. Compared to other school subjects, character strengths (e.g., grit) may have a more significant influence on FL learning, which involves more challenges, difficulties, lengthy processes, and failure, as well as discouraging, anxious, bored, and shy experiences [11, 32, 33]. Positive institutions (e.g., CE), as the third pillar of positive psychology, suggest approaches to setting up institutions (e.g., classrooms, schools, and families) that empower flourishing [8]. Well-being, affective improvement, and personal progress are contextually shaped by supportive institutions [8, 34]. Further, positive institutions have a solid communicative feature since people and their affective experiences are embedded in social contexts [35]. However, among the three pillars, institutional factors have been the least studied in L2/FL acquisition [36].

The prevailing paradigm of positive psychology, as articulated by the three pillars theory, has served as a foundational framework for numerous studies that have examined the interconnection among the three pillars of positivity and its implications for L2/FL outcomes, including achievement [37] and WTC [12, 13]. These studies emerged as a response to the prevailing cognitive perspective and the disproportionate focus on adverse emotions within the domain of applied linguistics, illuminating the manner in which psychological factors and environments promote individual flourishing [8, 28, 29]. In alignment with this research trajectory, this study adopts a comprehensive positive psychology framework encompassing its three pillars.

Willingness to communicate (WTC)

WTC was initially developed in communication research in first language (L1) contexts and subsequently extended into L2 contexts [38]. In L2/FL education, WTC is defined as “an intention to communicate when an opportunity arises” in the FL [1]. MacIntyre et al. (1998) [38] proposed a seminal six-layered pyramid model of L2 WTC and divided WTC antecedents into two categories, including situational factors (layer I, II, and III; e.g., state communicative self-confidence, and desire to communicate with a specific person), and enduring factors (layer IV, V, and VI; i.e., motivational propensities, affective-cognitive context, and social and individual context). Recently, to examine multilingual individuals’ WTC influenced by situational demands and broader context (i.e., out-of-school community), such a two-dimensional WTC model was revised into 3D graphics [1]. Through longitudinal in-depth case studies, [1] examined instantaneous fluctuations in WTC, communicative behaviors, and language choice between English (L2) and Swedish (L3) of eight women immigrants in a Swedish multilingual community. Such a book-length study indicates the WTC research was expanded into multilingual or community contexts [2].

Since high target language competence does not necessarily result in actual communication [39], WTC is deemed as the vital precondition for actual L2/FL use [40]. Building on the tenet of communicative language teaching, researchers further asserted that WTC is “the ultimate goal of the language learning process” [41]. Along with its paramount importance in target language acquisition is the complexity of WTC. That is, WTC in FL is a dual-faceted construct encompassing trait and situational features. In a very recent review of WTC research developments in 1998–2024, [2] classified six pivotal perspectives influencing WTC theorizing, i.e., social psychological, ecological, dynamic, cultural, digital, and multimodal perspectives. Given the merit and complexity of WTC, a substantial body of literature has sought to identify its underlying determinants, including, but not restricted to, communicative competence [5], anxiety [42], motivation [43], grit [44], motivational intensity [33, 45], classroom environment [16, 46], circumstances of migration [1], learner beliefs [47], enjoyment [10, 12, 46], emotional intelligence [48], AI literacy [49], and AI tools [50]. Additionally, as cognitive factors have been unable to adequately interpret WTC and its variation across individuals and contexts, WTC researchers have recently turned their attention to positive psychological factors. In the following sections, we will present the three psychological constructs: FLPOM, language-specific grit, and CE, as well as their relationships with WTC.

The rationale for validating FLPOM, language-specific grit, and CE as WTC predictors

Foreign language peace of mind (FLPOM) and WTC

FLPOM, which represents the pillar of positive emotions in our theoretical framework of positive psychology, is conceptualized as a positive learner emotion of inner peace and harmony in the Chinese FL context [18]. It is indicated that inner peace captures low-arousal positive (LAP) emotional states such as calmness and peacefulness, while harmony captures a state of internal balance both within oneself and between oneself and the environment. POM entails emotional regulation and control [51] and has been found to be closely linked to individuals’ basic psychological needs for autonomy, relatedness, and competence in the Chinese context, relative to high-arousal positive (HAP) emotions [52] suggested that an individual with a high level of POM is “an integrated self with life energy in a latent form” [52].

Based on Fredrickson’s [30] broaden-and-build theory of positive emotions, [18] proposed that, on the one hand, FLPOM can broaden learners’ thought-action repertoires by creating an urge to focus on what is intrinsically valuable and to downplay task-irrelevant distractions, especially negative ones such as stress or negative feedback; on the other hand, FLPOM can build personal resources by maintaining academic engagement and contributing to resource conservation (with higher resource stability and lower susceptibility to resource loss) due to its superiority in emotional regulation and control. These personal resources can take the form of psychological resilience or self-efficacy. A relevant intervention study found that inner peace training improved mental qualities (i.e., creativity) displayed during classroom activities in primary and secondary schools in the USA [53].

Thus, we can hypothesize that learners with FLPOM have a stable, balanced, and self-regulated emotional state and are better able to avoid the influences of negative internal or external stimuli and maintain focus on the task at hand, and thus are more willing to engage in L2 communication, which is essential for successful language learning. However, there is an urgent need for empirical support for this hypothesis because no prior empirical study has linked FLPOM and WTC, particularly from a positive psychology perspective. The only exceptional published literature is [54], which focused on these two constructs in informal online learning settings and also dealt with FL burnout as an unpleasant emotion and how these two emotions mediated the link between informal digital learning and digital WTC. Apart from the aforementioned research gap, there are compelling reasons to focus on FLPOM and its correlation with WTC in this study. Firstly, the correlations between WTC and certain emotions (e.g., FLE and anxiety) have been well-documented in the existing literature (see meta-analytic evidence in [5, 43]). Secondly, there have been calls to study emotions other than enjoyment and anxiety in both the overall L2 field [14, 55, 56] and WTC research [57, 58], especially because the latter is a multidimensional construct impacted by diverse emotions. Finally, from the standpoint of the culture-specific emotion of FLPOM, the recent call for expanding its nomological network was also considered [23, 59].

Language-specific grit and WTC

As a personality trait falling under the second pillar of positive psychology, grit encompasses two key sub-facets—enduring volition to achieve something challenging despite setbacks (i.e., perseverance) and continuous dedication to a valuable and long-term goal (i.e., passion). The passion facet and expanded timeframe distinguish grit from similar constructs, including self-control, conscientiousness, and growth mindset [60].

In target language contexts, L2 grit promotes better L2 accomplishment [61], including vocabulary knowledge [62, 63], pragmatic comprehension [64], L2 proficiency in the overall [65, 66], and specific levels of speaking, listening, and grammar [67]. It is further argued by Teimouri et al. [67] that L2 grit’s role in affecting FL achievements was as important as talent in the form of L2 aptitude. Moreover, in addition to facilitating language proficiency, studies evidenced that L2 grit was also correlated to various motivational and psychological factors, including higher intended effort, greater growth mindset, lower anxiety, lower fixed mindset, higher engagement, and enhanced WTC [68,69,70]. Additionally, empirical evidence demonstrated that language-specific grit outperforms conscientiousness, trait intelligence, and domain-general grit in predicting L2 learning outcomes, including L2 proficiency/motivation/engagement [61]. Considering this as well as Botes et al.’s [61] advocacy for more domain specificity of grit, language-specific grit rather than domain-general grit was therefore employed in this study.

As discussed above, aligned with the positive psychology tenets, grit as a personality trait may promote learners’ WTC, which can be considered as a vital non-linguistic outcome of FL learning [see literature evidence in 12, 13]. However, existing literature demonstrates inconsistent findings on the direct path from academic grit to WTC. For instance, [13] stated that grit saliently contributed to WTC among two Arabic dialect samples. In contrast, via a double mediation analysis, [11] revealed only indirect pathways from language-specific grit to WTC. These inconsistent findings highlight the necessity for further investigation into the role of grit in influencing WTC, especially in the context of under-researched Chinese learners in multi-ethnic regions. Another rationale for the selection of grit in the present study is that, compared with learning in other subjects, FL learning is a protracted process, and replete with unanticipated failures, challenges, setbacks, hardships, as well as frequent communication failures accompanied by losing face and discomfort. Considering this, grit in conjunction with continuing perseverance and consistent passion may play an even stronger influence in FL acquisition and use than in other subjects and learning processes [68, 69], and deserves more empirical investigations [71, 72].

Classroom environment (CE) and WTC

Positive CE belongs to the positive psychology pillar of institutional factors. Positive CE, a three-faceted construct, is characterized by consistent support from teachers to learners, positive student interaction, and well-organized and meaningful classroom activities [10]. In a broad sense, CE can be defined as the “assessment of the psychological, social, emotional, and organizational/managerial state of the classroom” [73].

In alignment with the tenets of positive psychology, we argued that the psychological classroom environment may contribute to learners’ FL development, as well as their well-being and personal development. Literature has documented that positive CE, as an enabling institution and primary platform for EFL learning, may benefit FL learners’ WTC in various contexts [10, 74, 75]. Nevertheless, there is a paucity of quantitative WTC literature on language learning that explicitly acknowledges CE as a potentially positive institution (e.g., ecological perspective in [76]). Furthermore, from a positive psychology perspective, positive institutions are the least explored among the three positive psychology pillars in L2/FL learning [36]. Therefore, our understanding of the applicability of positive institutions in WTC research is relatively limited.

In addition, dissimilar results on the direct association between CE/CE-related factors and WTC were reported in extant literature. Specifically, two studies by Wang and colleagues [14, 77] posited that class social climate (CSC) had only an indirect effect on WTC in Chinese and English L2 learning. In contrast, it was also indicated that CE is an immediate factor related to WTC [e.g., 10]. Given these mixed results, further scrutinizing classroom climate’s role in communicative intentions is needed, particularly among the under-explored Chinese learners in multi-ethnic regions. Apart from both theoretical and empirical concerns, there are several other reasons to examine CE, instead of other positive institution variables, as a means of promoting WTC in this study. Firstly, for EFL learners in exam-oriented learning contexts, the predominant site for their FL learning is the classroom [78], instead of other settings (e.g., communities, cram schools, digital, and out-of-class settings). Secondly, [79] explicitly suggested that peer support, and classroom climate instead of merely teacher support should be taken into account in FL communication theorizing. The multi-faceted construct of CE, involving supportive teachers, favorable peers, and well-organized classroom tasks and activities, was therefore considered in this study to address such a research call.

FLPOM, grit, and CE as co-predictors of WTC

In addition to the pivotal role of positive psychology in FL education, as previously discussed, a holistic lens of positive psychology integrating its three pillars of positivity is feasible and necessary in WTC research. The current investigation is based on the following theoretical and empirical considerations.

Firstly, WTC is a multi-faceted construct susceptible to volitional, emotional, and contextual factors [2]. MacIntyre and his colleague claim that the target language is acquired when learners integrate their emotions, motivation, and relationships with teachers and peers (e.g., FLPOM, grit, and positive CE) into each move of the learning and communication process [7]. Secondly, the merit and necessity for considering the three core facets of positive psychology in WTC research were further pointed out by researchers—“models of the learning and communication process are incomplete without explicit consideration of positive emotions, individual strengths, and the various institutions and contexts of learning” [80]. The last consideration is inspired by the L2EPP (second language emotions and positive psychology) theoretical model [81]. Based on Pekrun’s [82] control-value theory and Fredrickson’s [30] broaden-and-build theory, [81] developed an emotion-centered theoretical model that integrates the three pillars of positive psychology (positive emotions, character strengths, and empowering institutions) with causes, effects, and intervention of L2 emotions [81]. The model suggested that positive institutions, personality traits, and pleasant emotions may directly or indirectly (via pleasant emotions) co-impact target language learning outcomes, such as WTC. Hence, this study adopts a holistic approach that integrates the three pillars of positive psychology to understand WTC.

Despite the flourishing of positive psychology in the L2/FL field, only a few studies have employed such a holistic perspective. Specifically, only two of the three pillars of positivity were jointly investigated in those WTC studies. For instance, in a recent study, the longitudinal role of one pleasant emotion (language anxiety) and the cross-sectional effect of one character trait (grit) on WTC were confirmed [83]. Similar to that study, three other studies [12, 13, 57] investigated the role of positive emotions (e.g., enjoyment and boredom) and personality traits (grit and growth language mindset), yet neglected the value of positive institutions in promoting learners’ willingness to talk. Notably, there is only one empirical study [84] that has jointly examined the significance of the three essential facets of positive psychology (i.e., teacher support, grit, and enjoyment) in WTC in a single research.

However, to the best of our knowledge, no previous literature has included FLPOM, language-specific grit, and CE in a single study and examined their potential co-shaping role in learners’ WTC, especially from a comprehensive perspective that merges three positive psychology pillars. This study proposes a multifaceted conceptual model to elucidate the mechanisms underlying the growth of WTC among multi-ethnic FL students. The model integrates the concept of FLPOM as a pleasant experience, FL-specific grit as a character strength, and CE as an enabling institution. By examining the interplay between these three antecedents from a holistic positive psychology perspective, this study aims to offer nuanced insights into the growth process of WTC among multi-ethnic FL students.

The relative importance of predictors of WTC

As previously reviewed, some extant literature has revealed the predictive effects of FLE [though different, yet related to FLPOM; 23], FL-specific grit, and CE in WTC, though inconsistent findings in relatedly proximal patterns were also reported [e.g., 11, 12, 14]. In the majority of those studies, FLPOM-related FLE exerted a stronger effect size in promoting WTC, compared with FL-specific grit [12, 13] or CE [10] in a way of pairwise comparison. For instance, as observed in the study of [12], FLE showed a larger effect size in facilitating WTC in comparison to domain-general grit (β = 0.554 and 0.161, respectively; all ρ < 0.001). Similarly, through parallel triple mediation analyses, [10] suggested enjoyment yielded a stronger effect size in affecting WTC than CE (β = 0.421 and 0.211, respectively; all ρ < 0.001). However, the effect size difference between such patterns was inconsistent. As one example, in the [57] study, enjoyment showed a modest and pretty similar effect size in contributing to WTC than grit among Saudi EFL students (β = 0.11 and 0.12, respectively; all ρ < 0.01).

However, these findings regarding the stronger effect size of enjoyment in eliciting WTC compared to grit or CE demonstrated no statistical significance. Statistically, to accurately ascertain dominance, examination of predictors’ relative importance through dominance analyses is indispensable [85]. Traditional effect size measures (e.g., correlation and regression analyses) demonstrate limitations in addressing such an issue, particularly given inter-correlated predictors [86]. In those cases, calculating standardized regression values relies on other predictors within the same model [87]. Further, contrary findings were uncovered, implying that FLE perhaps is not more dominant than grit and CE (e.g., [57, 88]), not to mention that FLE can not replace FLPOM under investigation in this study. Therefore, it has yet to be conclusively determined that FLE or FLPOM is the most dominant antecedent of WTC, nor whether is grit or CE is more important.

This issue of relative importance for latent predictors can be solved using latent dominance analysis [19], which allows us to evaluate the “proportionate contribution that each predictor makes to the R-squared” by"consider[ing] the possible subset models to which a predictor could be added” [19]. Thus, a predictor is considered more important than its counterparts if it contributes more to possible subset models [19]. Further, compared to other importance measures, the dominance analysis method provides an intuitive explanation of predictor importance and defines different modes of relative importance [87, 89]. In the domain of social sciences, dominance analysis was initially developed by [85], expanded by [90], and latterly further extended by [19] to SEM models in the latent variable context. [89] recently introduced dominance analyses for multiple regression models with manifest variables into L2 research and developed an R-rested web application.

Yet, in applied linguistics research, compared with traditional effect size metrics, very little attention has been given to dominance analysis [37, 89] is the initial attempt at using dominance analysis, as they claimed. That opening study focused on three FL emotions’ dominance in predicting FL achievement. In another study, compared with three constructs (i.e., domain-general grit, cognitive ability, and conscientiousness), L2 grit’s dominance was identified in predicting four FL outcomes [61]. In WTC theorizing, only a handful of existing research touched on general dominance analysis. For instance, [32] gave very little weight to using dominance analysis and compared two FL emotions’ (enjoyment and boredom) importance in promoting online in-class WTC. Recently, a non-parametric data study [91] gave some weight to dominance analysis and confirmed intercultural sensitivity’s dominance in triggering out-class face-to-face and digital WTC in two samples from Spain and Poland. In summary, within the context of applied linguistics research, there is a paucity of studies that employ dominance analyses. The extant literature primarily utilizes hierarchical models, multiple regression models involving observed variables [32], or non-parametric analyses [91]. These approaches, however, lack the parsimony and robustness characteristic of structural equation modeling (SEM) approaches, with a single exception [37].

However, to our knowledge, no WTC studies have used dominance analysis with SEM models to calculate the relative dominance of WTC’s latent predictors, let alone a follow-up Bayes testing, which was recently recommended by [86]. An ample reason for the scarce usage of such analysis in this field is that the dominance analysis was only recently extended to the latent regression model in the study of [19] in the general domain. Further, such methodological gaps might lead to inaccurate insights regarding the importance ranking of positive emotions, enabling traits, and favorable learning environment in predicting learners’ communicative intentions and subsequent actual L2/FL use in classes. Thus, in line with Li et al.’s [92] call for more novel methodological attempts in the language-specific domain, latent dominance analysis and Bayesian testing were used in this study.

Research questions

Based on the above-discussed research gaps, this study aims to answer the following research questions:

• RQ1. Do FLPOM, L2-specific grit, and CE jointly predict WTC among Chinese EFL learners in multi-ethnic regions?

• RQ2. Which psychological variable (i.e., FLPOM, L2-specific grit, and CE) plays the dominant role in predicting WTC among Chinese EFL learners in multi-ethnic regions?

Methodology

Research design

We use a questionnaire survey to elicit data on the three positive psychological aspects (FLPOM, language-specific grit, and CE) and WTC in English learners in Chinese multi-ethnic regions. We perform structural equation modelling to investigate the influences of the three positive factors on WTC. Furthermore, we use latent dominance analysis and Bayesian testing to determine the importance ordering of the three factors in influencing WTC. The former was employed because effect sizes in traditional regression analysis can be biased due to their dependence on other correlational variables in the same model [87], whereas dominance analysis can avoid these biases because it takes into account both independent and partial effects of each predictor on the outcome variable [19]. The latter was employed because it allows the use of Bayes factors to “quantify the support from the data for various importance orderings” [19].

Participants with multi-ethnic cultural backgrounds

Previous WTC research has primarily recruited Chinese participants from the ethnic majority group of Han while ignoring the ethnic minorities with diverse cultural and geographic backgrounds, thereby limiting the generalizability of the findings of relevant studies because the predictive effect might fluctuate across distinct ethnic participants. Given the multi-ethnic features of the Chinese population, the current study chose Chinese respondents with multi-ethnic cultural backgrounds to extend insights into the prominence of FLPOM, language-specific grit, and CE in establishing FL learners’ WTC. Additionally, this study focuses on the western region of China, as it is home to 70.22% of the country’s ethnic minority population [93]. In comparison to their counterparts from the dominant Han ethnic group, English is a subject that is comparatively undervalued and underfunded for ethnic minority students in economically less developed western areas of China [94]. Most research on ethnic minority FL learners has focused on how to promote their FL proficiency, with very limited attention given to WTC development, especially based on the tenets of positive psychology.

Stratified sampling was employed to recruit respondents. Initially, 723 respondents were recruited from 7 universities distributed across three regions of Western China: northern, south-central, and southern. By principle, two universities from each region were selected. Nevertheless, three universities were selected from the south-central region, as that region has more universities. All these students were EFL learners and registered for a required College English course. Yet, 80 cases (11.07%) were identified as invalid and removed as they contained random responding, such as straight-lining, non-differentiation [95], and response style [96], as well as failure in response time analysis [95]. As such, 643 emerged as the final sample size, leaving a valid response rate of 88.93%.

Specifically, participants comprise 21 ethnic groups (51.5% Han, 24.9% Zhuang, 9.2% Zang, 2.2% Miao, 1.7% Hui, 1.6% Yao, 1.2% Buyi, 1.2% Dong, 1.2% Tu, 1.1% Yi, and 4.2% others). Among the top 12 minority ethnic groups by population in China [93], 9 of them were involved in the sample of this study. In addition, among all participants, 83% were recruited from universities that serve ethnic minority students and are under the direct or partial jurisdiction of the State Ethnic Affairs Commission of China. In those universities, specific ethnic cultures and customs are highly esteemed, such as recommending folk dances of ethnic groups throughout the whole university [97] and exploratory integrating of campus culture and ethnic traditional culture [98]. Meanwhile, students from those universities are more likely to hold more exposure to multi-ethnic peers. As such, ethnic and cultural factors may continue to influence ethnic minority students’ learning even after they become grown-ups [97]. The detailed demographic information of the participants is shown in Table 1.

Table 1 Demographic information of the participants (N = 643)

Instrument

A two-section questionnaire measured the participants’ background information and the four constructs. To ensure respondents’ complete comprehension, all items were translated into Chinese using backward and forward translation. Additionally, all four subscales were 7-point Likert-type, with 1 showing “strongly disagree” and 7 indicating “strongly agree”. Regarding subscales’ reliability, all variables achieved the threshold with Cronbach’s alpha over 0.70 [99]. These scales have demonstrated a high degree of validity in previous studies and have been extensively utilized to assess the target variables in FL learning. The subsequent instruments were employed.

FL peace of mind (FLPOM)

The FLPOM Scale developed by [18] was used. It is single-dimensional with eight items. One of the items is “I feel peace and comfort in the English class”. The Cronbach alpha reliability coefficient for FLPOM was 0.948.

Language-specific grit

Teimouri, Sudina, and Plonsky [100] claimed the merit of using a domain-specific measure of grit in language learning, that is, it “enhances its predictive and construct validity and better captures its differential effects in various domains and across languages” [100]. Based on such argument, the 10 items of the grit scale used in this study were slightly modified based on Teimouri et al.’s [69] L2-specific grit scale. Nine of the 10 items were slightly adapted from Teimouri et al.’s [69] scale. The last item measuring CI dimension was modified based on Duckworth & Quinn’s [15] scale while considering L2-specific contexts. Items included “I am a diligent English language learner”, and “I think I have lost my interest in learning English”. The Cronbach alpha reliability coefficient for PE and CI was 0.898 and 0.762, respectively.

Classroom Environment (CE)

The Classroom Environment Scale developed by [47] was adapted to gauge the three factors underlying CE, namely teacher support (4 items, e.g., “The teacher is patient in teaching”), students cohesiveness (4 items, e.g., “I am friendly to members of this class”), and task orientation (4 items, e.g., “Tasks designed in this class are useful”). One item measuring task orientation was deleted, because in the study of [47], it showed factor loadings lower than 0.50 [99]. The Cronbach alpha reliability coefficient for teacher support was 0.915, student cohesiveness was 0.901, and task orientation was 0.915.

Willingness to communicate (WTC)

The WTC Scale used in this study was adapted from WTC scales developed by [78, 101]. It measured one dimension with seven items, including 5 items adopted from [101] in measuring in-class WTC and 2 items adapted from [78]. One item included was “I am willing to make a presentation in English in front of the class”. The Cronbach alpha reliability coefficient for WTC was 0.919.

Data collection

This study was first approved by the research ethics committee of the corresponding author’s institution. Then, we contacted EFL teachers from different universities (see Table 1). A web-based survey, accessible on Wenjuanxing (www.wjx.com), was shared with the participants with the aid of those teachers. Before the questionnaire administration, potential participants were verbally notified about the informed consent form and research goal. They consented to participate in the study and were allowed to ask questions about the online survey.

Data analysis

To ascertain the feasibility of the measurement model, we report the factor loadings [acceptable fit > 0.60; 99] for each latent variable in the constructed four-factor model. The factor loadings for each latent variable are presented in Table 2.

Table 2 Factor loadings for each latent variable

We ran the normality test in preliminary analyses and performed descriptive and Pearson correlation analyses through Stata 14. To further validate the factor structure of the constructs, we conducted confirmatory factor analysis (CFA) using the maximum likelihood estimation method. The CFA model included four latent factors: FLPOM, language-specific grit, CE, and WTC. Each latent factor was measured by a set of observed variables, as defined in the instrument section.

The model fit was assessed using several fit indices, including the Comparative Fit Index (CFI), Tucker-Lewis Index (TLI), Root Mean Square Error of Approximation (RMSEA), and Standardized Root Mean Square Residual (SRMR). As Table 3 shows, the model with four discriminant factors fits the data best, with CFI = 0.917, TLI = 0.908, RMSEA = 0.073, SRMR = 0.074, and χ²/df = 1399.208/316 = 4.428. These values are within the acceptable range [see Table 3; 102, 103], indicating that the hypothesized factor structure is supported by the data.

Table 3 Results of confirmatory factor analysis

The factor loadings for each observed variable on its respective latent factor were all statistically significant (p < 0.001), ranging from 0.65 to 0.85. This indicates that the observed variables are reliable indicators of their respective latent factors.

To answer RQ1, SEM was conducted to detect each path from psychological antecedents (FLPOM, language-specific grit, and CE) to WTC co-existed in a single latent model. Maximum likelihood estimation combined with standard error served to test the model [102]. The model fit was assessed using the standards set by the literature [102, 103]. Multiple fit indexes included were χ²/df (acceptable fit < 5.0), Comparative Fit Index (CFI; acceptable fit > 0.90), Tucker-Lewis Index (TLI; acceptable fit > 0.90), Goodness of Fit Index (GFI; acceptable fit > 0.90), Adjusted Goodness of Fit Index (AGFI; acceptable fit > 0.90), Root Mean Square Error of Approximation (RMSEA; acceptable fit < 0.08), and Standardized Root Mean Square Residual (SRMR; acceptable fit < 0.08). Then, path significance was evaluated by ρ value [99].

To address RQ2, latent dominance analysis was performed as discussed above with a three-step procedure suggested by [86]. First, general dominance analysis was employed for each WTC predictor (FLPOM, language-specific grit, and CE). Second, by calculating a 5000 sample of bootstrapping [104], bias-corrected confidence interval (BCCI), Wald test z, and its ρ coefficients for each of the three predictors were obtained. Haven and Bain packages in R and Domin package in Stata were used. In turn, the hypotheses of relative importance were generated by referring to the results of the general dominance analysis, and through a bootstrapping method.

Third, the Bayesian test was also employed in the current study, as such a method allows for quantifying the support from data for diverse importance orderings, particularly when three or more predictors are involved in relative importance assessment [19]. In other words, such a method allows us to verify the proportion of the probability of one hypothesis to another [105]. According to [19], Bayes factors between 1 to 3, 3 to 20, 20 to 150, and > 150 are indicative of negligible, convincing, strong, and very strong evidence from the data for the importance ordering hypothesis, respectively (see Table 5 in [19]).

Overall, the first two steps established the predictors’ basic dominance estimates, while the last step specified the final dominance order of WTC predictors.

Results

Descriptive statistics, correlational, and validity analysis

Table 4 displays the descriptive estimates, the correlation coefficients between all researched variables formulated by the structural equation model, and coefficients regarding reliability and validity. According to skewness and kurtosis values shown in the table, no normal distribution issues emerged in the data [99]. All variables were linked statistically significantly (ρ < 0.001), positively, and medium-to-large (0.40 < around γ < 0.66; [106]). The correlation matrix for latent constructs also paved the way for the latent dominance analysis. In addition, these AVE values are above 0.50, indicating good convergent validity for the latent variables. To assess discriminant validity, we compared the square root of AVE for each latent variable with the correlation coefficients between latent variables. As shown in Table 4, the square root of AVE for each latent variable is greater than the correlation coefficients between that latent variable and other latent variables, indicating good discriminant validity.

Table 4 Results of descriptive statistics, reliability, and validity analyses (N = 643)

Research question 1: Structural equation modeling (SEM)

Figure 1 graphically presents the structural model. Through SEM analyses, all variables’ factor structures were specified as expected. Specifically, both FLPOM and WTC were identified as single-facet latent variables. In sequence, a two-factor structure was specified to assess grit, with the two first-order latent variables, PE and CI, representing a higher-order language-specific grit latent variable. Moreover, CE was also gauged with a higher latent variable suggested by the three lower latent variables of Teacher support, Student cohesiveness, and Task orientation.

Fig. 1

Structural equation model. Note: χ²/df = 2308.934/618 = 3.736, CFI =.911, TLI =.904, GFI =.920, AGFI =.910, RMSEA =.065, and SRMR =.071. All effect sizes are statistically significant at ρ <.001

In addition, the structural model obtained a good fit, with χ²/df below 5.0 (3.736), CFI (0.911), TLI (0.904), GFI (0.920), and AGFI (0.910) over 0.90. Furthermore, both RMSEA (0.065) and SRMR (0.071) were under 0.80 and thereby suggested an acceptable fit [102, 103]. Additionally, the factor loadings generally achieved acceptable indexes (see Fig. 1), with two exceptions of item G3 (0.173) in the lower order and CI (0.437) in the higher order of language-specific grit. Since all the five items underlying CI of grit (including item G3) were negatively keyed (e.g., “I am not as interested in learning English as I used to be”), the lesser factor loadings were in part anticipated. According to [107], item reversal and item negation easily result in unusual ways of responding.

Furthermore, the model denoted that all three positive psychological constructs co-shaped WTC statistically significantly and positively (FLPOM: β = 0.459; language-specific grit: β = 0.340; and CE: β = 0.293; all ρ < 0.001). The sequential order of effect size was FLPOM, followed by language-specific grit and CE. Yet, one can not conclude that one predictor is not dominant at all, even if its regression value was zero, as that predictor’s importance may be concealed by other correlated predictors [87]. Latent dominance analysis was thus carried out to address RQ2 fully.

Research question 2: Latent dominance analysis and Bayes testing

Table 5 demonstrates the findings of the latent dominance analysis. FLPOM had the highest dominance estimate (d_FLPOM = 0.234), followed by language-specific grit (d_Grit = 0.147), and lastly, CE (d_CE = 0.116). In addition, the general dominance indices were ascertained as statistically significant based on the Wald test z and p-values (p < 0.001) together with BCCI not overriding zero.

Table 5 General dominance analysis results for latent variables

On the basis of the general dominance estimates, we then established three hypotheses concerning the predictors’ dominance sequencing. Specifically, Hypothesis 1 was entirely based on the general dominance estimates to determine the importance ranking (see Table 5).

• H₁: d_FLPOM > d_Grit > d_CE.

Considering grit was oft-studied with positive emotion (e.g., FLE) and one of its two dimensions (CI; also named as passion) holds emotional property [e.g., 56], Hypothesis 2 was generated to specifically verify whether language-specific grit might be the most dominant predictor and to establish further the sequential ranking of importance between CE and FLPOM.

• H₂: d_Grit > d_FLPOM > d_CE.

Lastly, since CE (d_CE = 0.116) and language-specific grit (d_Grit = 0.147) achieved close values of the general dominance estimates, Hypothesis 3 was generated to explicitly testify to the extent to which the higher estimate of language-specific grit was a result of statistical bias and whether CE is more dominant than grit in predicting WTC.

• H₃: d_FLPOM > d_CE > d_Grit.

Table 6 presents the results for the Bayes factor. In the comparisons between H₁/H₂/H₃ and its counterpart null hypothesis (H_a), H₁ (d_FLPOM > d_Grit > d_CE) showed the highest Bayes factor (BF_1a = 4.205), which was also the only value falling in the range of 3–20. Based on the recommendation of [19], H₁ obtained the largest support and the only “convincing support” for null hypothesis comparisons from the data. In sharp contrast, H₂ captured the lowest support and should be directly rejected as its Bayer factor was less than 1 [BF2a = 0.025; 19]. In addition, H₃ gained “negligible evidence” from the dataset, as its Bayes factor was between 1 and 3 [BF3a = 1.315; 19].

Table 6 Bayesian test results for dominance ordering hypotheses

In turn, the hypotheses were compared with each other. The support for H₁ was found 3.194 times higher than that for H₃. Additionally, BF₁₂ (160.500) and BF₃₂ (50.250) both further verified the rejection of Hypothesis 2 in support of Hypothesis 1 and 3.

Based on the above analyses, it was found that FLPOM achieved the biggest general dominance estimate among all predictors, and the sequential order of relative importance for WTC predictors was FLPOM, followed by language-specific grit, and lastly, CE.

Discussion

This study is the initial attempt to highlight the value of FLPOM (as a positive experience), L2-specific grit (as a character strength), and CE (as a positive institution) in collectively shaping WTC in a single study, especially from a holistic positive psychology perspective. In addition, WTC among FL students in multi-ethnic regions has rarely been explored from such a theoretical basis. Furthermore, the extent to which latent variables (FLPOM, L2-specific grit, and CE) dominantly contribute to WTC (i.e., relative dominance) remains uncertain.

Regarding the first research question, the results of SEM analyses demonstrated that FLPOM, language-specific grit, and positive CE collectively promoted WTC. That is, multi-ethnic students with a positive state of inner peace and internal harmony, high grittiness, and situated in a supportive and favorable classroom climate were more likely to arouse higher behavioral intention to communicate in FL.

This result partially coincides with previous findings in that any two factors among FLE (a positive emotion that resembles FLPOM), grit, and CE were coherently associated with WTC enhancement (e.g., [10, 12, 44]). For instance, a mixed method study [44] found that classroom enjoyment and grit concurrently promoted in-class WTC among Hong Kong primary students. This finding also in part corresponded to Yu & Xia’s [12] result in that EFL students’ enjoyable experience and grittiness were jointly interconnected with their more classroom speaking engagement in three Chinese universities of different tiers.

Theoretically, the current finding further accentuates the tenets of positive psychology in relation to FL acquisition [8]. Aligned with positive psychology principles, this study shows that the three positive psychology pillars (i.e., enabling institutions, character strengths, and positive subjective experience), operationalized as FLPOM, language-specific grit, and CE in this study, may work in synergy, showing positive influence on the non-linguistic and psychological development of FL learners [8]. The three positive psychology pillars of constructs (e.g., FLPOM, language-specific grit, and positive CE) were positively intertwined, demonstrating a joint and positive benefit on WTC.

In addition, the result offers further evidence for Shao et al.’s [81] theoretical model of L2EPP in that within a favorable classroom environment featuring harmonious teacher-student and peer-peer relationships and well-designed language tasks, pleasant emotions, and affirmative characteristics can individually and jointly promote language learners’ outcome (e.g., WTC in this case).

It is equally important to note that the initial insight regarding the role of FLPOM in triggering WTC responds to MacIntyre and Wang’s [7] introspection. This is in reference to the incomplete heuristic WTC model proposed by [38], which did not consider emotional antecedents except anxiety.

This finding can be explained by empirical evidence from Pavelescu’s [74] case study. That is, a favorable, interactive, and encouraging language classroom environment tends to shape students’ enjoyable experience in FL learning, which subsequently contributes to learning motivation as well as learners’ intention to communicate in integrated and interactive ways from a qualitative perspective.

Concerning the second research question, FLPOM emerged as the positive psychological variable that exerts the strongest effect in predicting WTC among Chinese students in multi-ethnic regions, followed by language-specific grit and positive CE. In other words, when predicting Chinese multi-ethnic students’ intention to speak in FL, the FLPOM experience dominated compared to the gritty trait and favorable classroom climate.

Of note, since published literature has yet linked FLPOM with in-class or face-to-face WTC; further, both as a positive emotion, FLPOM, and FLE were highly correlated [18], we, therefore, replaced FLPOM with FLE in further discussion.

The current finding lent support to previous studies (e.g., [12, 13]) in that the positive emotion construct of FLE showed a stronger effect than domain-general grit in triggering WTC in three EFL contexts of Saudi Arabia, Morocco, and China. It is also slightly against the finding in [57] in that academic grit and FLE had almost the same (β difference = 0.01) and weak effects on Saudi EFL learners. The finding difference might be because of the predictor and sample differences—participants in the [57] study were composed of English majors in different education levels, while the current study focused on non-English majors, capturing less stress and higher positive emotional feelings. In addition, the current result confirms the findings of [75] and [10], where FLE exerted stronger predictive effects than CE in impacting WTC. Yet, our findings, in part, diverge from that of Wang and colleagues [14, 77], who reported FLE strongly, directly triggered WTC, while classroom social climate (CSC) had no direct but only indirect role in provoking WTC in FL Chinese and FL English learning. One possible reason for such path insignificance is that, in their studies, four emotions shared excessive weight originally belonged to the association from CSC to WTC.

One potential explanation for the current research finding is based on the fact that the three WTC predictors exhibit different properties—as a trait, language-specific grit is malleable in the long term but can not easily be changed in the short term; as a learner-external environment, CE is hardly shaped according to learners’ subjective volition; while FLPOM, as an emotion, is more likely than grit and CE to fluctuate significantly, thus showing more possibility to influence learners’ intention to speak than its counterparts.

The finding that FLPOM was more important in predicting WTC than learner personality and learning environment lends support to the study of learner emotions from a cultural-psychological perspective. The culture-oriented positive emotion of FLPOM seems to adequately meet the psychological needs of Chinese learners and increase their readiness to communicate in L2. This finding is particularly insightful because Chinese learners have historically been perceived as lacking initiative in language learning and communication [108]. Using classroom emotional interventions based on FLPOM, which is consistent with individuals’ valued psychological well-being in the Chinese cultural context, may be an effective way to improve Chinese learners’ WTC in L2 learning.

Conclusion, limitations, suggestions, and implications

This study offers pioneering insights by examining the relationship between multi-ethnic EFL students’ FLPOM and in-class WTC, along with language-specific grit and positive CE, as viewed through the lens of three positive psychology pillars: positive experience, character strengths, and enabling institutions. Furthermore, the study utilizes a novel method combining latent dominance analysis with Bayesian testing, a technique recently introduced within the latent variable context in the domain of applied linguistics. This combination serves to confirm the relative importance of three positive psychological predictors (FLPOM, language-specific grit, and positive CE) in predicting WTC in Chinese multi-ethnic students’ EFL classrooms. Specifically, the current findings demonstrate that while all three positive psychology factors impact learners’ WTC outcomes, FLPOM dominates, followed by language-specific grit, and, in turn, by CE. Further, the dominance of one positive emotion (FLPOM) predicting this specific non-linguistic outcome variable (WTC) should not conceal the reality that from the holistic positive psychology angle, the rest of two factors—language-specific grit (individual characteristics) and CE (favorable institutions)—ascertain the complete orbit which will steer to the final non-linguistic FL outcome of WTC.

However, this study has some limitations. Firstly, except for being trait-like and remaining stable, WTC in FL was also argued as being state-like with dynamic tendencies [1]. Yet, only quantitative data and onefold self-rated questionnaires for WTC can not fully address such two properties, thus limiting this study to gain a deeper understanding of WTC’s complex nature. Accordingly, future studies could use other data collection methods, for instance, observation from teachers’ perspective, and think aloud method. Furthermore, the role of peace of mind in WTC development needs to be validated through experimental designs using interventions. Secondly, cross-sectional data might confine this study to offering longitudinal evidence on the association patterns between WTC and its psychological antecedents. It is suggested that future research employ cross-lagged studies or latent growth curve analyses to probe the reciprocal or longitudinal relations between constructs (e.g., [46, 109,110,111]). Thirdly, WTC researched in this study refers to L2/FL speaking WTC, thus ignoring WTC in writing contexts. Since [38] argued the necessity of examining other skill-specific WTC, L2 writing WTC is a promising direction for future study (see details in [112]). Finally, from the standpoint of emotional constructs, only a cultural-specific and LAP emotion of FLPOM was focused in the current study, thus neglecting other kinds of emotional experience (e.g., social emotion, such as pride, shame, embarrassment, jealous, anger, and empathy). Given FL acquisition entails plentiful interpersonal interactions between educators and peers, some of which are infrequent in other school subjects [113], it is recommended that future studies could examine those social emotions in connection with WTC [e.g., 113], interaction engagement, and subsequent learning achievement.

Given the indispensable role of WTC in bumping up interactions and FL output and subsequent FL acquisition, confirmed by the principle of communicative language teaching [114], the findings of the current study are directly relative to FL practitioners. Given the prominence of FLPOM in WTC development, FL practitioners are suggestive of establishing a light-hearted and comfortable classroom atmosphere to help language learners not be afraid of teachers’ and peers’ reactions when they make mistakes. Such a climate is a viable strategy to enhance learners’ self-perceived FL proficiency and subsequent experience of FLPOM and is achievable through higher instructors’ autonomy in learner assessment, in-class activities, and teaching plans [18]. Another potential way to boost learners’ FLPOM experience is mindfulness training [115], such as tai chi, yoga, and mediation. Those practices are beneficial for improving learners’ sense of relaxation and a more well-adjusted and stable internal state [116]. Thereby, mindfulness-based practices may be slotted into students’ curriculum. Moreover, since FLPOM was verified to trump CE and grit in predicting WTC by latent dominance analyses, measures to cultivate FLPOM should be prioritized those interventions to promote grit (e.g., teachers can help students to understand grit’s merit in FL learning through examples of high-grit figures, school fellows, and themselves) and CE (educators can attentively select diversified activities, offer timely and enduring support to students, and promote closely and supportive relationships among students), especially in exam-oriented FL teaching contexts where available teaching and learning time is pressing.

The current findings also inform theoretical implications. Firstly, this research empirically supports that employing a holistic positive psychology perspective integrating the three positive psychology pillars (i.e., positive individual traits, positive subjective experience, and positive collective institutions) [27] is feasible and practical in FL learning, especially among language learners in multi-ethnic regions. Such a holistic view advances the “two-dimensional view of emotions” incorporating positive and negative emotions [80]. In other words, a genuine holistic positive psychology view should integrate those three related but distinct positive psychology pillars [8] instead of merely focusing on the positive experience (emotions) pillar. As explicitly argued by [80], future studies in applied linguistics should apply such a holistic approach to both communication (e.g., WTC) and learning processes (e.g., language engagement, achievement, and motivation). Secondly, the current design did favor us in providing an opening understanding that the experience of peace of mind in FL learning was a dominant predictor of multi-ethnic EFL learners’ WTC at a singular time point of the academic year.