Abstract
While teacher support, academic self-efficacy, and learning engagement are recognized as influencing L2 willingness to communicate (WTC) in computer-assisted language learning (CALL), the specific mechanisms linking teacher support to WTC via these mediators remain unclear. Drawing on a sample of undergraduates from a Chinese EFL context, this study investigated the chain mediating roles of academic self-efficacy and learning engagement, integrating control-value theory (CVT) and the WTC pyramid model. Data were collected from 709 undergraduates at three universities in Shandong Province, China, and analyzed using Mplus 8.3 and SPSS 24.0. Results revealed that teacher support had a moderate direct effect on WTC (β = .221, p < .001). Crucially, academic self-efficacy and learning engagement sequentially mediated this relationship. These findings, grounded in a Chinese cultural-educational setting, underscore the importance of teacher support in CALL environments and elucidate a cognitive-motivational pathway through which it enhances communicative willingness. Implications suggest educators should integrate technical and affective scaffolding (multimodal scaffolding) to leverage control-value synergies, while CALL designers should prioritize features that maintain appraisal salience to sustain engagement in autonomous learning contexts.
Plain Language Summary
This study looked at how teacher support helps students feel more willing to talk in a second language when using learning apps. We know that feeling confident (academic self-efficacy) and being actively involved (learning engagement) are important, but exactly how teacher support uses these to boost willingness to communicate wasn’t clear. Researchers surveyed 709 university students in China. They found that when teachers offer good support while students use language apps, it directly helps students feel more willing to talk. But importantly, teacher support also works indirectly: it first makes students feel more confident in their ability to learn. This increased confidence then makes students participate more actively in their learning. Finally, this active participation leads to a greater willingness to actually use the language and communicate. This shows teacher support in app-based learning is crucial. Teachers should blend both encouragement and help with the technology. App designers should focus on features that keep students motivated and engaged, especially when learning independently.
Introduction
The integration of computer-assisted language learning (CALL) into modern language education has redefined pedagogical paradigms, particularly in AI application contexts where blended and fully online environments have become normative (Godwin-Jones, 2023). While CALL’s capacity to enhance linguistic interactivity through multimodal resources is well-documented (Chapelle, 2001), its psychological ramifications on learners’ willingness to communicate (WTC) in a second language (L2) remain underexplored. This gap is critical given that WTC—conceptualized by (MacIntyre et al., 1998) as the volitional readiness to engage in L2 discourse—serves as the ultimate behavioral manifestation of language learning success (MacIntyre & Doucette, 2010). Grounded in a hierarchical pyramid model, WTC is posited to emerge from the interplay of situational antecedents, motivational orientations, and affective-cognitive states (MacIntyre et al., 1998). This motivational dimension is powerfully captured by Dörnyei (2009) L2 Motivational Self System (L2MSS), which conceptualizes motivation as being driven by the learner’s “Ideal L2 Self” and “Ought-to L2 Self.” The ultimate goal of L2 learning can be seen as the realization of the Ideal L2 Self, with WTC representing a critical behavioral step toward this ideal. In culturally distinct settings such as Chinese EFL classrooms, this model has been adapted to reflect sociocultural dynamics; Wen and Clément (2003) argue that the transformation of communicative desire into actual WTC hinges on societal norms (e.g., face-saving tendencies) and learner-specific traits. Among these traits, growth-oriented language mindsets—the belief that L2 ability is malleable—are crucial as they underpin the very possibility of self-development central to the L2MSS (Lou & Noels, 2019). Empirical studies corroborate the role of CALL environments in modulating these factors, with blended learning designs fostering supportive social ecosystems (Zhang, 2020) and digital tools mitigating anxiety to enhance emotional readiness for communication (Leeds & Maurer, 2009).
Within this evolving landscape, teacher support emerges as a pivotal yet multifaceted construct. In CALL contexts, instructors transcend traditional roles to provide technical guidance on digital tool usage, affective reassurance to alleviate technostress, and cognitive scaffolding to optimize task mastery (Joy & Seyed, 2018). From the perspective of the L2MSS, teacher support can be viewed as a key environmental factor that shapes the Ought-to L2 Self and facilitates the journey toward the Ideal L2 Self. While existing research acknowledges the correlation between teacher support and L2 outcomes (Guilloteaux & Dörnyei, 2008) the cognitive-motivational mechanisms underlying this relationship—particularly those linking support to WTC through sequential mediators—remain opaque. This opacity persists despite theoretical advances in two complementary frameworks: (MacIntyre et al., 1998) WTC pyramid and (Pekrun, 2006) control-value theory (CVT). The former situates teacher support within the social-individual layer of WTC antecedents, emphasizing its cultural salience in collectivist educational settings (Wen & Clément, 2003). The latter, CVT, provides a micro-level lens by positing that environmental inputs (e.g., teacher interventions) shape learners’ control appraisals (such as self-efficacy in navigating CALL tasks) and value appraisals (perceived relevance of L2 communication), thereby eliciting achievement emotions (e.g., enjoyment) that propel or inhibit WTC. Thus, CVT offers the precise psychological machinery—the appraisal processes—that connects the macro-level motivators of the L2MSS and the pyramid model to the ultimate outcome of WTC. Notably, recent work on language mindsets aligns with CVT’s tenets: Learners who internalize a growth mindset—viewing L2 competence as malleable rather than fixed—demonstrate stronger control appraisals and resilience in technology-mediated interactions (Lou & Noels, 2019).
Synthesizing these frameworks, this study proposes a chain mediation model wherein teacher support in CALL environments sequentially enhances self-efficacy (control appraisal), amplifies learning engagement (value-driven behavior), and ultimately elevates WTC. Such a model addresses two persistent gaps. First, prior studies have predominantly examined isolated mediators (e.g., self-efficacy or engagement alone) rather than their sequential interplay (Khajavy et al., 2018). Second, the integration of macro-level WTC antecedents (per the pyramid model) and macro-level motivational drivers (per the L2MSS) with micro-level appraisal-emotion processes (per CVT) remains theoretically fragmented, particularly in digital learning contexts. By bridging these gaps, the study provides a more comprehensive theoretical account that situates CVT’s micro-processes within the broader context of L2 motivation (L2MSS). This integrated model offers actionable insights for educators. For instance, designing scaffolded feedback mechanisms that reinforce growth mindsets may strengthen learners’ control appraisals, thereby channeling engagement toward authentic L2 communication. The present study aims to illuminate the nuanced pathways through which teacher support—when thoughtfully orchestrated in CALL ecosystems—can transform communicative potential into sustained L2 engagement.
Literature Review
Teacher Support and WTC: Direct and Indirect Pathways
In CALL, teacher support has technical, affective, and cognitive dimensions. Technical support helps with tool use but risks reducing autonomy (Kong et al., 2025; Zhai et al., 2024). Affective support eases anxiety and varies by modality (Carver et al., 2021). Cognitive support aids understanding but requires a balance with challenge to avoid shallow learning (Jeon, 2023). All dimensions should be adjusted to fit learners’ needs and technology.
The relationship between teacher support and L2 WTC in CALL environments is both direct and mediated by cognitive-motivational mechanisms, as posited by (MacIntyre et al., 1998) situational model and (Pekrun, 2006) CVT. Empirical studies consistently report a positive correlation between teacher support and WTC, yet the strength and nature of this association vary across CALL modalities. For instance, in synchronous video-based interactions, real-time teacher scaffolding (e.g., corrective feedback during Zoom debates) has been shown to directly enhance learners’ volitional readiness to communicate by reducing situational anxiety and increasing perceived safety (Hejazi et al., 2023). Conversely, in asynchronous text-based forums, where delayed feedback may attenuate immediacy, the direct effect of teacher support on WTC diminishes, suggesting that technological affordances moderate this relationship (Carver et al., 2021). These findings align with (MacIntyre et al., 1998) WTC pyramid, which positions teacher support as a critical social-individual antecedent within the broader ecosystem of situational factors (e.g., task type, communication medium).
However, the direct pathway alone inadequately explains how teacher support translates into sustained communicative behaviors. CVT provides a complementary lens by emphasizing the role of control-value appraisals as mediators. For example, learners who receive cognitive support (e.g., metacognitive prompts in AI-driven writing tasks) develop stronger control appraisals, particularly self-efficacy in navigating digital tools, which in turn foster positive achievement emotions (e.g., enjoyment) that propel WTC (Khajavy et al., 2018; Pekrun, 2006). Similarly, affective support (e.g., verbal encouragement in VR simulations) enhances learners’value appraisals by framing L2 communication as intrinsically rewarding, thereby amplifying behavioral engagement and WTC (Reinders & Wattana, 2015). Yet, prior research has predominantly examined these mediators—self-efficacy and engagement—in isolation, neglecting their sequential interplay. For instance, Sun and Wang (2020) demonstrated that teacher support boosts WTC via self-efficacy but overlooked how efficacy beliefs might further channel into engagement. Conversely, Svalberg (2009) engagement-focused studies treated self-efficacy as a static trait rather than a dynamic mediator shaped by teacher interventions.
This theoretical fragmentation becomes particularly salient in CALL contexts, where the integration of (MacIntyre et al., 1998) pyramid and CVT could reconcile macro-level antecedents with micro-level appraisal processes. The WTC pyramid identifies teacher support as a distal social factor, while CVT elucidates the proximal psychological mechanisms (control-value appraisals → emotions → engagement) through which such support operates. A synthesis of these frameworks suggests that teacher support in CALL does not merely predict WTC but activates a causal chain: Technical and cognitive scaffolding strengthens learners’ control appraisals, namely their self-efficacy, which then drive value-driven engagement (e.g., persistent participation in gamified tasks), ultimately culminating in WTC. Preliminary evidence for such a pathway exists in non-CALL contexts; Teng and Zhang (2020) found that classroom teacher support indirectly enhanced WTC through self-efficacy and engagement sequentially. However, CALL-specific dynamics—such as the role of multimodal feedback in shaping appraisals—remain underexplored, leaving a critical gap in understanding how technology-mediated teacher interventions uniquely orchestrate these mediators.
Self-Efficacy in CALL: Bridging Teacher Support and Engagement
Self-efficacy, defined as learners’ beliefs in their capability to organize and execute actions required to achieve specific goals (Bandura, 1997), serves as a critical mediator between teacher support and learning engagement in CALL environments (Gong & Xu, 2024). Grounded in (Pekrun, 2006) CVT, self-efficacy constitutes a core component of control appraisals—learners’ judgments of their ability to manage task demands—while engagement reflects value-driven behaviors sustained by perceived task relevance. Within (MacIntyre et al., 1998) WTC pyramid, self-efficacy operates at the affective-cognitive layer, modulating how social-individual antecedents (e.g., teacher support) translate into motivational readiness for L2 communication.
Teacher support in CALL directly cultivates self-efficacy through multiple pathways. Technical support contributes to reducing learners’ cognitive burden and improving their competence in using digital learning tools by boosting their self-efficacy (Lai, 2015). Cognitive support, including formative feedback on AI-generated writing tasks, further reinforces efficacy beliefs by helping learners attribute errors to controllable strategies rather than fixed deficiencies (Song & Song, 2023). Affective support, exemplified by instructors’ verbal encouragement during video-mediated role-plays, mitigates anxiety and fosters resilience, enabling learners to reinterpret challenges as opportunities for growth (Yeager et al., 2022). These mechanisms resonate with CVT’s emphasis on environmental inputs (teacher support) shaping control appraisals, notably self-efficacy, which subsequently regulate achievement emotions and engagement (Pekrun, 2006).
Empirical studies corroborate the mediating role of self-efficacy. For instance, Sun and Wang (2020) demonstrated that teacher scaffolding in online discussion forums enhanced learners’ self-efficacy, which partially mediated the relationship between support and WTC. Similarly, Shafiee Rad (2024) found that technical assistance in virtual exchange programs elevated learners’ confidence in cross-cultural communication, indirectly promoting sustained participation. However, these studies narrowly frame self-efficacy as a standalone mediator, neglecting its sequential interplay with engagement—a gap that becomes salient when viewed through (MacIntyre et al., 1998) pyramid. The pyramid posits that affective-cognitive factors (e.g., self-efficacy) must converge with behavioral engagement to actualize WTC, yet prior research often isolates these constructs.
This theoretical omission is particularly consequential in CALL contexts, where technology-mediated tasks demand continuous adaptation. For example, learners with high self-efficacy in using gamified vocabulary apps may initially engage behaviorally (e.g., frequent app usage), but without cognitive engagement (e.g., deep processing of semantic networks), such behaviors rarely culminate in authentic WTC (Fredricks et al., 2004). CVT clarifies this dynamic: Self-efficacy (control appraisal) initiates engagement (value-driven effort), but only when learners appraise tasks as meaningful (value appraisal) does engagement translate into volitional communication. Preliminary evidence for this chain exists in hybrid classrooms; Guo et al. (2023) reported that self-efficacy fostered by teacher feedback sequentially enhanced engagement and WTC in blended learning settings. Nevertheless, CALL-specific investigations remain scarce, particularly those examining how digital affordances (e.g., automated progress tracking) amplify the efficacy-engagement pathway by making value appraisals more salient.
Learning Engagement in CALL: From Self-Efficacy to WTC
Learning engagement, conceptualized as the quality of effort and attention learners invest in academic tasks (Fredricks et al., 2004), serves as a proximal predictor of L2 WTC in CALL contexts. Grounded in (Pekrun, 2006) CVT, engagement embodies value-driven behaviors—actions sustained by learners’ appraisals of task relevance and meaningfulness. Within (MacIntyre et al., 1998) WTC pyramid, engagement occupies the behavioral stratum, acting as the critical conduit through which affective-cognitive antecedents (e.g., self-efficacy) culminate in volitional communication.
Engagement in CALL environments manifests across three dimensions: behavioral (e.g., consistent participation in online forums), cognitive (e.g., deep processing of multimedia input), and affective (e.g., curiosity in exploring virtual simulations). These dimensions align with CVT’s assertion that engagement is both a product of control-value appraisals and a precursor to achievement-related outcomes like WTC (Pekrun, 2006). For instance, learners with high self-efficacy (control appraisal) are more likely to engage cognitively in AI-driven writing tasks, perceiving such activities as valuable for mastering rhetorical strategies (Barrot, 2023). This cognitive engagement, in turn, fosters linguistic confidence, thereby lowering affective barriers to spontaneous communication (Svalberg, 2009). Empirical studies corroborate this pathway Eghterafi et al. (2022) demonstrated that behavioral engagement in video-mediated peer interactions directly predicted WTC, while cognitive engagement mediated the effects of self-efficacy on communicative risk-taking.
CALL-specific affordances uniquely modulate the engagement-WTC relationship. In asynchronous environments, where learners exercise greater autonomy over participation timing, behavioral engagement often hinges on teachers’ multimodal feedback (e.g., audio-visual annotations) to sustain motivation (Wang & Xu, 2024). Conversely, in synchronous VR settings, affective engagement—driven by immersive storytelling tasks—has been shown to amplify WTC by simulating real-world communication pressures (Reinders & Wattana, 2015). Notably, technological tools that scaffold value appraisals (e.g., gamified progress trackers highlighting skill mastery) strengthen the link between engagement and WTC. For example, learners using adaptive vocabulary apps with real-time proficiency metrics reported higher engagement and subsequent WTC, as the apps rendered task value salient through personalized milestones (Shafiee Rad, 2024). These findings resonate with (MacIntyre et al., 1998) pyramid, which positions engagement as the behavioral actualization of layered antecedents (e.g., teacher support, self-efficacy), yet they also underscore CALL’s unique capacity to amplify value appraisals through interactive design.
However, prior research has inadequately addressed how engagement mediates the sequential effects of self-efficacy on WTC. CVT posits that control appraisals, with self-efficacy being a key driver, initiate engagement, which then channels value-driven effort into behavioral outcomes (Pekrun, 2006). In CALL contexts, this implies that self-efficacy fostered by teacher support must first translate into sustained engagement (e.g., persistent use of speech-recognition tools) before WTC can emerge. Preliminary evidence for this chain exists: Robson (2015) found that self-efficacy in online collaboration tasks predicted engagement, which subsequently enhanced WTC. Nevertheless, such studies remain scarce, particularly those examining how CALL’s technological mediation (e.g., anonymity in chat forums) alters the efficacy-engagement-WTC sequence. For instance, learners with low self-efficacy might still exhibit behavioral engagement in anonymous text chats due to reduced evaluation anxiety, decoupling the presumed causal chain (Carver et al., 2021). This underscores the need to interrogate CVT’s universality in digital spaces, where technological buffers may disrupt traditional appraisal-emotion dynamics.
An Integrated Framework: Bridging Macro-Structure and Micro-Process in CALL
The present study is predicated on the thesis that the (MacIntyre et al., 1998) WTC pyramid and (Pekrun, 2006) CVT offer complementary explanatory power that, when integrated, resolves a fundamental theoretical fragmentation in understanding L2 WTC. The WTC pyramid provides an indispensable structural map of the hierarchical antecedents influencing WTC, from momentary situational factors to enduring individual traits, rightly positioning teacher support as a key distal factor within the social-individual layer (Wen & Clément, 2003). However, this macro-level model, while robust in its taxonomy, leaves the proximal psychological mechanisms—the real-time cognitive-affective processes that translate a distal factor like teacher support into communicative behavior—conceptually underspecified (Khajavy et al., 2018). It delineates what factors constitute the ecosystem of WTC, but not precisely how their influence is dynamically mediated.
This is precisely the gap that control-value theory (CVT) is poised to fill. CVT provides a granular process engine, specifying the sequence through which environmental inputs (e.g., teacher scaffolding) are subjectively appraised as perceptions of control (e.g., self-efficacy) and value (e.g., task relevance), which in turn govern achievement emotions and subsequent behavioral engagement (Pekrun, 2006). For instance, in CALL contexts, technical guidance demystifies digital tools, directly boosting self-efficacy, which represents a core facet of control appraisal (Chao & Liu, 2022), while cognitive scaffolding can frame L2 tasks as meaningful challenges, enhancing value appraisals (DeKeyser, 2010). Yet, CVT often examines this process in relative isolation, without explicitly nesting it within the broader, hierarchical architecture of communicative willingness outlined by the WTC pyramid.
The proposed integration, therefore, is both logical and necessary. We bridge these frameworks by positioning teacher support—a well-defined element from the pyramid’s structure—as the critical distal environmental input in the CVT sequence. This allows us to theorize a specific psychological pathway: teacher support in CALL environments first strengthens learners’ self-efficacy (a key manifestation of CVT’s control appraisal), which in turn facilitates deeper learning engagement (manifesting as a value-driven behavior), ultimately culminating in heightened WTC. This chain (Teacher Support → Self-Efficacy → Learning Engagement → WTC) does not merely list variables; it specifies a testable, sequential mechanism that explains how a macro-level antecedent propagates its influence through micro-level appraisals (primarily self-efficacy, in our model) and behaviors to reach the apex of the pyramid. This synthesis is supported by emerging empirical logic; for example, studies in traditional classrooms have hinted at similar sequential mediations (Teng & Zhang, 2020), and research on language mindsets shows that growth-oriented beliefs—a key influencer of control appraisals—foster the very engagement that predicts WTC in online settings (Ebn-Abbasi et al., 2024). The unique affordances of CALL, however, from the potential decoupling of appraisals in asynchronous forums (Carver et al., 2021) to the motivational amplification of gamified platforms (Porter & Grippa, 2020), make testing this integrated model in a digital context a critical step for validating and refining our understanding of these theorized pathways. By unifying the WTC pyramid’s structural “what” with CVT’s mechanistic “how,” this study advances a holistic model that directly addresses theoretical fragmentation and offers a nuanced blueprint for pedagogical intervention in technology-mediated language learning.
Hypotheses
The present study proposes that the (MacIntyre et al., 1998) WTC pyramid and (Pekrun, 2006) CVT offer complementary, rather than competing, explanations for L2 communication behavior. Their integration addresses a critical theoretical gap: the WTC pyramid delineates the structural hierarchy of antecedents (from situational to stable) but offers less specificity on the proximal psychological processes that translate these antecedents into behavior. Conversely, CVT provides a precise process model of how environmental stimuli are appraised and translated into emotion and action, but typically operates at a more micro-level without always specifying how these processes are embedded within the broader, hierarchical ecosystem of WTC. We bridge this gap by positioning teacher support—a key social-individual antecedent in the WTC pyramid—as the distal environmental input in the CVT framework. We then specify the CVT’s core process sequence (control-value appraisals → achievement emotions/behaviors) using the constructs of self-efficacy (conceptualized here as a central control appraisal) and learning engagement (a value-driven behavior). This integrated model posits that teacher support in CALL environments fosters WTC by first strengthening learners’ self-efficacy (tapping into the control pathway), which in turn amplifies their engagement (the value pathway), ultimately culminating in heightened communicative willingness. Based on this synthesized framework, the following hypotheses are advanced:
First, teacher support in technology-mediated contexts directly enhances WTC by addressing situational barriers inherent to digital communication. Empirical evidence supports this direct link: For instance, Hejazi et al. (2023) demonstrated that real-time instructor scaffolding in synchronous video discussions significantly increased learners’ volitional participation, while Patrick et al. (2007) noted that positive teacher support alleviates learners’ affective distress and reduces anxiety-driven avoidance in communicative activities. These findings align with the WTC pyramid’s emphasis on social-individual antecedents (teacher support) as proximal drivers of communicative readiness. Thus, we posit:
Second, self-efficacy acts as a critical mediator, translating teacher support into learners’ perceived control over CALL tasks. CVT posits that environmental inputs (e.g., technical guidance) strengthen control appraisals, including self-efficacy, which in turn regulate engagement and outcomes (Pekrun, 2006). For example, Sun and Wang (2020)found that cognitive scaffolding in online writing tasks elevated learners’ self-efficacy, partially mediating the support-WTC relationship. Similarly, Poellhuber et al. (2008)observed that technical support in collaborative platforms enhanced efficacy beliefs, which predicted subsequent participation in peer interactions. These studies, however, isolate self-efficacy as a single mediator, neglecting its sequential interplay with engagement. We therefore hypothesize:
Third, learning engagement serves as both a direct mediator and a sequential conduit for self-efficacy. CVT’s value appraisal mechanism suggests that self-efficacy (control appraisal) must align with perceived task relevance (value appraisal) to sustain engagement (Pekrun, 2006). In CALL contexts, learners with heightened self-efficacy are more likely to invest effort in digital tasks (e.g., persistent use of gamified vocabulary apps), which reinforces communicative confidence (Papi & Khajavy, 2021). Empirical work by Pawlak et al. (2024) supports this pathway: Engaged learners in VR simulations exhibited higher WTC due to repeated practice and positive reinforcement. Yet, prior research rarely examines how engagement bridges self-efficacy and WTC. To address this gap, we propose:
These hypotheses collectively advance a chain mediation model that integrates the WTC pyramid’s hierarchical antecedents with CVT’s appraisal-emotion mechanisms (Figure 1). By testing this model, the study aims to reconcile macro-level contextual drivers (teacher support) with micro-level cognitive-motivational processes (self-efficacy → engagement), offering a holistic understanding of WTC development in technology-mediated language learning.
The chain mediation model.
Method
Sampling and Data Collection
Prior to data collection, ethical approval was obtained from the Research Ethics Committee of College of Foreign Languages at Qufu Normal University. To ensure confidentiality, all participants were assigned unique identification codes instead of personal names. Explicit written consent was secured from both participants and their guardians, with assurances that data would be accessible only to authorized research team members.
The study employed a stratified random sampling method to recruit 709 undergraduate students (mean age = 21.3 years, SD = 1.2) from three universities in Shandong Province, China, during October 2024. These institutions were selected to ensure diversity in academic disciplines, encompassing STEM (Science, Technology, Engineering, and Mathematics), arts, and humanities programs. The sample size was determined based on guidelines for structural equation modeling (SEM), which recommend a minimum ratio of 10 participants per observed variable (Kline, 2016). Participants were proportionally distributed across disciplines: 38% from STEM (e.g., computer science, engineering), 32% from humanities (e.g., literature, history), and 30% from arts (e.g., music, fine arts).
Data collection proceeded in two phases. First, questionnaires were distributed via institutional instructions, including standardized scales measuring teacher support, self-efficacy, learning engagement, and L2 communication willingness. Participants were informed of their right to withdraw voluntarily, and incomplete responses (n = 36) were excluded. A total of 673 valid responses were retained, yielding a response rate of 94.9%. Second, academic advisors provided anonymized demographic data, including participants’ majors and language proficiency levels, which were cross-referenced with survey responses. To enhance reliability, data were collected during regular class hours to minimize situational variability. All measures were administered in Mandarin to ensure comprehension, and response consistency was monitored through attention-check items embedded within the survey.
Research Instruments
Teacher Support Scale
The measurement of teacher support was conceptualized as a multidimensional construct integrating emotional, academic, and technology-facilitated dimensions. Emotional support was assessed using four items adapted from Patrick et al. (2007), capturing perceptions of teachers’ empathetic responsiveness through statements such as “Does your teacher try to help you when you are sad or upset?” and “Does your teacher respect your opinion?” Respondents rated these items on a 7-point Likert scale ranging from strongly disagree (1) to strongly agree (7). Academic support, also derived from Patrick et al. (2007), employed four items to evaluate instructional scaffolding, including “Does your teacher want you to do your best in school?” using the same 7-point scale. Technology-facilitated support, adapted from Lai (2015) framework, incorporated eight items to measure strategic integration of digital tools in language learning, such as “To expand opportunities to use the language through digital platforms.” Responses for this subscale utilized a 7-point frequency scale (never = 1 to always = 7) to enhance sensitivity to behavioral nuances. Prior to implementation, the composite scale underwent rigorous translation and back-translation protocols to ensure linguistic and conceptual equivalence in Mandarin. Confirmatory factor analysis (CFA) validated the three-factor structure, demonstrating acceptable model fit (χ2/df = 1.263, RMSEA = 0.020, CFI = 0.996).
L2 Willingness to Communicate Scale
The propensity of learners to engage in L2 communication was measured using an adapted version of the WTC scale (Peng & Woodrow, 2010). This instrument assesses spontaneous L2 interaction across formal and collaborative contexts, retaining its original two-factor structure while contextualizing items for CALL environments.
The first subscale, WTC in public speaking and formal tasks, evaluates learners’ readiness to participate in structured, performance-oriented activities. Example items include “I am willing to give a short self-introduction without notes in English to the class” and “I am willing to do a role-play standing in front of the class in English (e.g., ordering food in a restaurant).” In the original validation study, factor loadings for the items ranged from 0.426 to 0.933 (Peng & Woodrow, 2010). The second subscale, WTC in interactive and collaborative communication, measures learners’ inclination toward peer-to-peer or group-based interactions, with items such as “I am willing to ask my peer sitting next to me in English the meaning of an English word” and “I am willing to ask my group mates in English how to pronounce a word in English.” The items demonstrated factor loadings between 0.578 and 0.974 in prior research (Peng & Woodrow, 2010). The scale underwent rigorous translation and back-translation procedures to ensure semantic equivalence in Mandarin. CFA supported the two-factor structure within the current sample, yielding acceptable fit indices (χ2/df = 1.005, RMSEA = 0.001, CFI = 0.998). Responses were recorded on a 7-point Likert scale (1 = strongly unwilling to 7 = strongly willing) to capture nuanced variations in willingness. To enhance ecological validity, minor modifications were made to reflect CALL-specific contexts. For instance, references to “group mates” were replaced with “online peers” to align with digital interaction scenarios.
Academic Self-Efficacy Scale
Academic self-efficacy was measured using an adapted version of the Academic Self-Efficacy Scale developed by Patrick et al. (2007), which assesses students’ confidence in their ability to master domain-specific academic tasks. In the original study, the scale focused on mathematics; however, for this study, items were contextualized to reflect CALL environments while retaining the core construct. The scale comprises five items (e.g., “I’m certain I can master the language skills taught in my L2 class this year,”“Even if the L2 coursework is challenging, I can learn it through sustained effort”), rated on a 6-point Likert scale (1 = strongly disagree to 7 = strongly agree). Modifications to the original items included replacing subject-specific references (e.g., “math” with “L2 coursework”) to align with the study’s focus on language acquisition. Prior to administration, the scale underwent rigorous translation and back-translation procedures to ensure linguistic equivalence in Mandarin. CFA supported a unidimensional structure (χ2/df = 2.608, RMSEA = 0.049, CFI = 0.997).
Learning Engagement Scale
Learning engagement was measured using a three-dimensional scale adapted from Hiver et al. (2020), which assesses behavioral, emotional, and cognitive engagement in language learning contexts. Responses were recorded on a 7-point Likert scale (1 = strongly disagree to 7 = strongly agree), consistent with the original design by Hiver et al. (2020). To enhance ecological validity, the original scale was modified to align with CALL environments, with item wording adjusted to reflect both in-class and digital learning experiences, in which contextual phrases such as “language class” were expanded to include CALL-specific scenarios (e.g., “online language tasks”). This adaptation ensured the scale’s relevance to technology-enhanced language learning environments while maintaining its psychometric rigor. Behavioral engagement was evaluated through learners’ active participation and effort in language tasks. A single item retained from the original scale, “I put effort into learning in my language class,” captured this dimension. Emotional engagement, reflecting learners’ affective connection to language activities, was measured using the adapted item “I look forward to my language class.” Cognitive engagement, which emphasizes strategic and reflective investment in learning, incorporated two adapted items: “I go through the work for my language class carefully and make sure that it’s done right” and “In my language class, I think about different ways to solve a problem.” The scale underwent rigorous translation and back-translation procedures to ensure linguistic equivalence in Mandarin. CFA supported the structure of the questionnaire (χ2/df = 1.258, RMSEA = 0.001, CFI = 0.991).
Statistical Analysis
Analytical procedures were conducted in three sequential stages using Mplus 8.3 and SPSS 24.0. Preliminary analyses focused on data screening and assumption checks. Descriptive statistics (means, standard deviations) and distributional properties (skewness, kurtosis) were computed in SPSS to assess normality. All variables exhibited acceptable skewness (<|1.5|) and kurtosis (<|2.5|), aligning with structural equation modeling (SEM) requirements (Byrne, 2016).
The measurement model was rigorously evaluated via CFA in Mplus. Internal consistency was assessed using Cronbach’s α coefficients, with values exceeding .80 indicating strong reliability (Nunnally & Bernstein, 1994). Convergent validity was confirmed through composite reliability (CR >.70) and average variance extracted (AVE >0.50) (Fornell & Larcker, 1981). Discriminant validity was verified by ensuring the square root of each construct’s AVE exceeded its correlations with other latent variables.
The structural model was tested using robust maximum likelihood estimation (MLR) in Mplus to account for residual non-normality. Model fit was evaluated using multiple indices: χ2/df < 3.0, root mean square error of approximation (RMSEA) <0.06, standardized root mean square residual (SRMR) <0.08, comparative fit index (CFI) >0.95, and Tucker-Lewis index (TLI) >0.95 (Hu & Bentler, 1998). Chain mediation hypotheses were examined via a bias-corrected bootstrapping approach with 5,000 resamples, where 95% confidence intervals (CIs) excluding zero denoted statistical significance (Preacher & Hayes, 2008).
Common method bias was assessed using both procedural and statistical approaches. Procedurally, respondent anonymity was ensured and scale items were counterbalanced. Statistically, we first employed Harman’s single-factor test (Podsakoff et al., 2003). To provide a more robust check, we also conducted an unmeasured latent method construct (ULMC) analysis within a structural equation modeling framework (Podsakoff et al., 2012) The improvement in model fit was evaluated using the criterion that a change in comparative fit index (ΔCFI) of less than 0.01 indicates no meaningful improvement (Cheung & Rensvold, 2002).
Results
Common Method Variance Assessment
To mitigate concerns regarding potential common method bias inherent in cross-sectional self-report designs, we employed complementary statistical techniques. First, a Harman single-factor analysis was conducted via SPSS 24.0 following established psychometric protocols. The unrotated principal component analysis yielded a first factor accounting for 34.16% of total variance—substantially below the 40% threshold conventionally adopted as evidence of substantial common method effects (Fuller et al., 2016; Podsakoff et al., 2003). Second, to address the limitations of this standalone test, we performed a more rigorous ULMC analysis (Podsakoff et al., 2012). This method tests whether adding an orthogonal common method factor to the measurement model significantly improves model fit or alters the substantive relationships. The key finding was that the model fit did not significantly improve upon adding the method factor (ΔCFI = .001, ΔTLI = .001), indicating that common method variance did not substantially inflate the model’s relationships. Together, these psychometric evaluations demonstrate that common method variance is unlikely to have artificially inflated the observed relationships or compromised the validity of the model interpretations.
Measurement Model
The psychometric evaluation of the measurement model provided robust evidence for the reliability and validity of all latent constructs. As summarized in Table 1, both Cronbach’s alpha (α) and composite reliability (CR) coefficients surpassed the conventional threshold of .70, indicating strong internal consistency. Teacher Support (TS) and Willingness to Communicate (WTC) yielded particularly high coefficients (TS: α = .949, CR = .949; WTC: α = .912, CR = .913), while Self-Efficacy (SE) and Learning Engagement (LE) also demonstrated robust reliability (SE: α = .896, CR = .899; LE: α = .837, CR = .838). While such high alpha values attest to exceptional response homogeneity, it is methodologically prudent to address the potential concern that they might reflect item redundancy rather than optimal psychometric quality. This possibility was examined from both conceptual and empirical standpoints. Conceptually, the TS and WTC scales are well-established instruments whose validity and unidimensional structure have been consistently validated in prior L2 research (e.g., Lai, 2015; Patrick et al., 2007; Peng & Woodrow, 2010); their high reliability in the present context is thus consistent with their documented performance. Empirically, supplemental analyses of inter-item correlations and corrected item-total correlations were conducted for these two scales. For the TS scale, inter-item correlations ranged from .41 to .76 (M = 0.51), with corrected item-total correlations between .72 and .87. For the WTC scale, inter-item correlations ranged from .44 to .60 (M = 0.50), and corrected item-total correlations fell between .063 and .79. These patterns confirm that items are strongly interrelated in a manner congruent with unidimensional constructs, yet none of the correlations approach the extreme upper bound (e.g., >.90) that would signal problematic redundancy. Accordingly, the elevated alpha coefficients are interpreted as evidence of strong internal consistency within theoretically coherent constructs, rather than as an indication of measurement overlap that would compromise validity.
Evaluation of Reliability and Validity.
Note. SC = standardized coefficients.
Convergent validity was further supported by standardized factor loadings ranging from 0.638 to 0.878, each exceeding the conservative 0.60 cutoff. The average variance extracted (AVE) values for TS, SE, and LE met the 0.50 benchmark (0.543, 0.641, and 0.565, respectively). The AVE for WTC (0.513) also exceeded this threshold, albeit marginally. This borderline result is acknowledged and may be attributable to the multifaceted nature of the WTC construct, where items often tap into distinct yet interrelated communicative contexts, leading to slightly more dispersed indicator covariance. Nevertheless, the strong factor loadings and composite reliability (CR = .913) affirm the overall convergent validity of the WTC measure within the model. Discriminant validity was evaluated using the Fornell and Larcker’s (1981) criterion. As presented in Table 2, the square root of the AVE for each construct (diagonal values) exceeded its correlations with all other constructs (off-diagonal values). For instance, the highest correlation observed was between LE and WTC (r = .434), which remains substantially lower than the square root of AVE for either construct (0.752 and 0.716, respectively). This confirms that each latent variable shares more variance with its own indicators than with any other variable in the model, thereby establishing distinctiveness. Moreover, all inter-construct correlations were below the .70 threshold, alleviating concerns regarding multicollinearity. Collectively, these results affirm the soundness of the measurement model and justify proceeding to the examination of structural relationships.
The Test for Discriminant Validity of Potential Variables.
Note. The square root of the AVE of four latent constructs is given in the diagonal, and the correlation coefficient is given on the below diagonal.
The TS construct exhibited the lowest correlations with other variables (ranging from 0.276 to 0.373), suggesting its role as a distal contextual factor that operates through mediating mechanisms rather than direct associations. This observation resonates with hierarchical models of educational support, where teacher-student interactions often influence outcomes via intermediate psychological processes (Wentzel, 2022). The stringent adherence to discriminant validity criteria—coupled with previously established convergent validity (CR > .80, AVE > 0.50 across constructs; (Hair et al., 2018)—provides empirical justification for proceeding to structural path analysis without measurement model misspecification risks.
Structural Model
The structural equation model demonstrated an excellent fit to the observed data. As indicated in Table 3, the fit indices substantially surpassed conventional benchmarks: χ2/df = 1.036, RMSEA = 0.007 (90% CI [0.002, 0.012]), SRMR = 0.026, CFI = 0.998, and TLI = 0.998 (Hu & Bentler, 1999; Kline, 2016; West et al., 2012). We acknowledge that these indices approach the theoretical maximum, a result that, while indicating a high degree of model-data consistency, warrants deliberate consideration to preclude concerns regarding overfitting or an excessively parameterized model relative to the sample size (N = 673). Several substantive and methodological factors support the validity of this excellent fit. First, the hypothesized model is parsimonious, specifying only the theoretically essential structural paths (i.e., the proposed chain mediation) without extraneous or post-hoc modifications. Second, this structural model is built upon a strong and well-defined measurement model, characterized by high-reliability indicators and clear latent constructs, which provides a stable foundation for parameter estimation. Third, the sample size is more than adequate for the model’s complexity. With 35 observed indicators and a sample of 673 participants, the ratio of observations to freely estimated parameters is favorable and well above common recommendations for stability. To further address potential overfitting, a supplementary cross-validation procedure was employed. The sample was randomly split into two independent subsamples. The model was calibrated on the first half and then tested on the validation half. The excellent fit indices and the pattern of significant structural paths were consistently replicated across both subsamples (e.g., CFI > 0.98, RMSEA < 0.05 in the validation sample), demonstrating the model's stability and generalizability beyond the specific calibration sample. As such, the exemplary model fit is interpreted not as a statistical artifact but as robust evidence that the proposed integrative framework—specifying linkages among teacher support, self-efficacy, learning engagement, and WTC through a sequential mechanism—provides a highly accurate representation of the relationships within our data. This sound fit justifies the interpretation of the subsequent structural parameter estimates.
Goodness-of-Fit Indices for the Structural Mode.
Path analysis results (Table 4) revealed statistically significant associations across all hypothesized relationships. TS exhibited a direct association with WTC of β = .221 (p < .001), which can be interpreted as a small-to-moderate effect size according to conventional benchmarks in psychological research (β ≈ .10, .30, and .50 representing small, moderate, and large effects, respectively (Cohen, 1992). Simultaneously, TS showed significant indirect associations via SE and learning engagement (LE) as parallel mediators. The TS → SE → WTC pathway (β = .291 → .254) and TS → LE → WTC pathway (β = .186 → .280) demonstrated comparable indirect effects (statistically), which is consistent with social cognitive theory postulations about environmental influences on behavioral outcomes (Bandura, 1997). Notably, SE was significantly associated with WTC both as part of a mediation chain and independently (β = .254, p < .001), suggesting its potential dual role as a psychological mechanism and autonomous motivator of communicative behaviors (Pintrich & Groot, 1990).
The Test Results of Path Relationship.
Note.***p < .001.
The reciprocal SE ↔ LE association (β = .309, p < .001) is consistent with academic self-regulation frameworks positing dynamic interplay between cognitive and behavioral engagement components (Zimmerman, 2000). Learning engagement manifested the strongest total effect on WTC (β = .280), exceeding the magnitude of direct teacher support effects, which underscores the criticality of sustained cognitive investment in fostering interactive learning dispositions (Fredricks et al., 2004). All critical ratios (CR >4.0) exceeded the 3.29 threshold for p < .001 significance, with standardized estimates demonstrating practical significance beyond statistical thresholds (Cohen, 1992).
The structural model accounted for 63.2% of the variance in WTC (R2 = .632). This value far exceeds conventional benchmarks for explanatory power in social science research (e.g., R2 > .26 indicating a large effect size; Cohen, 1992) and the common thresholds applied in educational psychology (Hair et al., 2018). This robust predictive capacity is consistent with the theoretical integration of social-contextual factors (TS), psychological resources (SE), and behavioral states (LE) in modeling communication readiness—a finding consistent with recent advances in motivation-engagement interfaces (Martin & Dowson, 2009). The absence of specification errors (modification indices < 5.0) and acceptable multicollinearity levels (VIF < 2.1) further validate the model’s parsimony and theoretical integrity.
Analyses of the Mediating Effect
The structural model revealed a complex interplay of direct and indirect associations through which teacher support is linked to willingness to communicate. The decomposition of effects, as presented in Table 5, clarifies these mechanisms. The total effect of teacher support on WTC was significant (unstandardized estimate = 0.449, p < .001). This total effect was composed of a significant direct effect (estimate = 0.267, p < .001) and a significant total indirect effect mediated through self-efficacy and learning engagement (estimate = Thus, the direct pathway accounted for 59.4% (0.267/0.449) of the total effect, while the combined indirect pathways accounted for the remaining 40.6%. The significant direct effect aligns with social exchange theory, where supportive instructor behaviors can foster communicative willingness by establishing relational trust and a perception of safety (Cropanzano & Mitchell, 2005). The substantial indirect effect operated through sequential cognitive-motivational mechanisms. Self-efficacy served as the predominant mediator within the indirect pathway, accounting for approximately 48.8% of the total indirect effect (P2 in Table 5). This resonates with social cognitive theory, positing that teacher support enhances agentic beliefs, which in turn drive goal-directed communicative behaviors (Bandura, 1997). The specific sequential mediation path (TS → SE → LE → WTC) was also significant, explaining 16.6% of the total indirect effect, underscoring the hierarchical nature of motivation in which confidence (SE) fosters sustained engagement (LE), ultimately promoting WTC.
Total, Direct, and Indirect Effects of the Theoretical Model.
Note.***p < .001.
The significant indirect effect through learning engagement alone highlights the behavioral dimension linked to communicative readiness, where sustained academic involvement co-occurs with interactional dispositions. This finding complements recent work on engagement-behavior interfaces, demonstrating that cognitive immersion in learning tasks creates natural opportunities for communicative exchanges (Fredricks et al., 2004). The moderate proportion of distal mediation (17%) suggests teacher support is statistically linked to WTC through layered cognitive-motivational processes rather than singular pathways, reflecting the multidimensional nature of educational influence (Wentzel, 2022).
Bootstrap confidence intervals for all effects estimates excluded zero, confirming result stability across estimation methods. The total indirect effect’s magnitude (β = .182) meets empirical benchmarks for meaningful mediation in cross-sectional behavioral research (Shrout & Bolger, 2002), while the large total effect (β = .449) indicates the model’s strong explanatory power. The residual direct effect implies additional correlates or unmeasured mediators, potentially including emotional regulation or classroom climate factors, warranting future investigation. Methodologically, the integration of effect-size metrics and proportional mediation indices adheres to contemporary standards for complex model reporting (Hayes, 2017), ensuring both statistical rigor and theoretical interpretability.
Discussion
The present study corroborates the foundational premise that teacher support in CALL environments is significantly associated with learners’ WTC in a second language. This finding aligns with (MacIntyre et al., 1998) WTC pyramid, which positions teacher support as a critical social-individual antecedent within the hierarchical ecosystem of communicative readiness. Specifically, the direct pathway resonates with empirical evidence from synchronous CALL contexts, where real-time instructor interventions—such as scaffolded feedback during video-mediated discussions (Hejazi et al., 2023) or affective reassurance in virtual reality (VR) simulations (Reinders & Wattana, 2015)—have been correlated with reduced situational anxiety and amplify learners’ volitional readiness to engage. The technological immediacy inherent in these modalities appears to enhance the observed association with teacher support by providing learners with instantaneous safety nets, thereby aligning with (Pekrun, 2006) CVT: real-time scaffolding is linked to learners’control appraisals (perceived ability to manage task demands) while simultaneously being related to the affirmation of communicative risk-taking. However, this direct effect stands in contrast to studies conducted in asynchronous CALL environments, where delayed feedback loops and reduced social presence attenuate teacher support’s impact. For instance, Carver et al. (2021) documented negligible direct effects of instructor guidance on WTC in text-based discussion forums, suggesting that technological mediation moderates the support-WTC relationship. This divergence underscores a critical nuance in MacIntyre et al. (1998) model: while teacher support operates universally as a social-individual antecedent, its behavioral manifestation depends on the affordance-aligned design of CALL tasks. In East Asian EFL contexts, where face-saving norms heighten sensitivity to evaluative threats (Wen & Clément, 2003), synchronous teacher support may play an outsized role in mitigating situational anxiety—a mechanism less salient in individualistic educational cultures. Thus, the direct pathway not only validates the WTC pyramid’s structural hierarchy but also illuminates how CALL’s technological and cultural contingencies modulate teacher support’s efficacy.
The significant mediating role of self-efficacy in the relationship between teacher support and L2 WTC supports the theoretical proposition that learners’ control appraisals—central to (Pekrun, 2006) CVT—serve as a psychological conduit for translating instructional interventions into communicative behaviors. This finding aligns with (Bandura, 1997) assertion that efficacy beliefs are cultivated through mastery experiences and vicarious learning, both of which are systematically orchestrated in CALL environments through teacher scaffolding. For instance, technical guidance in navigating speech recognition software (Silviya Nancy et al., 2019) or metacognitive prompts during AI-mediated writing tasks (Hwang et al., 2023) provides learners with structured opportunities to attribute success to effort rather than innate ability, which may, in turn, be associated with reinforced self-efficacy. Such observed patterns are consistent with CVT’s emphasis on environmental inputs being related to control appraisals, which regulate achievement emotions and subsequent engagement. Notably, the study extends prior empirical work by isolating technology-specific efficacy as a distinct mediator. While Klassen and Usher (2010) demonstrated the general mediating role of self-efficacy in traditional classrooms, the current findings suggest that CALL environments may amplify this pathway through tool-mediated mastery experiences. For example, learners who received scaffolded support in virtual reality (VR) simulations (e.g., guided practice sessions for avatar-based interactions) developed heightened confidence not only in L2 communication but also in managing digital interfaces—a dual efficacy dimension absent in non-technological settings. This nuance challenges (Sun & Wang, 2020) narrow conceptualization of self-efficacy as task-specific competence, suggesting that technological literacy itself becomes a critical efficacy source in CALL contexts. However, the mediation effect was partial rather than full, implying that self-efficacy alone cannot exhaustively explain how teacher support fosters WTC. This aligns with (MacIntyre et al., 1998) WTC pyramid, which posits that affective-cognitive factors (e.g., self-efficacy) must interact with behavioral engagement to actualize communicative readiness. CVT further clarifies this partial mediation: while control appraisals, for which self-efficacy is a primary indicator, initiate the motivational sequence, value appraisals (task relevance) are required to sustain engagement and culminate in WTC. Thus, the findings underscore the necessity of examining self-efficacy within a chain mediation framework rather than as an isolated mechanism—a theoretical advancement that addresses fragmented interpretations in earlier studies.
The mediating role of learning engagement in the relationship between teacher support and L2 WTC underscores the centrality of value-driven behaviors in (Pekrun, 2006) CVT. The findings reveal that teacher support in CALL environments—particularly cognitive scaffolding and affective reinforcement—is associated with learners’ value appraisals, which in turn are linked to sustained engagement. For instance, learners exposed to gamified vocabulary tasks with progress metrics (Di Zou et al., 2021) reported higher behavioral engagement (e.g., frequent task completion) and cognitive engagement (e.g., deliberate practice of semantic networks), which were directly associated with WTC. This is consistent with (Fredricks et al., 2004) tripartite model, where value appraisals are linked to observable effort. The results partially reconcile conflicting evidence in prior literature. While Svalberg (2009) questioned engagement’s predictive power in traditional classrooms, the current study demonstrates its heightened salience in CALL contexts. Technological tools such as anonymized chat forums and VR simulations may reduce evaluation anxiety (Reinders & Wattana, 2015), allowing engagement to occur experimentally—a dynamic particularly potent in collectivist cultures (Wen & Clément, 2003). This may explain why engagement showed a robust mediating link in the present study. Furthermore, CVT’s emphasis on task value helps explain why certain CALL interventions succeed: When teachers frame AI-generated writing feedback as a tool for authentic communication (e.g., drafting emails for hypothetical internships), learners perceive engagement as instrumentally valuable, which is correlated with effort to WTC (Shafiee Rad, 2024). However, the mediation effect was contingent on technological design. In asynchronous discussion boards lacking immediate reinforcement, engagement often was observed to decay despite initial teacher support, echoing (Carver et al., 2021) observations of motivation attrition in self-paced modules. This suggests that engagement’s mediating role may depend on CALL tools’ capacity to sustain value appraisals through real-time feedback or adaptive difficulty—an insight absent in (MacIntyre et al., 1998) original pyramid. The WTC framework thus could benefit from augmentation to account for technology’s dual role as both conduit for and a potential disruptor of the associations along engagement pathways.
The identified sequential pathway—linking teacher support to WTC through self-efficacy and engagement in tandem—provides empirical support for the theoretical integration of (MacIntyre et al., 1998) WTC pyramid and (Pekrun, 2006) CVT. This chain mediation underscores that communicative readiness in CALL environments is associated with an orchestrated interplay of control appraisals and value-driven behaviors. For instance, learners who received technical scaffolding in virtual exchange programs (Wawrosz & Jurásek, 2021) first reported higher confidence in navigating cross-cultural dialogues, which was subsequently correlated with persistent participation in tasks, ultimately correlating with heightened WTC. This pattern is consistent with CVT’s proposition that control appraisals may theoretically initiate motivational sequences, which are then sustained by engagement through value reinforcement—a dynamic potentially amplified by CALL’s interactive affordances. The findings extend prior research in two critical ways. First, they clarify ambiguities in (Teng & Zhang, 2020) classroom-based model by demonstrating a sequential relationship between self-efficacy and engagement in CALL context. The current study indicates that in CALL contexts, self-efficacy may serve as a gateway to engagement: Learners who mastered VR simulation tools (control appraisal) were more likely to perceive gamified tasks as intrinsically valuable (value appraisal), thereby reporting deeper cognitive effort—a recursive loop absent in non-digital settings. Second, the results refine Alrabai’s (2022) assertion by suggesting that engagement’s role may depend on prior efficacy gains. This distinction highlights CALL’s potential to make psychological processes more observable. Technological design may further moderate this chain. In platforms with real-time feedback, the link between self-efficacy and engagement may be stronger due to immediate reinforcement (Wei, 2023). Conversely, in self-paced modules lacking interactivity, the associations in the chain weaken, as delayed feedback is linked to reduced appraisals’ salience—a phenomenon echoing (Carver et al., 2021) observations of motivation decay in asynchronous forums. This suggests that (MacIntyre et al., 1998) pyramid, while structurally sound, could be augmented to account for technology’s potential role in strengthening or weakening the observed efficacy-engagement-WTC associations.
Taken together, our findings—demonstrating a significant sequential mediation from teacher support to WTC via self-efficacy and engagement—provide a robust empirical basis for integrating the WTC pyramid and CVT in CALL contexts. The synthesis of (MacIntyre et al., 1998) WTC pyramid and (Pekrun, 2006) CVT within CALL contexts, supported by our findings, invites a theoretical proposition. Building directly on the confirmed chain mediation, we propose a technology-mediated control-value (TMCV) framework. This framework posits that digital affordances act as constitutive elements within the appraisal-behavior system, potentially transforming traditional sequential pathways into more dynamic, reciprocal relations (Larsen-Freeman, 2019) and altering the interaction between control-value appraisals and motivated behavior. For example, the immediacy and granularity of feedback in AI-driven systems can simultaneously reinforce control appraisals (e.g., self-efficacy) and heighten value appraisals (Hwang et al., 2023), while immersive or anonymous digital environments may facilitate identity experimentation, decoupling efficacy beliefs from social risk. Crucially, this TMCV framework introduces technological salience—the degree to which a tool renders app. We present this not as a validated model but as a generative theoretical proposition derived from our study. Its primary function is to consolidate our empirical results into a coherent lens and to chart a clear agenda for future research, specifically calling for direct tests of how specific technological affordances moderate the psychological pathways established here.
Conclusion
This study, situated within a Chinese EFL learning environment, substantiates a sequential mediation pathway, revealing that teacher support fosters L2 WTC through the chain effects of self-efficacy and learning engagement. The integration of (MacIntyre et al., 1998) WTC pyramid and (Pekrun, 2006) CVT elucidates how, in this context, technology-mediated environments amplify control-value appraisals, transforming instructional scaffolding into communicative readiness. Specifically, CALL tools such as AI-driven feedback systems and VR simulations accelerate this pathway by rendering self-efficacy gains visible (e.g., real-time proficiency metrics) and engagement value-laden (e.g., gamified task relevance), thereby collapsing traditional temporal gaps between psychological appraisals and behavioral outcomes.
Theoretically, these findings necessitate a paradigm shift toward recognizing technology as an architectural force in L2 motivation models. The proposed TMCV Model extends traditional frameworks by positioning digital affordances as co-constructors of appraisal-emotion dynamics, where tools like adaptive dashboards (Hwang et al., 2023) and anonymized chat interfaces (Reinders & Wattana, 2015) reconfigure agency hierarchies between learners, teachers, and technological systems. Practically, the study underscores the urgency of training educators in multimodal scaffolding—strategies that interweave technical guidance (e.g., VR navigation tutorials) with affective reinforcement (e.g., growth-oriented feedback on AI writing tools)—to maximize control-value synergies. CALL designers, conversely, must prioritize features that sustain appraisal salience, such as dynamic progress visualizations and context-aware difficulty adjustments, to prevent engagement attrition in self-paced modules.
Some limitations constrain the generalizability and contextual interpretation of these insights. First, and most critically, the cultural specificity of the sample—drawn from East Asian EFL learners in contexts that prioritize collectivist norms, face-saving, and elevated teacher authority (Wen & Clément, 2003)—fundamentally shapes the findings. The observed strong direct and mediated pathways involving teacher support may be partially amplified by these sociocultural conditions, where instructor guidance is both highly valued and culturally expected. This limits direct comparability with educational settings in individualistic cultures, where learner autonomy, peer collaboration, and self-regulated learning are often more central to motivation. Future research must test the proposed framework in diverse contexts (e.g., North American or European ESL settings) to disentangle universal psychological mechanisms from culturally contingent expressions. Specifically, studies could examine whether in individualistic or low-power-distance cultures, the mediation pathway via self-efficacy and peer engagement becomes more pronounced, while the direct effect of teacher support attenuates. Second, the study’s exclusive reliance on self-report surveys, while providing valuable quantitative evidence, limits the understanding of how teacher support is pedagogically enacted and subjectively experienced within authentic CALL practices. The mechanisms, nuanced teacher behaviors, contextual adaptations, and learner interpretations that give rise to the observed statistical pathways remain unexplored. To address this, future research should adopt mixed-methods designs. For instance, integrating the classroom observations could reveal specific supportive teacher behaviors in digital settings, while in-depth interviews or reflective journals could illuminate learners’ lived experiences and the meaning they ascribe to different forms of support. Third, the cross-sectional nature of the data constitutes a constraint. While the model fit and significant paths are consistent with the hypothesized causal sequence, this design cannot empirically establish temporal precedence or causality. The relationships observed could, in theory. Therefore, longitudinal or experimental replications are not merely beneficial but imperative to validate the proposed chain mediation as a causal pathway. Future studies should track learners over time, especially across key phases of CALL tool adoption and integration (e.g., from initial introduction to habitual use), to robustly test whether changes in teacher support predict subsequent changes in self-efficacy, which in turn foster engagement and ultimately enhance WTC. Such designs are essential to move from a model of plausible associations to one of demonstrated temporal dynamics.
Footnotes
Ethical Considerations
This manuscript is not under review elsewhere and the results have not been published previously or accepted for publication. This manuscript has been seen and approved by all authors. All methods were performed in compliance with the Helsinki Declaration. The questionnaire and methodology for this study were approved by the research ethics committee of College of Foreign Languages at Qufu Normal University before data collection.
Consent to Participate
Informed consent was obtained from all participants included in the study. And their parents provided their written informed consent to participate in this study.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Research Project of Humanities and Social Sciences of the Ministry of Education (Grant No. 24YJC740084); in part by the Higher Education Youth Innovation Team Project of Shandong Province (Grant No. 2023RW050); and by the International Chinese Language Education Research Program (Grant No. 23YH82C).
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Data Availability Statement
The datasets used and analyzed during the current study are available from the corresponding author on reasonable request.