Development and validation of the L2 grit scale in collective cultural context (L2GSC): evidence based on machine learning methods

1 Introduction

Currently, there’s a rising trend among second language acquisition (SLA) researchers recognizing the significance of positive psychology (PP) constructs (Padilla et al. 2020). Grit, as a distinctive instance of PP variables, plays a significant role in PP by contributing to happiness and life satisfaction (Khan and Khan 2017). It is initially conceptualized as a fundamental and enduring trait that manifests in a person’s unwavering commitment and effort toward achieving long-term goals, regardless of any obstacles, challenges, or slow advancements (Duckworth et al. 2007). The authors argue that there are two lower-order constructs: perseverance of effort (PE) and consistency of interest (CI). PE is characterized by an individual’s capacity to uphold dedication and continue to exert effort when faced with challenges. CI, on the other hand, pertains to the sustained focus and enthusiasm for a particular interest or goal over an extended period of time. According to Von Culin et al. (2014), grit is a relatively stable characteristic that persists over various circumstances and timeframes. This implies that individuals with higher levels of grit tend to succeed through a persistent and engaged effort, regardless of potential setbacks or challenges. In addition, research indicates that grit, potentially more than innate talent, contributes to student success by offering added predictive power for achievement beyond natural abilities (Duckworth and Quinn 2009; Duckworth et al. 2007).

Building on this concept, Datu et al. (2017) introduced the Triarchic Model of Grit Scale (TMGS), designing to measure an individual’s propensity to remain determined in pursuing long-term goals within a collectivist/group-oriented context. The reason why developing a TMGS is that individuals in a collectivist/group-oriented or an individualist-oriented context are different. According to Littlewood (1999), in a collectivist context, individuals perceive themselves as integral to their group, prioritizing the group’s perspectives, needs, and objectives over individual distinction. Individualism is characterized by an emphasis on the “I”, while collectivism emphasizes the “We” (Littlewood 1999). Therefore, this model emphasizes not just the importance of PE and CI in chasing long-term goals, but also the importance of adaptability to situations (AS), which means being adaptable and able to adjust one’s actions and strategies to align with varying social, situational, and environmental factors. As explained by Datu and Zhang (2021), in societies where people value a context-sensitive self (Suh 2007), the capacity to adapt is often seen as a precursor to achieving beneficial outcomes.

While general grit has been widely investigated, some researchers have posited that grit should be viewed as a trait that manifests differently across specific fields rather than a uniform, general quality (Duckworth and Quinn 2009; Teimouri et al. 2020). For example, in SLA, grit is posited as essential for sustained learning efforts in second or foreign language (henceforth L2) learners (Sudina et al. 2021). This has led to a rise in investigations into how grit affects language learning, revealing its importance in shaping a learner’s motivation (Khajavy et al. 2021), willingness to communicate (Lan et al. 2021), emotional well-being (Sudina and Plonsky 2021b), and overall language proficiency (Fathi and Hejazi 2024; Sudina and Plonsky 2021b). Nonetheless, Teimouri et al. (2020) evidenced that evaluating personality through measures specific to the language domain yields a more accurate assessment of students’ personality traits and behavioral characteristics in L2 contexts, compared to broad, out-of-context personality assessments. This approach underlines that grit levels can differ significantly across various life contexts (Cormier et al. 2019). To be specific, the domain-specific view of grit focuses on its varying degrees of intensity within distinct achievement areas, leading to a more precise and locally consistent measurement of traits (Cormier et al. 2019). Therefore, the effectiveness of a general grit scale is questionable in specific fields (Cormier et al. 2019; Li and Yang 2023).

Compared with domain-general grit, domain-specific grit (i.e., L2 grit) has increased power in predicting language academic achievement in L2 (Teimouri et al. 2020). L2 grit is defined as the specific application of passion and perseverance to the learning of an L2 (Teimouri et al. 2020). Prior research suggests a correlation between L2 grit and an extensive array of beneficial outcomes in L2 learning, showing that L2 grit is associated with lower foreign language anxiety (Liu et al. 2023a), higher foreign language enjoyment (Fathi and Hejazi 2024), increased growth mindset (Fathi et al. 2024), strengthened self-efficacy beliefs (Fathi et al. 2024), stronger academic engagement (Sadoughi and Hejazi 2023), and better foreign language performance (Fathi et al. 2024). These relationships can be explained through both theoretical and empirical perspectives.

Both self-determination theory (SDT, Deci and Ryan 1985) and broaden-and-build theory (BBT, Fredrickson 2003) focus on individual growth. SDT focuses on how the satisfaction of motivational needs drives personal growth, while BBT focuses on how positive emotions promote adaptation and resilience by expanding cognition and accumulating resources. The combination of the two provides a complete perspective for understanding how individuals can achieve lasting personal development with the support of intrinsic motivation and positive emotions.

SDT provides a robust framework for understanding the relationship between grit, enjoyment and engagement, particularly by explaining how perseverance in long-term goal pursuit fosters positive emotional experiences when autonomy, competence, and relatedness needs are met (Deci and Ryan 1985; Niemiec and Ryan 2009; Ryan and Deci 2000). Individuals with high grit typically pursue their chosen long-term goals voluntarily (autonomy), which enables them to maintain effort despite challenges, build skills, and gain experience (competence), often accompanied by social support from others (relatedness). Through persistent effort, achievement, and recognition, such individuals are likely to be more focused, active, thus more engaged in the process of achieving goals and experiencing enhanced enjoyment throughout the process.

According to the BBT (Fredrickson 2003), positive emotions broaden individuals’ cognitive and behavioral capacities, helping them build long-term resources such as psychological resilience, social connections, and skills, which can be beneficial in coping with challenges in the future. This theory provides a framework for understanding the relations between L2 enjoyment, L2 grit, and L2 engagement. On the one hand, in the context of language learning, L2 enjoyment, as a “positive learning-facilitative achievement emotion,” (Dewaele and Li 2021, p. 926) is closely associated with motivation, engagement, and L2 performance. L2 grit enables individuals to persist through challenges, which increases the likelihood of experiencing enjoyment as they progress toward their goals. As enjoyment increases, individuals may further broaden their experiences and invest more deeply in their pursuits, which strengthens grit over time. On the other hand, L2 grit as a personal resource enables individuals to broaden their experiences and cognitive responses, leading to sustained engagement in meaningful tasks. As people remain engaged, they build further emotional and psychological resources (e.g., enthusiasm and motivation), reinforcing their grit and ability to maintain focus and effort over time. As a result, the relationships among language L2 enjoyment, L2 engagement, and L2 grit are complementary, with active engagement enhancing learners’ grit, thereby fostering long-term success in language learning.

Accordingly, Liu et al. (2023a) highlight that emotional and instrumental support from teachers is crucial in fostering L2 grit, which significantly helps reduce foreign language anxiety. This underscores the need for educators to focus on both instructional content and providing emotional and practical support to improve student outcomes (Liu et al. 2023a). For instance, Zhao (2023) reviewed nine studies on the relationship between grit and foreign language enjoyment, concluding that these two variables are positively correlated. Thus, L2 enjoyment was used as a predicted outcome of L2 grit in this study. Similarly, Liu et al. (2023b) suggest that learners experiencing burnout struggle with engagement unless they build resilience and motivation. Resilience helps individuals stay committed despite setbacks (Eskreis-Winkler et al. 2014), while grit, emphasizing perseverance and overcoming obstacles for long-term goals (Duckworth et al. 2007), correlates positively with engagement and performance among EFL learners (Eren and Rakıcıoğlu-Söylemez 2023). Therefore, empirical studies demonstrated that grit supports enhanced student engagement, and L2 enjoyment was again employed as a predicted outcome of L2 grit in this study.

To date, two prominent L2-related grit scales have been developed, namely L2 grit (Teimouri et al. 2020) and L2 grit scale (Alamer 2021), both rooted in the seminal work of Duckworth et al. (2007). Teimouri et al. (2020)’s 9-item L2 grit scale, situated in the Iranian context, confirmed the existence of two significant components (i.e., PE and CI). In addition, Alamer (2021) devised and validated a 12-item L2 grit scale in the Saudi Arabian context based on the Grit–S scale developed by Duckworth et al. (2007), confirming that grit as a single construct with two specific scales, consistent with the initial conceptualization.

While prior studies (e.g., Liu et al. 2022) have confirmed the reliability and validity of the L2 grit scales developed by Teimouri et al. (2020) and Alamer (2021), the results have varied across different cultural and educational contexts. Specifically, on the one hand, by using the L2 grit scale (Teimouri et al. 2020), Sudina et al. (2021) posited that PE exclusively has a significant correlation with L2 achievement across different languages and contexts. Conversely, Sudina and Plonsky (2021a) highlighted that CI exhibits greater predictive and incremental validity to L2 achievement than PE. Adding complexity to this debate, Cheng (2021) found that both PE and CI facets notably contribute to predicting the willingness to communicate in L2 among English majors in Taiwan, merely PE had a more pronounced impact than CI. Similarly, Wu et al. (2024) recently found that PE rather than CI could predict language learning engagement. On the other hand, in the context of the Middle East, Alamer (2021)’s L2 grit scale also has mixed outcomes. For example, Khajavy et al. (2021) reported no correlation between grit (encompassing both PE and CI) and L2 achievement among Iranian university students. In a subsequent study, Khajavy and Aghaee (2024) indicated that PE alone was a significant indicator of L2 success. Therefore, it becomes evident that the construct of grit in the realm of language learning is multifaceted and its influence on L2 achievement is not universal. The variance in findings underscores the importance of considering cultural and contextual factors when assessing the role of grit in language acquisition. While some components of grit such as PE may stand out as key predictors in certain environments, the overall picture is one of a nuanced trait interacting with diverse educational settings and learner backgrounds. This suggests that educators and researchers should adopt a more localized approach to understanding and applying the concept of grit in L2 learning, recognizing that what may be a significant factor in one context may not hold the same level of influence in another.

Given the nascent stage of promoting grit in PP, more empirical research is needed to refine its definition and measurement in L2 learning. Notably, no research has explored how L2 grit develops and operates from a collectivist/group-oriented perspective, as existing L2-related grit scales were primarily developed and tested from an individualist-oriented perspective (Alamer 2021; Teimouri et al. 2020). Considering that people in East Asian countries exhibit a stronger tendency towards collectivism compared to those in Western countries (Littlewood 1999), and a substantial amount of L2 grit research is carried out in such environments (e.g., China, Japan, and Korea) (Zhao and Wang 2023), it raises the question of whether an L2 grit scale, if developed and validated within a collective society, might yield greater applicability and relevance in such settings. In other words, it requires the necessity to know the potential effectiveness of an L2 grit scale developed and validated within a collective society. To this end, this study aims to critically examine the core issues in L2 grit research. It highlights the oversight of collectivist cultural backgrounds in L2 grit research and the need for more culturally nuanced instruments to measure the personality traits of L2 learners from collective cultural backgrounds.

Machine learning (ML) is increasingly utilized in psychological and linguistic research to enhance the accuracy, efficiency, and predictive power of psychometric and language learning assessments (Lin 2021; Shatte et al. 2019). ML algorithms, such as random forests, can reveal subtle relationships between variables that traditional statistical methods may overlook (Shatte et al. 2019). For instance, a review points out ML has been successfully applied in applied linguistic studies and enjoyed a promising future (Lin 2021). Unlike traditional linear regression with its rigid assumptions of linearity, ML demonstrates significant potential for future studies (Lin 2021). In addition, prediction with ML can lead to a deeper understanding of behavior (Yarkoni and Westfall 2017). In our study, we aim to apply ML in an innovative way to explore L2 grit scale in collective cultural context (L2GSC) to capture L2 learning-related variables that conventional methods might overlook, providing more precise predictions as noted above. Furthermore, many ML algorithms make minimal assumptions about data distribution, for example, allowing for effective handling of non-normally distributed data. Therefore, our approach is informed by both theoretical considerations and empirical evidence, leveraging ML to uncover insights that contribute to the advancement of L2 learning research.

Grounded in the foundational research on the construct of grit (Alamer 2021; Datu et al. 2016, 2017]; Duckworth et al. 2007; Duckworth and Quinn 2009; Teimouri et al. 2020), the study aims to conceptualize, develop, and empirically validate a L2GSC. This endeavor is informed by the TMGS proposed by Datu et al. (2017). Within this framework, L2 grit can be posited as a nuanced trait characterized by the PE, CI, and AS within SLA. In addition to a new model of grit specific to L2 contexts, its relevance and potential impact on the behaviors of L2 learners will also be examined, considering that L2 grit may imply different meanings in different cultural contexts (Datu et al. 2017).

2 Phase 1 item pool of L2GSC

3 Phase 2 initial test of L2GSC

The primary objective of phase 2 was to assess its factor structure and score reliabilities. This involved employing exploratory factor analysis (EFA) and conducting reliability analyses using Cronbach’s alpha.

4 Phase 3 refining the L2GSC

Phase 3 focused on reinforcing the structural validity of the L2GSC via Confirmatory Factor Analysis (CFA). Subsequent evaluations of construct validity were conducted through correlational analysis, while criterion-related validity was assessed using hierarchical regression and ML techniques, with criteria including foreign language engagement and enjoyment.

4.4 Results

4.4.1 CFA results

In the proposed measurement model, L2GSC was conceptualized with three latent factors, each representing one of the three sub-dimensions: perseverance, consistency, and adaptability. These latent factors are indicated by 5, 4, and 4 observed indicators, respectively.

To evaluate the model fit, established criteria for acceptable fit were applied (Schermelleh-Engel et al. 2003): 1) a non-significant χ²; 2) Comparative Fit Index (CFI), Goodness of Fit Index (GFI), and Tucker-Lewis Index (TLI) exceeding 0.90; and 3) Root Mean Square Error of Approximation (RMSEA) below 0.08. The initial CFA revealed suboptimal fit indices: χ² = 220.01, df = 62, p < 0.001, CFI = 0.89, GFI = 0.88, TLI = 0.89, and RMSEA = 0.10 (CI 90 % = 0.089–0.12). A scrutiny of the modification indices (MI) suggested that the inclusion of Item 9 in the consistency of interests was detrimental to the model fit (MI = 35 > 12.0). Following the methodological framework presented by Steger et al. (2006), Item 9 was omitted, leading to another CFA with the revised 12-item version of the L2GSC. This revision yielded improved and acceptable fit indices: χ² = 148.25, df = 51, p < 0.001, CFI = 0.93, GFI = 0.91, TLI = 0.93, and RMSEA = 0.09 (CI 90 % = 0.073–0.107) (Figure 1).

Figure 1:

Outcome of confirmatory factor analysis in phase 3.

Following refinement, the reliability of the enhanced L2GSC is as follows: the overall scale exhibits a Cronbach’s alpha of 0.87, and the alphas for the sub-dimensions are reported as follows: α for perseverance is 0.87, α for consistency is 0.66, and α for adaptability is 0.83.

The correlation analyses demonstrated that PE, along with AS, exhibited positive associations with PE and AS of general grit, growth mindset, and conscientiousness. This reflected that PE and AS have strong correlations with other relevant psychological traits, indicating that these subscales can effectively measure relevant psychological traits in L2 learning. This supported the construct validity of L2 grit as an independent trait. However, consistency was not observed to correlate with adaptability in TMGS, or growth mindset. This may indicate that the subscale of consistency of interest may not be as closely related to relevant psychological variables in L2 grit as the other two subscales. Detailed results of descriptive statistics, reliability estimates, and correlational coefficients for the study variables are presented in Table 2.

Table 2:

Descriptive statistics, reliability coefficients, and correlations of L2GSC with general grit, growth mindset, and conscientiousness.

	Mean	Std. Deviation	α	1	2	3	4	5	6	7	8
1. Perseverance	3.22	0.71	0.87	1
2. Consistency	2.88	0.69	0.66	0.356**	1
3. Adaptability	3.42	0.65	0.83	0.658**	0.189**	1
4. gPerseverance	3.55	0.74	0.79	0.474**	0.300**	0.496**	1
5. gConsistency	2.70	0.76	0.79	−0.090	0.172**	−0.239**	−0.419**	1
6. gAdaptability	3.46	0.67	0.88	0.352**	0.031	0.444**	0.589**	−0.502**	1
7. Growth mindset	4.00	0.99	0.84	0.468**	0.103	0.503**	0.501**	−0.340**	0.485**	1
8. Conscientiousness	3.23	0.37	0.61	0.292**	0.269**	0.324**	0.455**	0.066	0.300**	0.165*	1

1. The gPerseverance, gConsistency and gAdaptability refer to corresponding constructs in TMGS respectively. *p < 0.05, **p < 0.01.

4.4.2 Hierarchical regression results

Before the hierarchical regression analyses, descriptive statistics, reliability indices, and correlations for L2GSC dimensions, general grit (TMGS), L2 engagement, and L2 enjoyment were computed to ensure data quality, confirm the reliability of the measurement instruments, and identify preliminary relationships among the variables (Table 3). The outcomes of hierarchical regression analyses demonstrated that L2 grit had a strong positive effect in both models, that is, it was an important predictor of L2 engagement and enjoyment. This connection persisted even when considering demographic characteristics and trait grit. In addition, consistency and adaptability also show significant positive effects in Model 1. The introduction of L2 grit dimensions into the regression models highlighted additional variance (ranging from 4.7 % to 29 %) in the mentioned outcomes, emphasizing the substantial contribution of L2 grit in predicting L2 learning engagement and enjoyment (refer to Table 4).

Table 3:

Descriptive statistics, reliability coefficients, and correlations of L2GSC with L2 engagement, L2 enjoyment, and general grit.

	Mean	Std. Deviation	α	1	2	3	4	5	6	7	8	9	10
1. Perseverance	3.22	0.71	0.87	1
2. Consistency	2.88	0.69	0.66	0.302**	1
3. Adaptability	3.42	0.65	0.83	0.658**	0.138*	1
4. L2 enjoyment	3.34	0.67	0.91	0.480**	0.220**	0.504**	1
5. Cognitive engagement	3.32	0.81	0.90	0.451**	0.239**	0.507**	0.585**	1
6. Emotional engagement	3.13	0.58	0.61	0.451**	0.370**	0.401**	0.443**	0.603**	1
7. Behavioral engagement	3.28	0.77	0.90	0.473**	0.262**	0.453**	0.561**	0.787**	0.587**	1
8. Social engagement	2.83	0.79	0.81	0.050	0.300**	-0.021	-0.076	-0.154*	0.092	-0.129*	1
9. L2 engagement	3.14	0.50	0.86	0.514**	0.423**	0.487**	0.550**	0.819**	0.796**	0.820**	0.311**	1
10. General grit	3.24	0.36	0.83	0.479**	0.299**	0.446**	0.528**	0.509**	0.422**	0.538**	0.109	0.578**	1

1. *p < 0.05, **p < 0.01.

Table 4:

Hierarchical regression analyses results.

Model		B	t	R2	△R2
1	step1			0.38	0.38***
	Age	-0.12	-0.59
	Sex	0.049	0.80
	Grade	0.52	2.90**
	Major	0.066	2.83**
	General grit	0.76	10.35***
	step2			0.51	0.13***
	Age	-0.247	-1.39
	Sex	0.033	0.60
	Grade	0.402	2.50*
	Major	0.058	2.76**
	General grit	0.455	5.94***
	L2PoE	0.088	1.90*
	L2CoI	0.199	4.93***
	L2AtS	0.154	3.15**
2	step1			0.31	0.29***
	Age	-0.431	-1.552
	Sex	0.006	0.069
	Grade	0.252	0.994
	Major	0.075	2.262*
	General grit	0.960	9.294***
	step2			0.41	0.10***
	Age	-0.46	-1.731
	Sex	-0.005	-0.064
	Grade	0.14	0.609
	Major	0.063	2.045
	General grit	0.61	5.402***
	L2PoE	0.11	1.648
	L2CoI	0.070	1.176
	L2AtS	0.27	3.725*

1. The outcome variable of model 1 and 2 is L2 engagement, and L2 enjoyment, respectively. *p < 0.05, **p < 0.01, ***p < 0.001.

4.4.3 Machine learning results

The analysis initiated a classification task aiming to predict outcomes from a dataset with two distinct targets: L2 enjoyment and L2 engagement. The predictive variables for these objectives are derived from both the three sub-dimensional scores and the overall scores of L2GSC and TMGS. This preprocessing step was crucial for converting the research aim (examine the adaptability of L2GSC to the L2 learning environment) into a multiclass classification scenario. The dataset was split 80–20 into training and testing sets for thorough model evaluation. The models were implemented in Python using libraries for Random Forest, XGBoost, and LightGBM.

In the task of predicting L2 enjoyment (Table 5), employing L2GSC proved to be more advantageous than using TMGS. Specifically, in the performance of L2GSC, the Random Forest model achieved the highest accuracy at 72.92 %, precision at 75.50 %, recall at 72.92 %, and F1-score at 73.30 %. By contrast, the performance of TMGS, as highlighted by the LightGBM model, achieved the highest accuracy at 66.67 %, precision at 63.81 %, recall at 66.67 %, and F1-score at 65.20 %. It proved that the L2GSC evidently overweighed TMGS in predicting L2 enjoyment.

Table 5:

Model performance comparison in predicting L2 enjoyment using data from L2GSC and TMGS.

Predicting L2 Enjoyment: Model Performance Comparison (Weighted Average)
	Performance of L2GSC				Performance of TMGS
	Accuracy	Precision	Recall	F1-score	Accuracy	Precision	Recall	F1-score
RandomForest	72.92 %	75.50 %	72.92 %	73.30 %	60.42 %	58.89 %	60.42 %	59.58 %
XGBoost	68.75 %	73.49 %	68.75 %	70.09 %	60.42 %	60.84 %	60.42 %	60.50 %
LightGBM	68.75 %	74.14 %	68.75 %	69.40 %	66.67 %	63.81 %	66.67 %	65.20 %

In terms of the analysis of predicting L2 engagement (Table 6), in the performance of L2GSC, the Random Forest model achieved accuracy at 70.83 %, precision at 75.20 %, recall at 70.83 %, and F1-score at 72.84 %. The TMGS performed best with the LightGBM model. It achieves an accuracy of 77.08 %, precision at 75.25 %, recall at 77.08 %, and F1-score at 76.16 %. This suggests that TMGS might be slightly more predictive of L2 engagement.

Table 6:

Model performance comparison in predicting L2 engagement using data from L2GSC and TMGS.

Predicting L2 Engagement: Model Performance Comparison (Weighted Average)
	Performance of L2GSC				Performance of TMGS
	Accuracy	Precision	Recall	F1-score	Accuracy	Precision	Recall	F1-score
RandomForest	70.83 %	75.20 %	70.83 %	72.84 %	77.08 %	73.51 %	77.08 %	75.25 %
XGBoost	72.92 %	75.49 %	72.92 %	74.08 %	77.08 %	73.51 %	77.08 %	75.25 %
LightGBM	70.83 %	67.46 %	70.83 %	69.11 %	77.08 %	75.25 %	77.08 %	76.16 %

5 Discussion

The aim of this research was to present initial evidence supporting the reliability and validity underlying a newly developed L2 grit scale for measuring L2 grit within a collectivist culture. Overall, the findings indicate that the L2GSC scores are psychometrically robust, offering a credible measure of L2 grit within the context of a collectivist culture (i.e., China in this study).

By integrating EFA data from Phase 2 with CFA data in Phase 3, the research has affirmed the second-order three-factor model of grit, which further replicated the findings of Datu et al. (2017), who found similar factor structures in samples from collective societies. The 12-item L2GSC, encompassing three distinct factors (i.e., perseverance of effort, consistency of interest, and adaptability to situations), was validated through evidence of strong psychometric properties, including internal and external validity and reliability. Furthermore, the scale’s adequacy is highlighted by its composition of an appropriate number of items across three factors. In addition, this research not only distinguishes L2GSC from the general Triarchic Model of Grit by affirming and extending the conceptual coherence of Datu et al. (2017) but also sets it apart from other constructs such as L2-related grit scales described by Teimouri et al. (2020) and Alamer (2021), thereby contributing to the theoretical and empirical understanding of grit in SLA contexts from a collectivist/group-oriented perspective.

This work not only reaffirms the dual facets of PE and CI within the framework of L2 grit as outlined by Teimouri et al. (2020) but also corresponds with the Triarchic Model of Grit proposed by Datu et al. (2017). Aligning with the new dimension (i.e., adaptability to situations) proposed by Datu et al. (2017), an intriguing question arises regarding the role of adaptability within L2 learning. Pettigrew (1979) contends that culture is crafted by human action and, reciprocally, shapes human identity; that is, self-perception is significantly influenced by the cultural environment. Therefore, within the context of collectivist cultures in which individuals’ learning habits are often underpinned by Confucian values, the construct of AS potentially serves as a critical component in the acquisition and application of an L2 for the following two reasons.

Specifically, the concept of Confucian dynamism refers to the extent to which a society emphasizes a future or long-term orientation or a past or short-term orientation and reflects the Confucian ethic of diligence (Hofstede and Bond 1988). Individuals from East Asian cultures are particularly noted for their strong success orientation, often placing a higher premium on hard work than typically observed in Western cultures (Cheng 1990). This study posits that effective L2 learning outcomes are a reflection of educational success. Such outcomes are frequently linked with attributes of perseverance and the ability to overcome challenges in the L2 learning journey, traits that are encapsulated by the term “grit”. Consequently, the high value that culture places on achievement might inspire students to persist in their English language studies, despite encountering unforeseen obstacles. Second, Confucian culture emphasizes people’s ability to cope with changes and challenges though individuals should respect tradition and maintain social order (Niu 2012). It posits that individuals ought to exhibit flexibility and open-mindedness, proactively engaging with and molding their surroundings, whilst also permitting environmental influences to inform their development (Niu 2012). This idea of adaptability and flexibility can be considered a response of Confucian culture to the constant change and unpredictability of real life. Therefore, in the L2 learning process, learners are encouraged to remain adaptable and responsive to each emerging opportunity.

In addition, this study examined the interplay between language learners’ language growth mindsets, conscientiousness, engagement, enjoyment, and their association with L2 grit. The outcomes of the correlational analysis revealed that both PE and AS demonstrated positive correlations with language growth mindset while consistency is not. From the perspective of collectivism, people emphasize the importance of interpersonal relationships within groups and the extended family (Boucher 2010), which can be regarded as an “other-directed self” (Wen and Clément 2003, p. 19). The robust network of communal relations offers consistent support to individuals, who in turn are expected to demonstrate unwavering loyalty to their group (Tang 2012). In such contexts, personal identity is less about individual uniqueness and more about external, socially defined attributes like roles, status, and affiliations with family and other collectives (Boucher 2010). Therefore, L2 learners may exhibit a diligent pursuit of successful language acquisition outcomes, striving to overcome any challenges to meet the expectations of others, such as family members and teachers, which can be labeled as subjective norms (Ajzen and Fishbein 1975).

Furthermore, both PE and AS showed significant positive associations with conscientiousness, whereas CI had a minor positive association with the same trait. Teimouri et al. (2020) reported no correlation between conscientiousness and L2 grit, diverging from past research indicating a strong link between conscientiousness and general grit. It’s important to consider future research to explore if L2 grit within a collective context is connected to conscientiousness in the same manner as general grit or conscientiousness has the potential to predict L2 grit within a collective culture, examining this in different environments.

The regression results showed that students with higher L2 grit were more attentively engaged in classroom activities and exhibited a more pronounced predilection for enjoying the language learning journey compared to their counterparts with lower grit. These outcomes align with the corpus of grit literature and L2 learning, which implies satisfactory L2 learning outcomes (Fathi and Hejazi 2024; Sadoughi and Hejazi 2023). This suggests that L2GSC is a reliable indicator of L2 learners’ behavioral patterns, confirming its appropriateness for the L2 context.

Based on ML analysis, the results demonstrate that the L2GSC outperforms the TMGS in predicting L2 enjoyment. The superior predictive power of L2GSC suggests that it is a dependable measure for anticipating the behavioral trends of L2 learners within language learning contexts. An interesting finding was that TMGS outperformed L2GSC in predicting L2 engagement. While this aspect of the findings may seem ambiguous, three potential reasons could explain this. First, the participants in this study were primarily from China, where English is a mandatory subject in universities. This requirement aims to enhance their language proficiency, which in turn improves their employability post-graduation. As a result, their engagement in learning English may play an important role in fostering their future career success, which motivates them to dedicate time and effort to their studies. In this context, students likely study diligently not out of interest in specific subjects, but rather with the general goal of achieving academic success across all courses. This suggests that their grit may be more generalized rather than specific to any one domain. However, future research should examine whether this finding holds true across different subjects and areas of study. In addition, we speculate that L2 enjoyment might be a more domain-specific variable than L2 engagement. This is supported by an analysis revealing that enjoyment is a more powerful predictor of achievement in L2 vocabulary than engagement (Oyama 2022), affirming the significant role of enjoyment in the learning process, making it a more suitable predictor of L2 grit. The varying degrees of domain specificity among different variables within the same domain is an intriguing question that deserves future research. This also demonstrates the robust predictive power of ML methods, which can help SLA researchers and educators better understand and recognize the underlying relationships between different language learning conditions. Third, this result was achieved through the innovative application of ML in scale validation, marking a significant methodological advancement. Comparing with using hierarchical regression in predicting engagement, ML provided distinct findings that deepen our understanding of the data. This approach offers a novel contribution, as prior research has typically relied on regression or correlation analyses for scale validation (e.g., Datu et al. 2017; Datu and Zhang 2021), rarely extending to the integration of ML techniques. By conducting a more comprehensive analysis with ML, we were able to reveal a new perspective that highlights the added value and necessity of employing diverse analytical methods to examine the same research question. This not only enhances the robustness of our findings but also underscores the potential of ML to refine and extend traditional validation processes.

In brief, the research has elucidated several key findings regarding the characteristics and outcomes associated with L2 grit. First, students exhibiting grit can demonstrate sustained persistence in their L2 learning endeavors, maintain a passion for engaging with the target language, and show an increased awareness of and responsiveness to evolving opportunities and changes within their L2 learning context. Second, gritty students can experience a heightened sense of engagement and derive more enjoyment from their L2 learning activities and may harbor the conviction that their language learning achievements can be significantly enhanced through increased effort. Therefore, it is clear that grit is not merely a peripheral trait but is central to fostering persistence, enthusiasm, and an adaptive response to the learning environment among students.

6 Limitations and future directions

This study encounters certain limitations. First, the self-reporting method may introduce biases, such as social desirability, and memory recall potentially affecting data accuracy. Second, use of convenience sampling, while practical, may limit the representativeness of the dataset. Random sampling in future studies could address this issue. Also, to establish a more robust cultural foundation, incorporating students from Chinese Departments may yield stronger evidence of Chinese collectivism. Third, the study employed a cross-sectional design and did not include language achievement measures, such as English test scores. Future research could adopt longitudinal or mixed-method approaches to better assess the L2GSC’s predictive capacity for language learning outcomes. Finally, this study represents a pioneering effort to create models using ML methods to evaluate scalability and predictability. The psychometric validation of the L2GSC was conducted within a single collectivist cultural context (China). Future research should gather more data to build more models and enhance predictive performance and test the L2GSC across diverse cultural settings to evaluate its cross-cultural applicability and replicability of findings.

7 Conclusions

The present study developed and validated L2GSC, a tool designed to assess grit in the context of collectivist culture specifically within SLA. The findings indicate that students exhibit higher grit levels in language learning scenarios than in general contexts when assessed with this scale. This suggests that a domain-specific measure of grit yields more definitive and significant outcomes. Building on prior studies, this work integrates the concept of L2 grit into the collectivist societal framework and investigates its impact on the learning behaviors of L2 learners. In short, the results suggested that grittier individuals tend to be more engaged and enjoyable in language learning compared to their less gritty peers.

This research contributes to theoretical understanding, practical application, and pedagogical implications. Theoretically, it enriches the construct of L2 grit with cultural nuances, potentially enhancing SLA processes. Practically, the L2GSC offers a valuable instrument for researchers and educators to measure individual variances in PE, CI, and AS toward long-term objectives in SLA. Should further replication studies consistently validate the factorial structure of the L2 grit scale across various collective contexts, this scale will become a valuable contribution to the current measures for grit specific to the language domain. Pedagogically, the belief in the power of perseverance of effort, consistency of interest, and adaptive approaches, a hallmark of gritty students, suggests a promising avenue for enhancing language learning outcomes. Therefore, these insights advocate for the incorporation of strategies to cultivate grit within language learning curricula, aiming to unlock the full potential of learners in mastering an L2.