Growth mindset intervention and academic engagement: the mediating role of l2 motivational self system among non-English majors Authors: Lifang Ye Published: 2026 ABSTRACT BackgroundNon-English major students in Chinese higher education frequently exhibit insufficient motivation and declining academic engagement in compulsory English courses. Although growth mindset theory and the L2 Motivational Self System (L2MSS) have each been associated with language learning outcomes, intervention research integrating these traditions remains scarce, and the psychological pathways linking implicit ability beliefs to engagement behaviors are not well understood. Drawing on possible selves theory and self-discrepancy theory as a unifying framework, this study examined the longitudinal effects of a growth mindset intervention on non-English major students’ L2MSS and academic engagement, and tested whether the L2MSS mediates the intervention–engagement relationship.MethodsA quasi-experimental design with cluster randomization at the class level assigned 120 first-year undergraduates to an experimental group (n = 60) and a control group (n = 60). The experimental group received a 12-week growth mindset intervention integrated into regular English instruction, comprising four modules: cognitive restructuring, attribution training, growth-oriented feedback, and metacognitive strategy instruction. Data were collected at three time points across one semester (T1 baseline, T2 mid-intervention, T3 post-intervention), measuring growth mindset, L2MSS components (ideal L2 self, ought-to L2 self, and L2 learning experience), and academic engagement dimensions (behavioral, emotional, and cognitive). Repeated measures ANOVA examined intervention effects, and bias-corrected bootstrap mediation analyses with a temporally ordered model (T1 → T2 → T3) tested mediating pathways.ResultsRepeated measures ANOVA revealed significant Group × Time interactions for all three L2MSS components and all three engagement dimensions, indicating that the intervention was associated with gains in motivational selves and engagement. The largest effects were observed for ideal L2 self and L2 learning experience. Bootstrap mediation analyses suggested that the L2MSS, particularly the ideal L2 self, may have functioned as a mediating mechanism, accounting for approximately 78.5% of the total intervention effect on engagement (total indirect effect β = 0.665, 95% CI [0.521, 0.821]), with the ideal L2 self alone contributing about 51%.ConclusionsFindings indicate that a classroom-integrated growth mindset intervention can meaningfully reshape non-English major students’ L2 motivational selves and academic engagement, with effects transmitted primarily through enhanced ideal L2 self construction. These results extend theoretical understanding of how implicit beliefs, domain-specific motivational constructs, and learning behaviors connect, and they offer practical guidance for educators seeking to address motivation and engagement challenges in foreign language classrooms. Causal inferences should be tempered by the quasi-experimental design and absence of delayed follow-up assessment. Introduction The landscape of foreign language education in Chinese higher education has undergone notable transformation over the past decade, yet these reforms have also exposed challenges that call for deeper scholarly inquiry. Non-English major students, who form the majority of English learners in Chinese universities, continue to grapple with insufficient motivation and declining academic engagement—a reality that hampers the attainment of desired learning outcomes [1]. Institutional assessments indicate that many students approach English courses with passive attitudes, treating them as compulsory hurdles rather than opportunities for meaningful skill development [2]. This disconnect between educational objectives and learner dispositions raises pressing questions about how motivational deficits in second language acquisition might be more effectively addressed. Two parallel lines of inquiry have shaped contemporary thinking about language learning motivation. Dörnyei’s L2 Motivational Self System (hereafter L2MSS) reconceptualizes motivation through the lens of possible selves [3], offering a tripartite structure of ideal L2 self, ought-to L2 self, and L2 learning experience [4]; Yousefi and Mahmoodi’s [5] meta-analysis affirms its predictive value while flagging the scarcity of intervention-based work. Running alongside this tradition, Dweck’s implicit theories framework distinguishes growth from fixed mindsets [6], and growth mindset has been linked to persistence, resilience, and various academic outcomes [7]. Within language education specifically, Lou and Noels [8] and Lou and Zarrinabadi [9] extended these ideas into the construct of language mindsets, while Kırmızı, Irgatoğlu, and Atalmış [10] examined how mindsets, grit, and the L2MSS structurally cohere. Yet despite an established line of work documenting that mindsets matter for self-regulation and achievement [11, 12], the literature remains thin on intervention studies that trace how mindset shifts might cascade into domain-specific motivational systems, and the recent meta-analytic exchange between Macnamara and Burgoyne [13] and Burnette et al. [14] indicates that growth mindset effects vary considerably across implementation contexts—a heterogeneity that, as Tipton, Yeager, and colleagues [15] caution, demands attention to when and for whom such interventions prove beneficial. Such variability is precisely what motivates the present study’s emphasis on the psychological pathways through which intervention effects unfold rather than the mere question of whether they exist. Academic engagement—a multidimensional construct spanning behavioral, emotional, and cognitive components—represents a third critical variable, with active participation, emotional investment, and deep cognitive strategies all predicting academic success and persistence [16, 17]. Despite accumulating evidence for each of these constructs independently, however, research examining their interrelationships remains fragmented; studies have largely treated growth mindset, the L2MSS, and engagement as separate phenomena rather than examining their dynamic interplay, a gap that Zhou, Hiver, and Al-Hoorie [18] recently flagged in calling for more process-oriented longitudinal research on engagement in L2 classrooms. Connecting these strands requires a unifying theoretical lens rather than a parallel listing of constructs. The present investigation adopts possible selves theory and self-discrepancy theory as the central framework (developed in Sect. 2), under which mindset operates as a precondition shaping self-construction, the L2MSS captures the motivational architecture activated by that construction, and engagement emerges as the behavioral expression of self-guided action. This framing supplies a principled basis for examining the L2MSS as a mediating mechanism between mindset intervention and engagement outcomes. Against this backdrop, the present study investigates the longitudinal effects of a growth mindset intervention on non-English major students’ L2MSS and academic engagement, treating these constructs as dynamically interrelated rather than independent. Three objectives guide the inquiry: to determine whether systematic mindset intervention produces meaningful changes in students’ L2 motivational selves; to assess whether such intervention enhances behavioral, emotional, and cognitive engagement in English learning; and to elucidate the mediating role of the L2MSS in linking mindset to engagement. A quasi-experimental design with measurement at three time points across one semester serves these aims, allowing intervention effects to be traced as they unfold rather than captured at a single moment. The study contributes to the field on several fronts. Theoretically, it offers an integrated account—anchored in possible selves and self-discrepancy theory—of how implicit beliefs, motivational self-constructs, and engagement behaviors connect. Methodologically, the longitudinal structure transcends the limitations of prevalent cross-sectional approaches and reveals temporal dynamics often invisible to single-wave studies. Practically, an effective intervention would furnish educators with an implementable strategy that can be folded into existing pedagogical frameworks without major curriculum overhaul, and the analysis of mediating pathways would allow more precise targeting of intervention components. Theoretical foundation and research hypotheses Growth mindset theory and its educational intervention Dweck’s implicit theories of intelligence framework distinguishes between two fundamentally different beliefs about the nature of human abilities [19]. Individuals with a growth mindset believe that intelligence and capabilities are malleable qualities that can be developed through dedication, strategic effort, and learning from mistakes. In contrast, those holding a fixed mindset view these attributes as static traits—essentially predetermined and resistant to change regardless of effort invested [20]. This distinction goes beyond mere optimism or self-confidence; it reflects deeper epistemological assumptions about the nature of human potential that shape how individuals interpret challenges, setbacks, and feedback. The developmental trajectory of growth mindset theory has expanded considerably since its initial formulation. Early research primarily documented associations between mindset orientations and achievement patterns, demonstrating that growth mindset predicted greater persistence when facing difficulties and stronger academic performance across domains [21]. Subsequent investigations have delved into the cognitive and affective mechanisms underlying these effects, revealing that mindset influences how students attribute success and failure, the types of goals they pursue, and their emotional responses to academic challenges [22]. More recent work has examined contextual factors that shape mindset development, including cultural influences, classroom climate, and instructional practices [23]. Educational interventions targeting growth mindset have emerged as promising approaches for enhancing student motivation and achievement. These interventions typically employ several core strategies: exposing students to scientific evidence about brain plasticity and the malleability of intelligence; teaching them to reframe challenges as opportunities for growth rather than threats to self-worth; and guiding them to attribute academic outcomes to controllable factors such as effort and strategy rather than fixed ability [24]. Meta-analytic evidence suggests that such interventions can produce meaningful improvements in academic outcomes, particularly for students facing motivational difficulties or academic risk [25]. However, effect sizes vary considerably depending on implementation quality, duration, and contextual factors [26]. The applicability of growth mindset intervention to foreign language education warrants careful consideration. Language learning presents challenges that may make it particularly amenable to mindset intervention—it inherently involves sustained effort, tolerance for errors, and a willingness to engage with unfamiliar material [27]. Some students approach language learning with fixed beliefs about language aptitude, viewing themselves as either “good” or “poor” language learners based on perceived innate talent [28]. Preliminary empirical work has yielded encouraging results, with studies documenting positive associations between growth mindset and language learning outcomes [8]. Importantly, Lou and Noels [8] proposed the Language Mindset Meaning System, a motivational framework that systematically maps how language mindsets relate to goal orientations, effort beliefs, and emotional responses. A meta-analysis by Elahi Shirvan, Robat, Alamer, Lou, and Barabadi [29] found a moderate positive correlation between growth language mindset and various language learning outcomes. Yet questions persist about how mindset intervention interacts with domain-specific motivational systems in language learning [30]. The recent debate between competing meta-analyses—Macnamara and Burgoyne [13] reporting negligible effects after accounting for study quality, versus Burnette et al. [14] finding meaningful effects for targeted subpopulations with high implementation fidelity—underscores that intervention context and design choices matter considerably for observed outcomes. L2 motivational self system theory Dörnyei’s L2 Motivational Self System represents a significant reconceptualization of second language motivation, grounding motivational processes in self-related constructs derived from possible selves theory [31]. This framework moves beyond traditional integrative and instrumental motivation dichotomies, instead proposing that learners’ visions of themselves as future L2 users constitute the primary driving force behind sustained engagement with language learning. The model’s elegance lies in its synthesis of psychological theories of the self with applied linguistic concerns, creating a framework that resonates across diverse learning contexts [32]. The tripartite structure of the L2 Motivational Self System comprises three interrelated components, each contributing uniquely to motivational dynamics. The ideal L2 self encompasses learners’ aspirational images of themselves as competent L2 users—representations of who they would like to become [33]. When this ideal self is vivid, elaborate, and perceived as attainable, it generates powerful approach motivation that propels learners toward language study. The ought-to L2 self, by contrast, reflects attributes that individuals believe they should possess to meet external expectations or avoid negative outcomes [34]. This dimension captures more obligation-based motivation, often associated with parental expectations, academic requirements, or professional necessities. The third component, L2 learning experience, concerns situated, context-specific motives related to the immediate learning environment, including relationships with teachers and peers, curriculum design, and success experiences [35]. These three components do not operate in isolation but interact in complex ways that shape overall motivational intensity and quality. Research suggests that the ideal L2 self typically exerts stronger influence on motivated behavior than the ought-to self, particularly in contexts where learners possess autonomy in their educational choices [36]. However, cultural factors moderate these relationships, with some collectivist contexts showing stronger effects for ought-to selves reflecting family and societal obligations [37]. The learning experience dimension often serves as a foundational element, influencing whether learners develop clear future self-guides in the first place. Cross-cultural validation studies have substantiated the L2MSS’s applicability across diverse settings. Investigations conducted in East Asian, European, Middle Eastern, and other contexts have generally confirmed the theorized three-factor structure, with some variation in component strength and interrelationships [38]. Yousefi and Mahmoodi [5] conducted a meta-analysis confirming that the three L2MSS components predict motivated behavior, though ideal L2 self consistently showed the strongest predictive power. More recently, Liu [39] mapped the L2MSS literature through bibliometric analysis, identifying emerging trends including increased focus on multilingual future self-guides and growing connections to Complex Dynamic Systems Theory. Despite its enduring influence, the L2MSS has also attracted recent critical scrutiny. Al-Hoorie and colleagues [40] questioned whether L2 motivation can be adequately modeled as a self-system, prompting important debates about the framework’s conceptual boundaries. These developments underscore the need for empirical studies that test the L2MSS not merely as a descriptive framework but as a dynamic mediating mechanism in intervention research. The connections between growth mindset and the L2MSS warrant deeper attention. Possible selves theory suggests that growth-oriented beliefs influence how learners construct their ideal L2 selves: those who view abilities as malleable can envision more ambitious future selves and treat these visions as attainable through effort [41]. Self-discrepancy theory complements this account by clarifying when the gap between actual and ideal L2 selves functions as a motivating force rather than as a source of helplessness—growth orientation transforms the gap from threat into challenge. Growth mindset may also reshape ought-to L2 selves by altering how learners interpret external expectations, recasting them as developmental targets rather than fixed judgments about capability. Empirical work has begun to support these theoretical links: Kırmızı et al. [10] found that growth mindset positively predicted both ideal L2 self and L2 learning experience among EFL learners, and Zarrinabadi, Lou, and Ahmadi [42] reported that growth language mindset was associated with stronger ideal L2 selves and more positive affective responses to language learning. These findings provide an empirical foundation for the present study’s hypothesized mediational pathway. Academic engagement and research hypotheses Academic engagement reflects the psychological investment and active participation students demonstrate in learning activities, representing a critical mediator between motivational dispositions and actual achievement [43]. Scholars conceptualize this construct as multidimensional, typically distinguishing among behavioral, cognitive, and emotional facets that collectively characterize the quality and intensity of student involvement in educational processes. Behavioral engagement manifests through observable actions—class attendance, homework completion, participation in discussions, and persistence when encountering difficulties [44]. Cognitive engagement encompasses the mental effort students invest, including their deployment of deep learning strategies, self-regulation, and metacognitive monitoring [45]. Emotional engagement captures students’ affective responses to learning experiences, ranging from enthusiasm and interest to anxiety and frustration [46]. The relationship between motivation and engagement has attracted considerable scholarly interest, with motivation generally positioned as an antecedent that energizes and directs engagement behaviors. Students with strong intrinsic motivation or well-developed achievement goals tend to exhibit higher engagement across all three dimensions [47], although this relationship is not strictly linear: engagement experiences can recursively shape motivation, creating feedback loops in which successful engagement strengthens motivational intensity while disengagement erodes it. Zhou, Hiver, and Al-Hoorie [18] showed that teacher motivational practice and student engagement operate in a reciprocal fashion across time, challenging unidirectional models. In second language contexts, these dynamics become particularly salient given the sustained effort proficiency demands [48], and Dao [49] has further extended engagement conceptualization into online L2 settings. Within the present framework, the L2MSS supplies the domain-specific account of how motivation translates into engagement: the ideal L2 self generates approach-driven behavioral participation and emotional resonance, the ought-to L2 self contributes obligation-based persistence, and positive L2 learning experiences cultivate the sense of competence that feeds cognitive engagement [16, 43]. To anchor the proposed mediational pathway in a coherent theoretical structure, the present investigation places possible selves theory [31] at the centre of its framework, with self-discrepancy theory [50] supplying the motivation-to-action mechanism and Dweck’s implicit theories [6] entering as the upstream precondition rather than as a parallel construct. Within this unified lens, learners’ visions of future language-using selves serve as the primary motivational engine; the gap between actual and ideal (or ought-to) selves generates motivational energy that translates into behavioral, emotional, and cognitive engagement; and beliefs about whether ability is malleable determine whether such gaps register as energizing challenges or demoralizing verdicts. A growth-mindset orientation thus enables vivid ideal L2 self construction by lifting the psychological barrier that fixed-ability beliefs impose, and the elaborated ideal L2 self in turn produces the engagement behaviors that bring the actual self closer to the imagined one [51, 52]. The intervention is therefore expected to reshape engagement primarily through its impact on the L2MSS rather than independently of it, with the ideal L2 self carrying the largest share of the indirect pathway. Drawing upon this theoretical rationale and existing empirical evidence, three core hypotheses guide the present investigation:H1: Growth mindset intervention will exert positive effects on non-English major students' L2 Motivational Self System, including significant enhancements in ideal L2 self, ought-to L2 self, and L2 learning experience.H2: Growth mindset intervention will positively influence students' academic engagement in English learning, manifested through increased behavioral, cognitive, and emotional engagement.H3: The L2 Motivational Self System will mediate the relationship between growth mindset intervention and academic engagement, such that intervention effects on engagement operate partially or fully through changes in students' L2 motivational selves. H1: Growth mindset intervention will exert positive effects on non-English major students' L2 Motivational Self System, including significant enhancements in ideal L2 self, ought-to L2 self, and L2 learning experience. H1: Growth mindset intervention will exert positive effects on non-English major students' L2 Motivational Self System, including significant enhancements in ideal L2 self, ought-to L2 self, and L2 learning experience. H2: Growth mindset intervention will positively influence students' academic engagement in English learning, manifested through increased behavioral, cognitive, and emotional engagement. H2: Growth mindset intervention will positively influence students' academic engagement in English learning, manifested through increased behavioral, cognitive, and emotional engagement. H3: The L2 Motivational Self System will mediate the relationship between growth mindset intervention and academic engagement, such that intervention effects on engagement operate partially or fully through changes in students' L2 motivational selves. H3: The L2 Motivational Self System will mediate the relationship between growth mindset intervention and academic engagement, such that intervention effects on engagement operate partially or fully through changes in students' L2 motivational selves. Research design and methods Participants and intervention program design Participants were recruited from four intact classes of first-year non-English major undergraduates at a financial university in eastern China during the fall semester of 2022. Because random assignment at the individual level was not feasible within intact classroom settings, this study employed cluster randomization at the class level: two classes were randomly assigned to the experimental condition and two to the control condition. Selection criteria stipulated that participants must be full-time undergraduate students enrolled in compulsory English courses, have completed at least one semester of university English instruction, and have no prior exposure to formal growth mindset training programs [53]. Students with diagnosed learning disabilities or those planning to withdraw from the course were excluded. The final sample comprised 120 participants, with 60 students in the experimental group and 60 in the control group. Table 1 presents the demographic characteristics of both groups. As shown in Table 1, the two groups showed comparable distributions across key demographic variables, with no statistically significant differences in gender composition, age, or baseline English proficiency. This demographic equivalence strengthened the internal validity of the quasi-experimental design, minimizing potential confounding attributable to pre-existing group differences [54]. It should be noted, however, that cluster randomization at the class level introduces the possibility of clustering effects, which represents a limitation discussed later in this paper. Table 1Basic Characteristics of Experimental and Control GroupsCharacteristicCategoryExperimental Group (n = 60)Control Group (n = 60)GenderMale24 (40.0%)26 (43.3%)Female36 (60.0%)34 (56.7%)AgeMean (SD)18.7 (0.6)18.6 (0.7)Major CategoryScience & Engineering35 (58.3%)33 (55.0%)Liberal Arts25 (41.7%)27 (45.0%)Baseline English ProficiencyCET-4 Score Mean (SD)456.3 (42.1)452.8 (45.6)Prior English LearningYears Mean (SD)9.2 (1.1)9.4 (1.2)Weekly English Study HoursMean (SD)6.8 (2.3)6.5 (2.5)A note on the intervention’s framing: although recent scholarship distinguishes domain-specific language mindsets from general intelligence mindsets [8, 9], the present program adopted a general growth mindset framework adapted for language learning. This choice was guided by two practical considerations: non-English major students’ motivational difficulties tend to extend beyond language-specific beliefs into broader academic self-concepts, and the existing intervention literature offers more developed implementation guidelines for general mindset programs [14]. Materials were nonetheless tailored to language learning, incorporating language-specific examples, attribution scenarios, and feedback practices. Table 1Basic Characteristics of Experimental and Control GroupsCharacteristicCategoryExperimental Group (n = 60)Control Group (n = 60)GenderMale24 (40.0%)26 (43.3%)Female36 (60.0%)34 (56.7%)AgeMean (SD)18.7 (0.6)18.6 (0.7)Major CategoryScience & Engineering35 (58.3%)33 (55.0%)Liberal Arts25 (41.7%)27 (45.0%)Baseline English ProficiencyCET-4 Score Mean (SD)456.3 (42.1)452.8 (45.6)Prior English LearningYears Mean (SD)9.2 (1.1)9.4 (1.2)Weekly English Study HoursMean (SD)6.8 (2.3)6.5 (2.5)A note on the intervention’s framing: although recent scholarship distinguishes domain-specific language mindsets from general intelligence mindsets [8, 9], the present program adopted a general growth mindset framework adapted for language learning. This choice was guided by two practical considerations: non-English major students’ motivational difficulties tend to extend beyond language-specific beliefs into broader academic self-concepts, and the existing intervention literature offers more developed implementation guidelines for general mindset programs [14]. Materials were nonetheless tailored to language learning, incorporating language-specific examples, attribution scenarios, and feedback practices. Basic Characteristics of Experimental and Control Groups A note on the intervention’s framing: although recent scholarship distinguishes domain-specific language mindsets from general intelligence mindsets [8, 9], the present program adopted a general growth mindset framework adapted for language learning. This choice was guided by two practical considerations: non-English major students’ motivational difficulties tend to extend beyond language-specific beliefs into broader academic self-concepts, and the existing intervention literature offers more developed implementation guidelines for general mindset programs [14]. Materials were nonetheless tailored to language learning, incorporating language-specific examples, attribution scenarios, and feedback practices. The intervention program ran across 12 weeks and integrated four modules designed to cultivate growth mindset orientations within second language learning. The first module, mindset cognitive restructuring (Weeks 1–3), used multimedia presentations, guided discussions, and case-based reflection to expose students to evidence on brain plasticity and the malleability of language learning ability [55]. Students engaged with cases of language learners who overcame initial difficulties through strategic effort, then wrote brief reflections linking the neuroscience content to their own English learning histories—an exercise designed to challenge fixed-ability narratives carried over from prior schooling. The second module, attribution training (Weeks 4–6), taught students to reframe setbacks as natural components of the learning process rather than as evidence of inherent inadequacy [56]. Through structured reflection exercises—post-task attribution worksheets, draft comparison activities, and short self-assessments—students practised attributing difficulties to controllable factors such as strategy choice and practice intensity. The third module (Weeks 7–9) emphasized growth-oriented feedback, with instructors providing process-focused comments that highlighted effort, strategy use, and improvement trajectories rather than ability judgments or normative comparisons. The fourth module (Weeks 10–12) introduced metacognitive strategy instruction, equipping students with techniques for self-monitoring, goal setting, and adaptive strategy selection [57]; a weekly learning log supported these practices, and biweekly future-self visualization exercises asked students to articulate concrete scenarios in which they used English successfully, thereby strengthening ideal L2 self vividness. Detailed protocols and sample materials are available in the supplementary materials. Figure 1 illustrates the comprehensive structure and temporal sequencing of the intervention program. As shown in Fig. 1, the intervention followed a recursive design wherein foundational concepts introduced in early weeks were continuously reinforced and deepened through subsequent activities, ensuring sustained engagement with growth mindset principles rather than treating them as one-time lessons. Fig. 1Growth Mindset Intervention Program Implementation Framework Growth Mindset Intervention Program Implementation Framework The experimental group received the intervention within their regular English course, with 20-minute segments per twice-weekly class session devoted to intervention content. Two trained English instructors—one per experimental class—delivered the program after participating in a workshop on growth mindset theory and standardized activity protocols. Implementation fidelity was monitored through weekly observations and instructor session logs, with checks indicating that over 90% of planned activities were delivered as designed. To maintain instructional equivalence, control group students engaged in supplementary English language practice activities (e.g., vocabulary review, reading exercises) during the same 20-minute time slot, ensuring equivalent total instructional time across conditions. Control group instructors were informed that the study examined instructional approaches but were not given details about the growth mindset content to minimize contamination. Both groups received identical curricular content, assessment standards, and total class hours. This design aimed to isolate intervention effects while preserving ecological validity, though it should be acknowledged that perfect instructional equivalence is difficult to guarantee and the 20-minute time allocation represents a potential confound, as the specific content of replacement activities may have differed in cognitive demands. These issues are addressed as limitations below. Research instruments and measurement Growth mindset was assessed through a Chinese-adapted version of Dweck’s Implicit Theories of Intelligence Scale, which has demonstrated robust psychometric properties in prior research with Chinese university students [58]. This instrument comprises eight items rated on a six-point Likert scale ranging from 1 (strongly disagree) to 6 (strongly agree), with half the items reverse-coded to control for acquiescence bias. Sample items include “You have a certain amount of intelligence, and you can’t really do much to change it” (reverse-coded) and “You can always substantially change how intelligent you are.” Higher scores reflect stronger endorsement of growth mindset orientations. The L2 Motivational Self System was measured using Taguchi’s validated instrument, which operationalizes Dörnyei’s theoretical framework through three distinct subscales [37]; the questionnaire format and item phrasing followed established procedures for L2 motivation research [59]. The ideal L2 self subscale contains six items capturing students’ aspirational visions of themselves as proficient English users (e.g., “I can imagine myself speaking English with international friends or colleagues”). The ought-to L2 self subscale comprises five items addressing perceived obligations and external expectations (e.g., “I study English because close friends or family members think it is important”). The L2 learning experience subscale includes six items evaluating immediate learning environment perceptions (e.g., “I really enjoy learning English”). All items employed six-point Likert scales, maintaining consistency with the mindset measure. Academic engagement assessment drew upon Reeve’s multidimensional engagement framework, incorporating behavioral, emotional, and cognitive components [60]. The behavioral engagement subscale consisted of five items measuring observable learning actions (e.g., “I participate actively in English class discussions”). Emotional engagement was captured through six items reflecting affective investment (e.g., “I feel excited when learning new English concepts”). Cognitive engagement comprised five items assessing strategic mental effort (e.g., “I try to connect what I learn in English class to things I already know”). Respondents rated all engagement items on six-point scales parallel to those used in other instruments. Figure 2 depicts the theoretical relationships among measured constructs within the analytical framework. As shown in Fig. 2, growth mindset was conceptualized as an exogenous variable influencing both the L2 Motivational Self System components and academic engagement dimensions, with motivational selves serving potential mediating roles in the mindset-engagement pathway. Fig. 2Theoretical Measurement Model of Research Variables Theoretical Measurement Model of Research Variables Confirmatory factor analysis (CFA) was conducted to verify the factorial validity of measurement instruments, and internal consistency was assessed through Cronbach’s alpha coefficients. Table 2 summarizes the psychometric properties observed in the current sample. All scales demonstrated acceptable to excellent reliability, with Cronbach’s α values ranging from 0.82 to 0.91, exceeding conventional thresholds [61]. Individual-scale CFA yielded satisfactory model fit: comparative fit index (CFI) values exceeded 0.90, root mean square error of approximation (RMSEA) remained below 0.08, and standardized root mean square residual (SRMR) fell within acceptable ranges [62]. Table 2Reliability and Validity Indices for Measurement ScalesScaleNumber of ItemsCronbach’s αCFIRMSEASRMRGrowth Mindset80.860.950.0610.045Ideal L2 Self60.890.970.0550.038Ought-to L2 Self50.820.940.0670.051L2 Learning Experience60.870.960.0580.042Academic Engagement160.910.930.0710.048 Table 2Reliability and Validity Indices for Measurement ScalesScaleNumber of ItemsCronbach’s αCFIRMSEASRMRGrowth Mindset80.860.950.0610.045Ideal L2 Self60.890.970.0550.038Ought-to L2 Self50.820.940.0670.051L2 Learning Experience60.870.960.0580.042Academic Engagement160.910.930.0710.048 Reliability and Validity Indices for Measurement Scales Beyond individual-scale analyses, a comprehensive measurement model incorporating all latent variables (growth mindset, three L2MSS components, and three engagement dimensions) was tested to verify discriminant validity. This seven-factor model yielded acceptable fit (CFI = 0.91, RMSEA = 0.058, SRMR = 0.052), and all factor loadings were significant (p < .001, range: 0.58–0.87). The average variance extracted (AVE) for each factor exceeded 0.50, and AVE values surpassed the squared inter-factor correlations for all factor pairs, supporting adequate discriminant validity among the measured constructs. Given the longitudinal design with three measurement waves, measurement invariance was tested following the sequential procedure recommended by Vandenberg and Lance [63]. Configural invariance was established by fitting the measurement model at each time point separately; all three models showed acceptable fit (CFI > 0.90, RMSEA < 0.07). Metric (weak) invariance was then tested by constraining factor loadings to equality across time points. The change in CFI from the configural to metric model was ΔCFI = − 0.006, which fell within the recommended threshold of ΔCFI ≤ 0.01 [64], supporting the conclusion that factor loadings remained stable over time. Scalar (strong) invariance testing, which constrained item intercepts across waves, yielded ΔCFI = − 0.009, also within acceptable limits. These results indicate that observed changes over time reflect genuine developmental change rather than shifts in measurement properties. Data collection occurred at three time points across the semester: baseline assessment (T1) prior to intervention initiation, mid-intervention assessment (T2) after six weeks, and post-intervention assessment (T3) following the complete 12-week program. This longitudinal measurement design enabled examination of both immediate and sustained intervention effects while capturing potential developmental trajectories in the target constructs. Data collection procedures and analytical methods Data collection unfolded across three strategic time points designed to capture both immediate and sustained intervention effects. Baseline assessment (T1) occurred during the first week of the semester, prior to any intervention activities, establishing initial equivalence between experimental and control groups. Mid-intervention measurement (T2) took place at the six-week mark, coinciding with the transition from foundational mindset restructuring modules to more advanced metacognitive strategy components. Post-intervention assessment (T3) was conducted immediately following the 12-week intervention period, allowing evaluation of cumulative program effects [65]. Several quality control measures were implemented to ensure data integrity and minimize attrition. All questionnaires were administered during regular class sessions under standardized conditions, with trained research assistants providing uniform instructions and monitoring completion processes. Students received anonymous identification codes linking their responses across time points while protecting confidentiality. A 96.7% retention rate was achieved across all three waves, with only four participants (two from each group) dropping due to course withdrawal unrelated to the study. Missing data remained minimal and appeared to follow a missing completely at random pattern based on Little’s MCAR test [66]. Statistical analyses proceeded through a planned sequence. First, independent samples t-tests verified baseline equivalence between groups on all measured variables. Second, 2 (Group) × 3 (Time) mixed-design repeated measures ANOVAs tested for within-subject changes and between-group differences in developmental trajectories, with Group × Time interactions serving as the primary test of intervention effects (H1 and H2). Mauchly’s test of sphericity was evaluated for each ANOVA, and Greenhouse–Geisser corrections were applied when the sphericity assumption was violated. Third, to test H3, mediation analyses examined whether L2MSS components mediated the intervention-engagement relationship. Recognizing that cross-sectional mediation at a single time point would be inconsistent with the longitudinal design, the mediation model employed a temporal structure: intervention condition (T1 assignment) served as the independent variable, L2MSS components measured at T2 served as mediators, and academic engagement at T3 served as the outcome, controlling for baseline values at T1. Bias-corrected bootstrap procedures with 5,000 resamples, implemented in Mplus 8.3 [67], generated 95% confidence intervals for indirect effects [68]. Mediation was considered statistically significant when the confidence interval excluded zero [69]. Effect sizes were reported using \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\eta\:}_{p}^{2}$$\end{document} for ANOVA effects and Cohen’s d for between-group comparisons. Following Cohen [70], d values of 0.20, 0.50, and 0.80 were interpreted as small, medium, and large effects respectively, while \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\eta\:}_{p}^{2}$$\end{document} values of 0.01, 0.06, and 0.14 served as benchmarks for small, medium, and large effects. All statistical tests employed two-tailed significance criteria with α = 0.05, and exact p-values are reported throughout. Results Descriptive statistics and baseline equivalence testing Preliminary analyses examined the distributional properties of all measured variables and verified pre-intervention equivalence between experimental and control groups. Table 3 summarizes the descriptive statistics and bivariate correlations among study variables at baseline. As presented in Table 3, means for both groups fell within expected ranges based on scale properties, and standard deviations indicated adequate variability for subsequent analyses. Skewness and kurtosis values remained within acceptable bounds (|skewness| < 2.0, |kurtosis| < 7.0), supporting the normality assumption underlying parametric statistical procedures. Table 3Descriptive Statistics and Correlation Matrix of Study Variables at BaselineVariableM (SD) - Exp.M (SD) - Control1234561. Growth Mindset3.82 (0.71)3.79 (0.68)—2. Ideal L2 Self3.65 (0.84)3.61 (0.79)0.52**—3. Ought-to L2 Self4.12 (0.73)4.08 (0.76)0.34**0.41**—4. L2 Learning Experience3.48 (0.89)3.52 (0.91)0.47**0.58**0.29**—5. Behavioral Engagement3.71 (0.77)3.68 (0.82)0.43**0.56**0.38**0.62**—6. Emotional Engagement3.56 (0.93)3.59 (0.88)0.51**0.63**0.31**0.69**0.71**—7. Cognitive Engagement3.62 (0.81)3.57 (0.85)0.48**0.61**0.35**0.64**0.68**0.74**Exp. Experimental group (n = 60), M Mean, SD Standard DeviationCorrelations are based on the combined sample (N = 120).**p < .01 Table 3Descriptive Statistics and Correlation Matrix of Study Variables at BaselineVariableM (SD) - Exp.M (SD) - Control1234561. Growth Mindset3.82 (0.71)3.79 (0.68)—2. Ideal L2 Self3.65 (0.84)3.61 (0.79)0.52**—3. Ought-to L2 Self4.12 (0.73)4.08 (0.76)0.34**0.41**—4. L2 Learning Experience3.48 (0.89)3.52 (0.91)0.47**0.58**0.29**—5. Behavioral Engagement3.71 (0.77)3.68 (0.82)0.43**0.56**0.38**0.62**—6. Emotional Engagement3.56 (0.93)3.59 (0.88)0.51**0.63**0.31**0.69**0.71**—7. Cognitive Engagement3.62 (0.81)3.57 (0.85)0.48**0.61**0.35**0.64**0.68**0.74**Exp. Experimental group (n = 60), M Mean, SD Standard DeviationCorrelations are based on the combined sample (N = 120).**p < .01 Descriptive Statistics and Correlation Matrix of Study Variables at Baseline Exp. Experimental group (n = 60), M Mean, SD Standard Deviation Correlations are based on the combined sample (N = 120).**p < .01 Independent samples t-tests confirmed baseline equivalence across all measured constructs. For growth mindset, no significant difference emerged between experimental (M = 3.82, SD = 0.71) and control groups (M = 3.79, SD = 0.68), t(118) = 0.24, p = .811, d = 0.04. Similarly, the three L2 Motivational Self System components showed comparable baseline levels: ideal L2 self, t(118) = 0.27, p = .789, d = 0.05; ought-to L2 self, t(118) = 0.29, p = .770, d = 0.05; and L2 learning experience, t(118) = -0.25, p = .805, d = -0.05. Academic engagement dimensions also demonstrated initial equivalence, with behavioral engagement t(118) = 0.21, p = .835, d = 0.04, emotional engagement t(118) = -0.18, p = .854, d = -0.03, and cognitive engagement t(118) = 0.33, p = .742, d = 0.06. These non-significant differences validate the assumption that observed post-intervention differences would not stem from pre-existing group disparities. Correlation analyses at baseline revealed patterns consistent with theoretical expectations. Growth mindset showed moderate to strong positive associations with all L2MSS components (r = .34 to 0.52, all p < .01), and the ideal L2 self exhibited particularly strong correlations with both learning experience (r = .58, p < .01) and emotional engagement (r = .63, p < .01). Engagement dimensions correlated strongly with one another (r = .68 to 0.74, all p < .01), supporting their conceptualization as interrelated facets of a broader engagement construct while remaining sufficiently distinct to justify separate analysis. Ought-to L2 self, by contrast, showed weaker correlations than ideal L2 self and learning experience—a pattern echoing the theoretical expectation that externally driven motivational orientations connect less tightly to self-determined engagement than internally generated aspirational identities. All three engagement dimensions correlated comparably with growth mindset (r = .43 to 0.51), suggesting that beliefs about ability malleability may influence behavioral, emotional, and cognitive engagement in roughly uniform fashion. These bivariate associations are non-directional and should not be read as causal claims; they primarily confirm that the constructs relate in theoretically expected ways. Effects of growth mindset intervention on L2 motivational self system Before examining intervention effects on L2MSS components, it is essential to verify that the intervention successfully altered its proximal target—growth mindset itself. A 2 (Group) × 3 (Time) repeated measures ANOVA on growth mindset scores revealed a significant Group × Time interaction, F(2, 236) = 42.15, p < .001, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\eta\:}_{p}^{2}$$\end{document} = 0.263. Simple effects analyses showed no group difference at T1, F(1, 118) = 0.09, p = .811, d = 0.04, confirming baseline equivalence. By T2, the experimental group (M = 4.35, SD = 0.63) scored significantly higher than the control group (M = 3.84, SD = 0.66), F(1, 118) = 21.47, p < .001, d = 0.79, representing a medium-to-large effect. At T3, this gap widened further, with the experimental group (M = 4.72, SD = 0.58) substantially exceeding the control group (M = 3.81, SD = 0.70), F(1, 118) = 67.82, p < .001, d = 1.41, a large effect by Cohen’s [70] standards. Figure 3 illustrates these diverging trajectories. These results confirm that the intervention effectively reshaped students’ implicit theories about ability, serving as a necessary precondition for examining downstream effects on L2MSS and engagement. Fig. 3Developmental Trajectories of Growth Mindset Scores for Experimental and Control Groups Developmental Trajectories of Growth Mindset Scores for Experimental and Control Groups As shown in Table 4, the repeated measures ANOVA on growth mindset scores yielded significant main effects for both group, F(1, 118) = 35.41, p < .001, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\eta\:}_{p}^{2}$$\end{document} = 0.231, and time, F(2, 236) = 51.26, p < .001, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\eta\:}_{p}^{2}$$\end{document} = 0.303, as well as a significant Group × Time interaction, F(2, 236) = 42.15, p < .001, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\eta\:}_{p}^{2}$$\end{document} = 0.263, confirming that the two groups followed divergent developmental trajectories in mindset change across the semester. Table 4Repeated Measures ANOVA Results for Growth MindsetSourceF \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\eta\:}_{p}^{2}$$\end{document} Group35.41***0.231Time51.26***0.303Group × Time42.15***0.263df for Group = 1, 118; df for Time = 2, 236; df for Group × Time = 2, 236. ***p < .001 Table 4Repeated Measures ANOVA Results for Growth MindsetSourceF \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\eta\:}_{p}^{2}$$\end{document} Group35.41***0.231Time51.26***0.303Group × Time42.15***0.263df for Group = 1, 118; df for Time = 2, 236; df for Group × Time = 2, 236. ***p < .001 Repeated Measures ANOVA Results for Growth Mindset df for Group = 1, 118; df for Time = 2, 236; df for Group × Time = 2, 236. ***p < .001 A series of 2 (Group: experimental vs. control) × 3 (Time: T1, T2, T3) mixed-design repeated measures ANOVAs were conducted to examine intervention effects on the three components of the L2 Motivational Self System. Mauchly’s test indicated that sphericity assumptions were met for all analyses (all p > .05), permitting interpretation of univariate F statistics without corrections. Table 5 summarizes the omnibus ANOVA results for each motivational dimension. Table 5Repeated Measures ANOVA Results for L2 Motivational Self System ComponentsSourceIdeal L2 SelfOught-to L2 SelfL2 Learning ExperienceF \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\eta\:}_{p}^{2}$$\end{document} F \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\eta\:}_{p}^{2}$$\end{document} F \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\eta\:}_{p}^{2}$$\end{document} Group28.67***0.19612.43**0.09531.52***0.211Time47.82***0.28823.16***0.16452.39***0.307Group × Time35.21***0.23018.74***0.13738.65***0.247df for Group = 1, 118; df for Time = 2, 236; df for Group × Time = 2, 236. **p < .01. ***p < .001 Table 5Repeated Measures ANOVA Results for L2 Motivational Self System ComponentsSourceIdeal L2 SelfOught-to L2 SelfL2 Learning ExperienceF \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\eta\:}_{p}^{2}$$\end{document} F \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\eta\:}_{p}^{2}$$\end{document} F \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\eta\:}_{p}^{2}$$\end{document} Group28.67***0.19612.43**0.09531.52***0.211Time47.82***0.28823.16***0.16452.39***0.307Group × Time35.21***0.23018.74***0.13738.65***0.247df for Group = 1, 118; df for Time = 2, 236; df for Group × Time = 2, 236. **p < .01. ***p < .001 Repeated Measures ANOVA Results for L2 Motivational Self System Components df for Group = 1, 118; df for Time = 2, 236; df for Group × Time = 2, 236. **p < .01. ***p < .001 The results in Table 5 indicate significant main effects and interactions across all three motivational dimensions, though with varying magnitudes. For ideal L2 self, both the group main effect, F(1, 118) = 28.67, p < .001, and the time main effect, F(2, 236) = 47.82, p < .001, reached statistical significance. The Group × Time interaction was also significant, F(2, 236) = 35.21, p < .001, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\eta\:}_{p}^{2}$$\end{document} = 0.230, indicating that developmental trajectories differed between experimental and control conditions. This interaction accounted for approximately 23% of the variance in ideal L2 self changes. Simple effects analyses decomposed the significant interactions by examining group differences at each time point. At baseline (T1), experimental (M = 3.65, SD = 0.84) and control groups (M = 3.61, SD = 0.79) did not differ on ideal L2 self, F(1, 118) = 0.12, p = .732. By mid-intervention (T2), a marginally significant difference emerged, with the experimental group (M = 4.08, SD = 0.78) exceeding the control group (M = 3.73, SD = 0.81), F(1, 118) = 9.23, p = .003, d = 0.44. This gap widened considerably by post-intervention (T3), where experimental participants (M = 4.52, SD = 0.71) scored markedly higher than controls (M = 3.68, SD = 0.79), F(1, 118) = 58.47, p < .001, d = 1.12, reflecting a large between-group effect. Ought-to L2 self displayed a similar but somewhat attenuated pattern. The significant Group × Time interaction, F(2, 236) = 18.74, p < .001, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\eta\:}_{p}^{2}$$\end{document} = 0.137, indicated differential change trajectories. Simple effects testing showed that while groups remained equivalent at T1, F(1, 118) = 0.14, p = .711, the experimental group demonstrated significantly higher ought-to L2 self scores at both T2, F(1, 118) = 6.82, p = .010, d = 0.38, and T3, F(1, 118) = 24.31, p < .001, d = 0.71. L2 learning experience showed the largest Group × Time interaction effect size among the three components, F(2, 236) = 38.65, p < .001, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\eta\:}_{p}^{2}$$\end{document} = 0.247. Within-time comparisons revealed no baseline differences, F(1, 118) = 0.09, p = .769, but significant divergence at T2, F(1, 118) = 12.54, p = .001, d = 0.51, and at T3, F(1, 118) = 63.72, p < .001, d = 1.16. These results are consistent with Hypothesis 1, indicating that growth mindset intervention was associated with increases in all three components of the L2 Motivational Self System. The largest effect sizes were observed for ideal L2 self and learning experience, with more moderate effects on ought-to L2 self. Effects of growth mindset intervention on academic engagement and mediation analysis Parallel repeated measures ANOVAs examined intervention effects on academic engagement. For overall engagement, significant main effects emerged for both group, F(1, 118) = 32.18, p < .001, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\eta\:}_{p}^{2}$$\end{document} = 0.214, and time, F(2, 236) = 54.63, p < .001, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\eta\:}_{p}^{2}$$\end{document} = 0.316. The crucial Group × Time interaction achieved significance, F(2, 236) = 40.87, p < .001, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\eta\:}_{p}^{2}$$\end{document} = 0.257, indicating differential developmental trajectories between conditions. Simple effects analyses revealed that while groups showed baseline equivalence at T1, F(1, 118) = 0.18, p = .674, the experimental group demonstrated significantly higher engagement at T2, F(1, 118) = 14.76, p < .001, d = 0.56, with this advantage amplifying at T3, F(1, 118) = 68.34, p < .001, d = 1.20. Examining engagement dimensions separately yielded consistent patterns. Behavioral engagement showed a significant Group × Time interaction, F(2, 236) = 28.53, p < .001, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\eta\:}_{p}^{2}$$\end{document} = 0.195, with the experimental group outperforming controls by T3 (d = 0.98). Emotional engagement demonstrated even stronger intervention sensitivity, F(2, 236) = 44.21, p < .001, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\eta\:}_{p}^{2}$$\end{document} = 0.273, reflecting substantial affective gains in the experimental condition (T3: d = 1.15). Cognitive engagement paralleled these results with F(2, 236) = 35.67, p < .001, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\eta\:}_{p}^{2}$$\end{document} = 0.232, and post-intervention Cohen’s d = 1.08. These findings collectively supported Hypothesis 2, indicating that growth mindset intervention enhanced academic engagement across behavioral, emotional, and cognitive dimensions. Hypothesis 3 was tested by examining whether L2 Motivational Self System components mediated the relationship between intervention condition and academic engagement. Mediation analyses employed bias-corrected bootstrap procedures with 5,000 resamples within a temporally ordered model: intervention assignment at T1 predicted L2MSS components at T2 (mediators), which in turn predicted academic engagement at T3, with baseline values of all variables included as covariates. The total effect of intervention on engagement was decomposed into a direct effect and indirect effects operating through the three L2MSS components. Table 6 presents the bootstrap mediation analysis results. As presented in Table 6, the total intervention effect on academic engagement proved substantial (β = 0.847, 95% CI [0.673, 1.028]), indicating that the experimental group scored approximately 0.85 standard deviations higher on engagement than controls after accounting for all pathways. This total effect decomposed into a modest direct effect (β = 0.182, 95% CI [0.041, 0.336]) and significant indirect effects operating through motivational mediators. Table 6Bootstrap Mediation Analysis Results for L2 Motivational Self SystemPathwayPoint Estimate (β)SE95% CI Lower95% CI UpperTotal Effect (c)0.8470.0890.6731.028Direct Effect (c’)0.1820.0740.0410.336Indirect via Ideal L2 Self0.4280.0610.3150.556Indirect via Ought-to L2 Self0.0960.0380.0270.178Indirect via L2 Learning Experience0.1410.0450.0580.236Total Indirect Effect0.6650.0760.5210.821 Table 6Bootstrap Mediation Analysis Results for L2 Motivational Self SystemPathwayPoint Estimate (β)SE95% CI Lower95% CI UpperTotal Effect (c)0.8470.0890.6731.028Direct Effect (c’)0.1820.0740.0410.336Indirect via Ideal L2 Self0.4280.0610.3150.556Indirect via Ought-to L2 Self0.0960.0380.0270.178Indirect via L2 Learning Experience0.1410.0450.0580.236Total Indirect Effect0.6650.0760.5210.821 Bootstrap Mediation Analysis Results for L2 Motivational Self System Among the three mediating pathways, ideal L2 self showed the largest indirect effect (β = 0.428, 95% CI [0.315, 0.556]), accounting for approximately 51% of the total effect. L2 learning experience contributed a moderate indirect effect (β = 0.141, 95% CI [0.058, 0.236]), representing about 17% of the total effect, while ought-to L2 self showed the smallest mediating influence (β = 0.096, 95% CI [0.027, 0.178]), accounting for roughly 11%. Figure 4 presents a conceptual illustration of the mediation model with standardized path coefficients from the estimated model. It should be noted that this figure represents the regression-based mediation results rather than a full structural equation model; it is included to visualize the relative magnitudes of different pathways. Fig. 4Standardized Path Coefficients for the Mediation Model (Conceptual Illustration). Note. Values on paths represent standardized regression coefficients from the bootstrap mediation analysis with temporal structure (T1 intervention → T2 mediators → T3 outcome), controlling for baseline values. This figure serves as a conceptual summary; it does not represent a full latent variable SEM model Standardized Path Coefficients for the Mediation Model (Conceptual Illustration). Note. Values on paths represent standardized regression coefficients from the bootstrap mediation analysis with temporal structure (T1 intervention → T2 mediators → T3 outcome), controlling for baseline values. This figure serves as a conceptual summary; it does not represent a full latent variable SEM model The total indirect effect through all three motivational components (β = 0.665, 95% CI [0.521, 0.821]) was larger than the direct effect, indicating that intervention effects on engagement were transmitted primarily through motivational pathways. The proportion of the total effect accounted for by indirect pathways was 78.5%. These findings are consistent with Hypothesis 3, suggesting that the L2MSS—particularly the ideal L2 self—may function as a mediating mechanism in the intervention-engagement relationship. However, it is important to acknowledge that even with temporal ordering of variables, these results do not establish definitive causal mediation; they are best interpreted as providing evidence consistent with the hypothesized mediational account rather than confirming a causal chain. Discussion The present investigation yielded evidence that a growth mindset intervention can meaningfully reshape non-English major students’ L2MSS and academic engagement patterns. These findings invite careful consideration of the underlying psychological mechanisms and their broader implications for second language pedagogy. Figure 5 presents the comparative trajectories of all three L2MSS dimensions across groups and time points. The experimental group exhibited consistent upward movement across all three motivational facets, while the control group remained relatively stable. The steepest growth gradient appeared for ideal L2 self and learning experience in the experimental condition, with ought-to L2 self showing more modest gains. Fig. 5Group Comparisons of L2 Motivational Self System Dimensions Across Time Group Comparisons of L2 Motivational Self System Dimensions Across Time The particularly strong intervention effects on ideal L2 self merit close examination through the lens of possible selves theory and self-discrepancy theory. By dismantling fixed ability beliefs, the intervention may have removed psychological barriers that previously constrained students’ aspirational visions. When learners believe linguistic competence is largely predetermined, envisioning ambitious future identities as proficient English users may feel unrealistic. Once students internalized that ability can develop through strategic effort, they may have gained a sense of permission—so to speak—to construct more elaborate ideal L2 selves. This interpretation aligns with Kırmızı et al.’s [10] finding that growth mindset positively predicted ideal L2 self among Turkish EFL learners, and with Zarrinabadi et al.’s [42] observation that growth language mindset was associated with stronger L2 future self-guides. The attribution training component likely enhanced ideal self accessibility by altering how students processed past learning experiences. Rather than interpreting previous difficulties as evidence of inherent inadequacy, students learned to view setbacks as temporary states reflecting insufficient strategy deployment. This reframing may have transformed failure narratives from identity threats into developmental feedback, creating psychological space for constructing positive future selves—a process consistent with the attributional retraining mechanisms described in Lou and Noels’ Language Mindset Meaning System framework [8, 9]. The metacognitive strategy instruction further provided concrete pathways connecting present selves to ideal futures. Vivid ideal selves motivate behavior only when accompanied by plausible action plans for bridging the discrepancy [51]; by equipping students with specific learning strategies and self-regulation techniques, the intervention made aspirational identities feel genuinely achievable. The comparatively weaker intervention effects on ought-to L2 self deserve more extended consideration than a simple appeal to self-discrepancy dynamics might suggest. Several factors likely contributed to this pattern. First, ought-to L2 selves primarily reflect internalized external expectations and perceived obligations—constructs rooted in social relationships and institutional structures that are less directly amenable to shifts in individual epistemological beliefs. Parents still maintain expectations about English proficiency, employment markets still demand language credentials, and academic programs still impose requirements. These external realities persist regardless of how students think about the malleability of ability. Second, in the Chinese educational context, ought-to L2 selves often carry particularly strong weight given the collective orientation of family expectations and the high-stakes nature of English examinations [37]. The intervention may have had limited leverage over these deeply embedded social-contextual factors. Third, from a self-determination perspective, ought-to selves typically carry an avoidance orientation focused on preventing negative outcomes, whereas growth mindset intervention—with its emphasis on possibility, development, and achievement—naturally resonates more strongly with approach-oriented ideal selves. Fourth, the intervention design itself may not have specifically targeted ought-to self dynamics. Future research might explore whether supplementing growth mindset intervention with components that directly reframe external expectations—for example, helping students reinterpret parental pressure as developmental support—could strengthen effects on this dimension. The substantial improvements in L2 learning experience were notable. This dimension captures students’ immediate, situated engagement with classroom activities—precisely the arena where growth mindset principles were enacted and reinforced daily. Students in the experimental condition encountered repeated experiences of growth-oriented feedback, challenge-embracing encouragement, and strategic effort recognition. These accumulated microexperiences appear to have reshaped the emotional and cognitive texture of their English learning. This finding resonates with the implementation fidelity literature reviewed by Burnette, Knouse, and colleagues [71], who emphasized that the manner of delivery—particularly integration with daily practice rather than isolated lessons—substantially moderates intervention effectiveness. The mediation findings carry theoretical weight, though they should be interpreted with appropriate caution given the quasi-experimental design. That intervention effects on engagement operated primarily through motivational pathways, particularly ideal L2 self, illuminates a plausible psychological architecture linking implicit theories to behavioral outcomes. Students did not simply become more engaged because they believed ability was malleable—the relationship appears more complex. Growth beliefs may first reshape motivational self-constructs by expanding perceived possibilities, and these transformed motivational identities may then energize sustained engagement by providing compelling reasons for investing effort. This interpretive account is consistent with the mediational pathway proposed in Lou and Noels’ [8] framework and with recent findings showing that growth language mindset affects L2 outcomes through autonomous motivation and engagement [72]. It should be stressed, however, that the L2MSS functions not solely as a mediating mechanism in this study but also as a meaningful outcome in its own right—enhanced L2 motivational selves have intrinsic value for learners beyond their role in predicting engagement. There is an important caveat: the finding that growth mindset intervention operated primarily through motivational mediators does not mean that interventions failing to trigger shifts in domain-specific motivational identities will necessarily be ineffective. Alternative pathways—such as direct effects on self-regulation or resilience—may also contribute to engagement enhancement, though they were not measured in the present study. Comparing these results with the broader growth mindset literature reveals both convergences and tensions. The effect sizes observed here (d = 1.08 to 1.20 at T3 for engagement dimensions) substantially exceed those reported in recent meta-analyses. Macnamara and Burgoyne [13] found a near-zero average effect of mindset interventions on achievement after correcting for study quality, while Burnette et al. [14] reported more favorable effects (d = 0.14 for academic achievement) among targeted subpopulations with high-fidelity implementation. Several factors may explain the larger effects in the present study: the intervention was substantially longer (12 weeks vs. typical brief interventions), deeply integrated into classroom instruction, and measured motivational and engagement outcomes rather than distal achievement. Recent commentary by Tipton, Yeager, and colleagues [15] has emphasized that heterogeneity in mindset intervention effects should be expected, and that context-specific implementations targeting at-risk populations may yield larger effects than broad-based approaches. The non-English major population in this study—students facing documented motivational challenges in English learning—may represent a subgroup particularly responsive to mindset-based approaches. The differential intervention effects across engagement dimensions also warrant consideration. Emotional engagement showed the largest gains, which may reflect the affective transformation accompanying reconceptualized challenge encounters and enhanced self-efficacy. Behavioral engagement improvements were also pronounced, perhaps because observable actions like participation and persistence respond relatively directly to motivational shifts. Cognitive engagement gains, while significant, were somewhat smaller. Deep cognitive processing strategies—elaboration, critical analysis, knowledge integration—demand metacognitive sophistication beyond motivational willingness alone. This pattern suggests that maximizing cognitive engagement gains might require supplementing growth mindset intervention with explicit instruction in advanced learning strategies. Conclusion This investigation employed a quasi-experimental longitudinal design to examine how growth mindset intervention influences non-English major students’ L2MSS and academic engagement across a semester-long period. The results indicate that systematic intervention targeting implicit theories of language learning ability can generate meaningful improvements in both motivational self-constructs and engagement behaviors. Students who participated in the growth mindset intervention demonstrated enhanced ideal L2 selves, ought-to L2 selves, and L2 learning experiences compared to controls. These motivational gains were accompanied by elevated academic engagement across behavioral, emotional, and cognitive dimensions. The mediation analyses revealed that the L2MSS, particularly the ideal L2 self, may serve as a psychological pathway through which growth mindset intervention influences academic engagement. The bulk of the intervention effect on engagement was accounted for by indirect pathways through motivational variables, with the ideal L2 self contributing the largest share. This pattern suggests—though does not definitively establish—that growth mindset intervention reshapes the motivational identities that energize and sustain engagement, rather than simply prompting students to exert more effort. This study advances existing scholarship through several contributions. Theoretically, it integrates growth mindset theory and L2MSS within a framework grounded in possible selves and self-discrepancy theories, demonstrating that these traditionally separate research traditions offer complementary insights when combined. Methodologically, the longitudinal design with three measurement waves captures developmental trajectories that cross-sectional approaches cannot access, revealing progressive strengthening of effects over time. Mechanistically, the temporally structured mediation analyses provide evidence for hypothesized psychological pathways, extending beyond the predominantly correlational evidence available in prior work. On practical grounds, the intervention’s effectiveness suggests that educators can enhance student motivation and engagement through accessible adjustments to classroom practice. Three principles deserve particular attention: integrating brief, evidence-informed lessons on the malleability of language learning ability into regular instruction; replacing normative or ability-focused feedback with process-oriented comments that highlight effort, strategy, and improvement; and cultivating ideal L2 self vividness through periodic future-self visualization paired with metacognitive strategy guidance. These elements can be folded into existing curricula without major resource investment, and they can be calibrated to the constraints of specific instructional contexts. Several limitations temper interpretation and generalization. The sample came from a single university in eastern China, which may limit generalizability to different institutional types, regions, or cultural contexts. The cluster randomization at the class level means that only four classes were included, raising the possibility of clustering effects that were not formally modeled in the analyses; future studies should consider multilevel modeling approaches when randomizing at the class level. The 20-minute intervention segments, while designed to maintain instructional equivalence, may have introduced confounds related to the specific nature of replacement activities in the control group. Critically, the post-intervention assessment (T3) was conducted immediately after the 12-week intervention period. This timing means that the observed effects reflect cumulative immediate post-intervention outcomes rather than sustained effects after the intervention has ended. Without delayed follow-up assessments, it remains uncertain whether the documented gains would persist, intensify, or dissipate over time without ongoing reinforcement. Future studies should incorporate follow-up measurements at intervals of three to six months or longer. Additionally, the mediation analysis, while employing temporal ordering of variables (T1 → T2 → T3), does not fully resolve causal ambiguity inherent in quasi-experimental designs. The absence of random assignment at the individual level and the possibility of unmeasured confounders at the class level mean that causal claims should be made cautiously. Future investigations might productively extend this work by: expanding sample diversity across multiple institutions and proficiency levels; employing individual-level randomization where feasible; examining moderators such as initial mindset orientation, baseline proficiency, and cultural value orientations; and comparing domain-general growth mindset interventions with language-specific mindset interventions [8, 9] to determine whether domain specificity enhances effectiveness. Despite these limitations, the present findings suggest that thoughtfully designed growth mindset interventions can reshape the motivational and engagement patterns of non-English major students in Chinese higher education, operating in part through enhanced construction of ideal L2 selves. This evidence provides both theoretical insight into motivational dynamics in second language learning and practical guidance for educators seeking to address persistent challenges of student motivation and engagement.