Exploring the Impact of VR Scaffolding on EFL Teaching and Learning: Anxiety Reduction, Perceptions, and Influencing Factors

Abstract

This study examined the use of virtual reality (VR) scaffolding in English as a Foreign Language (EFL) instruction, focusing on its effects on speaking anxiety and learner perceptions and the need for tailored assessment methods. The study involved 34 Taiwanese university medical students and utilized quantitative and qualitative questionnaires. The quantitative results indicated a significant reduction in speaking anxiety and positive perceptions of VR-assisted learning. Qualitative findings revealed that students experienced dual anxieties related to language and technology during VR learning and nervousness during performance evaluations in a VR setting. This study highlights the importance of creating interactive scaffolding that considers individual learner differences and supports personalized learning experiences. It also underscores the necessity of adopting assessment strategies that align with VR environments’ unique, immersive nature in language instruction. Our findings contribute to the growing body of research on VR applications in language learning, offering valuable insights for educators and researchers aiming to leverage this innovative technology in the EFL context.

1. Introduction

Virtual reality (VR) is a well-established technology that has the potential to revolutionize traditional educational methods by providing immersive and interactive learning experiences. This study aimed to investigate the impact of VR learning interventions on learners’ English-speaking anxiety; explore the relationships among learners’ perceptions, anxiety levels, and experiences with VR in language learning; and examine how learners’ backgrounds influence the benefits of VR-based language learning. Students underwent a series of pre- and post-tests to assess the effectiveness of VR interventions. The pre-tests measured their initial English-speaking anxiety levels, English proficiency background, and perceptions of VR as a learning tool. The post-tests evaluated changes in anxiety levels and perceptions after participating in the VR learning sessions. Previous studies have highlighted the potential of VR in educational settings. For instance, Makransky and Lilleholt [1] found that VR enhanced student engagement and motivation. Similarly, Chen et al. [2] demonstrated that VR reduced anxiety and improved language learning outcomes among English as a Foreign Language (EFL) learners. These findings underscore the importance of further exploring the effects of VR on language learning and its potential to alleviate speaking anxiety.

1.1. Research Objectives

• To explore the relationships between learners’ perceptions, anxiety levels, and experiences when using VR for language learning.
• To determine the impact of VR scaffolding interventions on English language learning and English-speaking anxiety.
• To investigate how learners’ backgrounds (e.g., VR experience and English proficiency) affect the benefits of VR-assisted language learning and how the effectiveness of VR instruction is related to reliance on scaffolding (usage frequency and process).

1.2. Hypothesis and Framework

We hypothesized that VR scaffolding interventions would reduce English-speaking anxiety, positively correlate with learners’ perceptions and experiences, and provide greater benefits to learners with prior VR experiences and higher English proficiency. This study aimed to integrate VR with scaffolding theory and differentiated instruction, positing that students could engage in self-directed learning through VR even after the scaffolding was removed. This study also sought to elucidate student learning processes in virtual and physical classrooms by utilizing the Plan-Do-Check-Act cycle quality management model and learning theory frameworks. Continuous adjustments to personalized learning paths were made based on each student’s progress, ensuring that learning occurred within the zone of proximal development, and was supported by updated scaffolding. The timely removal of scaffolding was implemented to enhance student learning outcomes. The implementation framework is illustrated in Figure 1.

1.3. Literature Review

1.3.1. Scaffolding Theory

Scaffolding theory, also known as scaffolded instruction, encompasses a variety of definitions, and is challenging to pinpoint precisely because of its application across diverse teaching methodologies and integrative learning approaches. Puntambekar and Kolodner [3] noted that the complexity of classroom environments and varying circumstances mean that a single scaffold might not necessarily guarantee improved learning outcomes. Wood et al. [4] initially introduced the concept of scaffolding, and emphasized that support educators should facilitate students’ acquisition of new skills and concepts. In the context of problem-solving, scaffolding refers to a process that enables novice learners to tackle problems or perform tasks. Educators create “temporary, adjustable, and supportive” structures tailored to students’ needs, which are gradually removed as learners become more proficient [5]. Discussions of scaffolding theory often refer to Vygotsky’s concept of the zone of proximal development, which describes the gap between a learner’s current abilities and their potential development, achievable through guidance from more knowledgeable peers or experts [6].

Furthermore, examining the operational design of scaffolding, Belland [7] identified two primary functions: aiding students in enhancing their skills and engaging them in the execution of target skills. Common types of scaffolding include one-on-one, peer-assisted, and computer-assisted scaffolding. Effective scaffolding should possess three distinct characteristics: contingency, intersubjectivity, and the transfer of responsibility. Contingency refers to the continuous diagnosis of students’ learning stages and the provision of appropriate scaffolding support [7,8]. Intersubjectivity involves interactive collaboration between educators and students, emphasizing shared learning rather than unidirectional transmission. The transfer of responsibility denotes the gradual removal of scaffolding by the educator as the student becomes more proficient, ultimately leading the student to take full responsibility for their own learning.

For effective scaffolding designs, it is essential to consider the type of scaffolding VR that is most suitable for the objectives of instructional research. Hannafin et al. [9] categorized scaffolding into four functional types: conceptual scaffolding, which helps students identify fundamental core themes and relevant background knowledge; metacognitive scaffolding, which helps students monitor and reflect on their learning processes; strategic scaffolding, which offers alternative approaches to handling learning tasks; and procedural scaffolding, which assists students in using resources and tools for learning, such as providing directions for system functionalities and features. This implies that educators must consider numerous influencing factors when designing scaffolding, including the difficulty of the subject matter, expected learning outcomes, and styles and types of learners. Additionally, careful consideration must be given to the timing of the implementation and removal of scaffolding, as these factors could significantly affect the scaffolding’s success.

1.3.2. Scaffolding in Language Education

The application of scaffolding theory to language education has a long history of providing significant benefits to learners. The use of VR as a scaffolding tool presents a promising opportunity to integrate VR into real-world learning environments. This Section reflects on past experiences and compares them with relevant literature to determine their alignment with classroom observations. It is essential to discuss the scaffolding providers and recipients. As previously mentioned, there are three types of scaffolding: one-on-one, peer-assisted, and computer-assisted. In various teaching experiences, students’ preferences for these types have been observed to vary according to their English proficiency levels and learning styles. Some students prefer visual aids or auditory methods, and some learn best through hands-on activities, whereas others favor peer assistance or are more socially reserved. In alignment with the literature, I have encountered similar cases in which my students have requested to avoid group work. They express concerns such as, “I usually complete the tasks while others are still thinking, which is unfair to me”. This observation suggests that students’ pre-university socialized learning habits and perceptions of the role of knowledge providers might influence their preferences. Therefore, it is crucial to consider the provision of differentiated support tailored to the needs of all students.

Scaffolding is commonly implemented in classrooms in three ways: as an expert, reciprocally, and self-scaffolding. Expert scaffolding occurs when an expert assists learners in the learning process. In reciprocal scaffolding, peers help each other based on their strengths. Self-scaffolding refers to an individual’s ability to apply existing knowledge to solve new problems [10]. These forms align with observed classroom experiences.

In a typical classroom setting, one-on-one scaffolding is provided through consultation hours between the instructor and the students, typically held a week before the submission of the final report, to assess each student’s overall learning progress and offer personalized advice. Peer-assisted scaffolding occurs almost weekly, allowing students to leverage their strengths to assist each other. Additionally, over the past five years, computer-assisted scaffolding has significantly increased, and it has been integrated into the classroom without time constraints whenever different student needs arise. Tools such as artificial intelligence (AI) pronunciation correction apps have been particularly well received by students for their convenience and utility, functioning as English tutors accessible via mobile devices. Previous studies have shown that virtual gaming environments could effectively enhance students’ spelling and vocabulary. Utilizing VR as a scaffolding tool not only makes learning more practical and accessible but also promotes student-centered learning, encouraging students to interact autonomously with the virtual world.

Based on its sources and impact, scaffolding could be categorized into three types: humans, artifacts, and a combination of humans and artifacts. Instructors and peers provided human scaffolding. Regardless of peer assistance, students often prefer corrective feedback from instructors to peer evaluations, especially when the task requires a specific correct answer or when students have higher English proficiency. However, for open-ended reflective tasks or activities that are not tied to semester grades, students are generally less concerned with sources of scaffolding. In language education, scaffolding is effective in meeting students’ interests and needs, providing authentic language information, offering clarification, highlighting errors, and delivering corrective feedback [11]. To enhance students’ trust in and acceptance of scaffolding, it is crucial to design scaffolding source attributes based on the target audience.

Dynamic assessments within scaffolding could also facilitate iterative adjustment of teaching strategies. Artifact scaffolding, which provides immediate corrective feedback, is sometimes welcomed by students, particularly in large classroom settings. For instance, AI-based grammar correction tools help students to continue to produce written work despite spelling difficulties, thereby building confidence without undermining their learning motivation. In distance learning classes, students with poor listening skills are encouraged to use automatic subtitle software, which enhances their comprehension of lecture content through differential support. Doo et al. [12] found that computer-assisted scaffolding is relatively more effective than human scaffolding. The students also responded positively to the recommended use of ElsaSpeak, an English-speaking practice application, noting that speaking with a machine reduced their anxiety. The effectiveness of VR in supporting second language acquisition is influenced by learners’ vocabulary in their native language and is particularly beneficial for students with lower language proficiency [13]. Therefore, VR could be effectively utilized as artifact scaffolding to assist less proficient or confident students by familiarizing them with prerequisite knowledge before class. Complementing this with human scaffolding during class could further alleviate anxiety associated with immediate output and reduce disparities among peers.

1.3.3. Scaffolding in VR and Creativity

The existing literature on VR use in education suggests that the design of VR scaffolding should be diverse and differentiated to address the varying learning backgrounds and needs of students and to enhance its effectiveness. The following key points were synthesized from the literature:

• Diversifying Types of Scaffolding: Emphasizing the four language skills (listening, speaking, reading, and writing) equally and tailoring scaffolding to different English proficiency levels and learning styles. Incorporating multimedia elements such as images, videos, and step-by-step demonstrations could facilitate active learner engagement and a sense of accomplishment in achieving educational goals.
• Enhancing Interaction and Immediate Feedback: Creating authentic learning scenarios in which students experience virtual communication contexts can guide students to connect these virtual scenarios to real-world communicative practices, thereby enhancing the practical application of language skills.
• Increasing Autonomy in VR Learning: Unlike in traditional classroom settings, VR environments do not have a universal outcome-oriented or standard answer. Learning participation varies among individuals. Regular monitoring and encouraging students to report their self-learning outcomes could boost their motivation and learning effectiveness [12,14,15,16,17,18,19].

Based on the literature, it can be summarized that instructional content should link VR to an actual classroom when planning VR scaffolding. Course assessments should evaluate English proficiency, peer-assisted task-based learning, and reflective assessments of self-improvement and growth (VR logs). The simplicity and user-friendliness of VR operations are the core considerations when designing instructional strategies. This includes detailed explanations of the headset and controller buttons, visualized standard operation procedures for the expected operations, and descriptions of the learning resources and processes available at the self-learning center. VR self-learning is integrated with the thematic knowledge content of general education courses, differentiation of instructional materials for diverse student needs, rigorous alignment of course assessments with educational objectives, and clear criteria for VR performance evaluation. Additionally, continuous interactions between teachers and students for learning diagnostics and the timing of reinforcing or withdrawing scaffolding significantly influenced the students’ autonomous learning. Therefore, examining these variables could help to develop innovative and culturally appropriate VR communication teaching plans for institutions.

In addition to well-designed lesson plans, evaluation of the effectiveness of scaffolding interventions should play a crucial role. Numerous scholars have demonstrated that using VR as a scaffolding tool can enhance learning outcomes, and this approach has been applied in various fields. For example, Arora et al. [20] developed a VR environment for teaching introductory concepts in C++ programming. Mohamed Shaari and Badioze Zaman [21] used VR to teach poetry, whereas Duke et al. [10] employed VR to help nursing students practice pediatric patient interview techniques. Alemdag and Cagiltay [22] found that visual cues in VR could focus students’ attention, with visual and verbal prompts positively correlated with student learning outcomes. Similarly, Bacca Acosta et al. [23] observed that image-based scaffolding designs in VR positively affect students’ academic performance. These studies indicate that VR as a scaffolding tool positively affects students’ learning performance. However, the design and implementation of scaffolding are complex and vary among individuals. Guidelines on how teachers can use VR as a scaffolding tool and the specific design steps remain unclear.

Quintana et al. [24]’s framework provides the only clearly defined guidelines in the current literature to understand the effectiveness of VR as a scaffolding tool and the aspects of its design that are particularly effective. This framework includes seven guidelines, each accompanied by implementation strategies, and relevant examples. According to these guidelines, scaffolding should use forms and languages that are accessible and comprehensible to learners; organize effective semantic links and tools to highlight the subject knowledge to be learned; allow learners to examine information in various ways; reveal the significant attributes or meanings behind the information; structure complex tasks and related functions in a coherent manner; embed expert guidance on how to perform tasks; automate routine and non-essential tasks; and promote ongoing expression and reflection throughout the intervention process. This framework was adopted as a reference for the scaffolding design in this project and was incorporated into subsequent instructional experiment questionnaires (see Appendix A).

This project defined creativity as the ability to generate innovative ideas through cognitive processes, specifically the use of creative thinking to solve problems. It encompasses four dimensions: fluency, flexibility, originality, and elaboration [25,26,27]. Numerous studies have indicated that integrating VR into education can enhance students’ learning motivation and interactivity. When individuals enter a state of flow, they tend to exhibit higher levels of creativity [28,29]. Therefore, assessing this flow state might help us better understand creativity. In addition, research suggests that VR could enhance creativity through its inherent features of spatial ability, increased focus, and enhanced motivation and engagement [30,31,32]. Although the relationship between the spatial capabilities of VR and creative stimulation is significant, research on this topic is limited [33].

In summary, appropriate scaffolding significantly affects student learning outcomes, particularly when applied at appropriate times. Research indicates that integrating VR into education can enhance learning motivation and engagement and stimulate creative thinking about unfamiliar environments. However, enabling students to independently undertake task-based learning in a VR setting could be challenging and lead to disengagement. Therefore, combining flipped classroom methodologies with VR, linking virtual language tasks to actual classroom activities, and providing real-time diagnostics and differentiated support might make VR teaching more effective. Research has confirmed that VR integration could alleviate speaking anxiety [34] and combining it with scaffolding theory could further enhance teaching effectiveness. Although a scaffolding design could improve educational outcomes, its application in VR instruction remains unclear. Additionally, the existing literature often focuses on student learning but lacks guidance on how teachers should design integration and lesson plans. Clarifying these aspects will facilitate the adoption of VR-integrated teaching across various subjects and classrooms, enabling a comparative analysis of teaching effectiveness in different contexts, and thereby standardizing and empirically validating the use of VR in education.

2. Materials and Methods

2.1. Participants

This study was conducted as part of an undergraduate course aimed at enhancing the literacy and communication skills of students at a medical university in Taiwan. The course targeted students whose English proficiency was at or below the B2 level according to the Common European Framework of Reference for Languages. The study involved 34 students (20 men and 14 women; age range: 18–21 years) who participated in the pre- and post-test assessments. Initially, participants completed a pre-test to evaluate their English literacy background, English-speaking anxiety levels, and perceptions of VR as a learning tool. Subsequently, they engaged in VR-based language learning sessions designed to provide immersive and interactive experiences. Following the VR sessions, a post-test was administered to assess any changes in the anxiety levels and perceptions of VR.

2.2. Instruments and Methods

This study employed a game-based learning model [35] in conjunction with pre- and post-questionnaires, and systematically collected qualitative data on learning processes to explore the relationship between scaffolding use and learning outcomes. VR-based teaching experiments require professional equipment, software, and instructional assessments. Consequently, the research tools were divided into two categories: system setup and instructional experimental tools. The system setup tools included hardware equipment such as desktop computers, monitors, mice, head-mounted displays (HTC VIVE), controllers, wearable microphones, and digital cameras, in collaboration with the digital self-learning center for the VR self-learning process. The VR self-learning courses included “Virtual Speech”, a user-friendly VR application developed by a UK developer primarily aimed at enhancing public-speaking skills. This application offers online interactive remote training, allows students to upload their presentation files into virtual scenarios, tracks learning progress, and provides immediate feedback on speaking performance, which was also evaluated using a rubric (see Appendix B).

The VR sessions were structured over four weeks using the Virtual Speech software. Students participated in a two-hour weekly class that focused on two specific communication scenarios, such as elevator speeches or job interviews. Students engaged in approximately 25 min of VR experience during each session, immersing themselves in the designated communication scenarios. Following the VR experience, the entire class participated in an interactive discussion led by the instructor, focusing on the communication context encountered during the VR session. This interactive teaching approach facilitated deeper understanding and skill development. Students were also encouraged to continue learning independently at the university’s digital self-learning center. Here, they could further explore the VR content and complete their learning profiles and VR logs, enabling the instructor to monitor their progress and provide ongoing scaffolding support.

The instructional experimental tools included a VR scaffolding design framework, anxiety scale, VR clinic, engagement and process VR logs, and scaffolding perception interviews. The VR scaffolding design framework by Quintana et al. [24] was used as a reference for scaffolding design, with its seven guidelines applied in the pre- and post-questionnaires to understand students’ responses and satisfaction with the VR scaffolding intervention and to assess self-learning outcomes (the course list is available in Appendix C). The Public Speaking Class Anxiety Scale developed by Yaikhong and Usaha [36] was used, which originally consisted of 17 items scored on a five-point Likert scale and was refined to 10 items based on previous studies [34] to better suit our context and students’ culture (Appendix D). A weekly real-time session for questions (VR Clinic) was established to understand the students’ use of scaffolding, diagnose issues, provide immediate feedback, and promote self-reflection. This study monitored the students’ VR self-learning hours, engagement levels, and feedback. One-on-one interviews with students confirmed their preparedness and direction for the final performance and clarified their understanding, perceptions, and satisfaction with the implemented scaffolding.

2.3. Data Analysis

This study employed multiple data collection methods to evaluate the effectiveness of the intervention and understand individual differences in the perceptions of scaffolding strategies. Data were collected through (1) questionnaires, (2) semi-structured interviews, and (3) teacher observations before, during, and after the class. The mixed-methods approach first analyzed quantitative data, followed by qualitative data, thereby emphasizing their integration. Questionnaire data were used to analyze the correlation between learning outcomes and four scaffolding interventions: corpora, VR clinics, VR SOP, and VR logs. These findings informed the development of semi-structured interview questions for end-of-term one-on-one interviews with students to verify the consistency between self-assessments and observed behaviors.

To analyze the data, Kernel Density Estimation (KDE) was used to assess whether VR scaffolding interventions could assist in learning English by evaluating the distribution and changes in anxiety levels and perceptions before and after the intervention. The KDE was chosen because of its non-parametric nature, making it suitable for data that do not fit a normal distribution, which is common in psychological and perception studies. Spearman’s rank correlation was used to explore the relationships between learners’ perceptions, anxiety levels, and experiences when using VR for language learning. This approach was selected because it does not assume a linear relationship, which is crucial for ordinal data or small sample sizes as in our study. McNemar’s test was conducted to determine whether there were significant shifts in students’ preferences for scaffolding strategies, VR learning styles, and feedback. This test is particularly useful for assessing changes in categorical data and determining whether the differences between the pre- and post-test distributions are statistically significant. Heatmaps provide visual representations of these shifts.

3. Results

3.1. Correlation Analysis of Pre- and Post-Test Scores in VR-Based English Language Learning

To explore the impact of VR on English language learning, we conducted a Spearman’s rank correlation analysis to investigate the relationships between various pre- and post-test scores related to anxiety, perceptions, and experiences with VR. The analysis revealed several noteworthy correlations that provided insights into the effectiveness and interconnections of VR interventions. Significant positive correlations were identified between students’ anxiety when speaking English compared to when speaking Chinese, and experiencing peer pressure when speaking English, both in the pre- and post-tests (r = 0.93, p < 0.001 and r = 0.74, p < 0.001, respectively) (Figure 2A,B). This suggests that learners who initially felt more anxious about speaking English than Chinese experienced more peer pressure throughout the study. Another strong correlation was observed between the use of VR and software guides in the post-test (r = 0.82, p < 0.001) (Figure 2B), indicating that participants who used both types of guides believed that pre-class VR guides made the learning process more efficient.

Moderate-to-high correlations were evident between the post-test comparisons of L1 proficiency and the VR guidance language (r = 0.60, p < 0.001). The analysis indicated that students with higher L1 proficiency might require better VR guidance to enhance their learning experiences. A moderately positive relationship was found between L1 proficiency and the VR guidance language (r = 0.69), suggesting that students who were more proficient in their first language benefitted more from the well-structured VR guidance. This implies that students with higher L1 proficiency were more likely to appreciate detailed and clear instructions in VR, which could support their learning processes more effectively. Additionally, there was a moderately positive relationship between promoting VR in English learning and the use of VR guides (r = 0.57, p < 0.001) (Figure 2B). This finding suggests that students who found the VR guides useful were more likely to believe in the effectiveness of VR in promoting English learning. This interrelation highlights the fact that students who valued the support provided by VR guides tended to advocate a broader application of VR in their language learning process. These findings highlight the interconnected nature of learning tools and strategies in VR environments. Moderate positive correlations (r = 0.30–0.49) further demonstrated the interconnectedness among various learning tools and strategies. For example, the initial promotion of VR in English learning was correlated with teaching assistant support (r = 0.53, p < 0.001), and the use of the “software-guided” item from the post-test was related to peer pressure (r = 0.53, p < 0.001) (Figure 2B). These results indicate that supportive elements and resources in VR learning are linked to the learners’ overall experiences and perceptions.

Comparing the pre- and post-test scores, our results showed that VR promotes English learning and maintains moderate positive correlations with most factors on both tests. The strongest correlation in the post-test was observed with the software guide (r = 0.61, p < 0.05). This suggests that students believed that the software guides significantly enhanced the effectiveness of VR in promoting English learning. The support provided by the software guides made the VR learning experience smoother and more beneficial, reinforcing the perception that VR could effectively enhance English language learning. Additionally, VR guides demonstrated moderate positive correlations with several factors in the post-test, particularly the software guides (r = 0.48, p < 0.05). Similarly, the software guides showed moderate positive correlations with the VR guides (r = 0.49, p < 0.05) in the post-test. The VR guidance language itself exhibited a strong positive correlation with VR in promoting English learning (r = 0.57, p < 0.001) in the post-test, indicating a significant association post-intervention (Figure 2C). These findings suggest that after the intervention, factors such as VR promoting English learning, VR guidance language, and VR guides were more interrelated, indicating stronger connections or higher degrees of association.

A comprehensive correlation analysis offered critical insights into the relationships between learners’ perceptions, anxiety levels, and experiences in VR-based language learning. These findings supported the hypothesis that VR interventions could reduce English-speaking anxiety and positively affect learners’ perceptions and experiences. Additionally, the results indicated that learners’ prior experiences and proficiency levels significantly influenced the effectiveness of VR-based learning. This analysis highlights the potential of VR to enhance language learning by creating immersive and interactive educational environments. The data supporting these conclusions were derived from a comprehensive correlation analysis and visualizations presented in heatmaps (Figure 2). The heatmaps displayed Spearman’s rank correlation coefficients and corresponding p-values for various factors in both pre- and post-test phases. The key findings are as follows.

• There was a significant correlation between L1 proficiency and the VR guidance language (r = 0.69, p < 0.05), suggesting the need for tailored VR guidance for students with higher L1 proficiency.
• A significant correlation was found between VR guides and the promotion of English learning through VR (r = 0.57, p < 0.05), indicating that students who favored VR guides also supported the use of VR in English learning.
• There was a significant correlation between the software guides and the promotion of English learning through VR (r = 0.61, p < 0.05), demonstrating that students perceived software guides as crucial for the effectiveness of VR in promoting English learning.

These correlations and their statistical significance provided empirical support for the points discussed.

Interview data corroborated these findings. Most of the interviewed students (n = 31) who supported the use of VR guides explicitly stated that they were more likely to advocate for the application of VR in English learning. In addition, students perceived the software guides as having the most significant impact on the efficacy of VR in enhancing their English learning. As consistently highlighted in the interviews, this endorsement appeared to stem from the observation that students who used guides performed better and showed greater support for the VR experiments. This is attributed to the guides providing clear instructions and support, thereby reducing confusion and anxiety, and boosting confidence and learning outcomes. According to Robert, an interview participant,

Initially, I was quite nervous about using VR for English learning. I had never used VR before and was worried I might not handle it well, potentially affecting my performance. However, the VR guide made a huge difference. It provided clear, step-by-step instructions on using the equipment and navigating the learning modules, along with troubleshooting tips. This clarity reduced my confusion and anxiety, and as I became more familiar with the technology, my confidence grew. I started participating more actively in VR activities, especially virtual English conversations, and noticed a significant improvement in my learning outcomes. The guide was crucial in helping me overcome my fear of new technology, ultimately making the learning experience enjoyable and effective.

Furthermore, student anxiety about VR-based learning stems primarily from several factors. First, many students lack prior experience with VR technology, leading to unfamiliarity and discomfort. They worry about their ability to operate VR equipment effectively, which could result in difficulties and setbacks during the learning process. Second, the VR learning approach differs significantly from traditional language learning methods, causing students to be uncertain of the adaptability and efficacy of this new mode of learning. Most critically, students fear that these technological challenges might negatively affect their academic performance, and ultimately, their grades. Consequently, these concerns have hindered the acceptance of and support for VR-based learning. According to another interview participant, Nancy,

As a first-year medical student, I felt very anxious about VR-based learning even with a Common European Framework of Reference for Languages (CEFR) C1 level of English proficiency. Having never used VR technology before, I worried about my ability to effectively operate the equipment. I feared that any technical difficulties might impede my learning progress. Additionally, the VR learning environment was vastly different from the traditional classroom settings I was accustomed to, making me uncertain about how well I could adapt and benefit from this new approach. Most importantly, I was deeply concerned that my performance in the VR modules might not meet my expectations, potentially affecting my overall semester grades. However, after receiving detailed guidance, my situation improved significantly. The guide provided clear steps and troubleshooting tips, which helped reduce my fear of the technology and gradually familiarize myself with the VR equipment. As I became more comfortable with the VR technology, my confidence grew, allowing me to participate more actively in learning activities. Ultimately, my performance improved, and so did my grades.

3.2. Overall Impact of VR Scaffolding in Assisting Students to Learn

We examined the impact of VR learning interventions on English language acquisition using pre- and post-intervention tests. Descriptive statistics indicated an increase in the mean score from 4.00 (pre-test) to 4.35 (post-test), with standard deviations of 0.78 and 0.73, respectively, suggesting the effectiveness of VR interventions (Figure 3. The KDE further validated these findings, showing a significant improvement in scores with a p-value of 0.0377. Box plot analysis also visually confirmed an increase in scores from the pre- to post-tests (Figure 3. These results highlight the significant positive impact of VR learning interventions on English language acquisition and suggest that incorporating VR technology into language learning curricula effectively enhances learners’ English speaking and comprehension proficiency.

We analyzed the correlations between pre-test and post-test scores in a VR-based English language learning intervention, organized into three clusters: “overall”, “anxiety”, and “user experience”. Spearman’s method was used for correlation testing (Figure 4). The “overall” cluster, which included questions on the general impact of VR on English learning, showed a moderate positive correlation between pre- and post-test scores in the “user experience” cluster (r = 0.53). This suggests that the respondents found user guides, including software, VR guides, and teaching assistants, crucial for learning English through VR. Additionally, pre-test scores in the “overall” cluster showed moderate positive correlations with post-test scores in the “overall” (r = 0.35), “anxiety” (r = 0.42), and “user experience” (r = 0.51) clusters, indicating that initial perceptions moderately influenced post-intervention outcomes. The “anxiety” cluster, focusing on anxiety related to speaking and learning English, revealed weak negative correlations between initial anxiety levels (pre-anxiety) and initial “overall” perceptions (r = −0.099) as well as post-test “overall” perceptions (r = −0.12). This suggests that higher initial anxiety slightly reduced overall perceptions. Pre-anxiety showed weak to negligible correlations with other factors, indicating a minimal influence on user experiences and perceptions before and after the intervention. The “user experience” cluster, centered on the VR user experience, showed moderate positive correlations between initial user experiences (pre-user experience) and pre-test “overall” (r = 0.53), post-test “overall” (r = 0.24), and post-test “user experience” (r = 0.39) scores. This implies that positive early experiences moderately predicted positive perceptions and post-intervention experiences. There was also a weaker positive correlation with post-test “anxiety” (r = 0.31), suggesting that initial user experiences had a modest impact on reducing anxiety levels after the intervention. These findings highlight the importance of user guides and positive initial experiences in enhancing the effectiveness of VR-based English-language learning interventions.

The post-intervention measures revealed stronger correlations. Post-test “Overall” scores showed significant positive correlations with post-test “Anxiety” (r = 0.59) and “User Experience” (r = 0.57) scores, indicating that overall post-intervention positive perceptions were associated with reduced anxiety and improved user experiences. Additionally, post-test “Anxiety” and “User Experience” exhibited a strong positive correlation (r = 0.71), suggesting that reduced anxiety was significantly linked to better user experiences. The heatmap (Figure 4) indicated that initial perceptions and user experiences with VR moderately predicted post-intervention outcomes. However, post-intervention measures were more strongly interrelated, suggesting that positive experiences and reduced anxiety were linked to better overall perceptions of VR-based language learning interventions. These findings support the hypothesis that VR interventions could enhance learners’ experiences and perceptions, with initial user experiences playing a crucial but moderate role in shaping these outcomes.

In summary, the data indicated that interventions using VR for English learning can effectively enhance learning outcomes and that user guides, along with the initial user experience, play a significant role in the results. Comparative interview data also revealed that most students (n = 29) stated that user guides helped them because they previewed VR instructional content and demonstrations by teaching assistants, thereby alleviating their anxiety. Interestingly, the students also mentioned that the introduction of VR into teaching complicated learning because of the dual anxieties related to language proficiency and technological literacy. According to Leo, an interview participant,

I think using VR for learning English really helps improve my skills. The user guide is very important. It helps because it shows me what to expect, and the teaching assistant’s demonstrations reduce my anxiety [I can just be a copycat (laughs)]. Sometimes, I’m not just worried about my English, but also about not being good with new technology. The VR lessons make learning more complicated because I must deal with language and technology worries.

4. Discussion

4.1. Analysis of English Proficiency Groups

The students were divided into four groups based on their English proficiency levels (A2, B1, B2, and C1), and Spearman’s rank correlation analysis was conducted to explore the relationships between their proficiency levels and other factors in both pre- and post-tests (Figure 5). This analysis provided several insights. In the pre-test condition, A2 proficiency showed weak correlations with “overall” (r = −0.013), “anxiety” (r = 0.26), and “User Experience” (r = 0.30); B1 proficiency exhibited a moderate positive correlation with “anxiety” (r = 0.13) and a very weak correlation with “User Experience” (r = 0.01); B2 proficiency had weak positive correlations with “anxiety” (r = −0.32) and “User Experience” (r = −0.22); and C1 proficiency showed moderate positive correlations with “overall” (r = 0.24), “anxiety” (r = 0.26), and “User Experience” (r = 0.23). In the post-test condition, A2 proficiency had weak positive correlations with “Overall” (r = 0.21), “Anxiety” (r = −0.14), and “User Experience” (r = 0.19); B1 proficiency showed very weak correlations with “Overall” (r = −0.13), “Anxiety” (r = −0.043), and “User Experience” (r = −0.088); B2 proficiency exhibited weak positive correlations with “Overall” (r = 0.17), “Anxiety” (r = 0.15), and “User Experience” (r = 0.096); and C1 proficiency showed moderate negative correlations with “Overall” (r = −0.31), “Anxiety” (r = −0.14), and “User Experience” (r = −0.23). The p-values for these correlations indicated that none were statistically significant (p > 0.05). Therefore, no significant differences were found among the English proficiency groups (A2, B1, B2, and C1) in their perceptions of and experiences with the VR learning tools.

4.2. Comparison of Scaffolding Preferences

McNemar’s test was conducted to determine whether there were significant shifts in students’ preferences for scaffolding strategies, VR learning styles, and feedback. The heatmaps in Figure 6 visually represent these shifts.

For scaffolding strategies, the comparison revealed an increase in the user-friendliness of the software interface (from zero in the pre-test to five in the post-test), and an increase in the diversity of the effectiveness assessment (from zero in the pre-test to two in the post-test). However, the p-value for McNemar’s test was 1.000, indicating no statistically significant shift in the preference for scaffolding strategies. In subsequent interview sessions, 18 students mentioned the importance of diverse assessments within the VR environment. They highlighted that diverse assessments provided a more comprehensive evaluation of skills, catered to different learning styles, and helped reduce anxiety by offering multiple ways of demonstrating understanding. Among these students, 12 specifically emphasized the value of immediate scaffolding support after the assessments (see Figure 7). Sandra, another interview participant, mentioned the following:

When I finish a VR task, I get instant feedback on my performance and suggestions for improvement. This helps me know what I did well and what I need to work on. For example, after practicing English speaking in VR, I received tips on how to improve my pronunciation. This feedback boosts my confidence and motivates me to keep trying. I feel less lost and more supported in my learning journey.

4.3. Comparison of VR Learning Styles

Regarding VR learning styles, there was a noticeable shift from individual to collaborative learning (from one pre-test to three post-tests). However, the p-value for McNemar’s test was 1.000, indicating no significant shift in learning style preference (Figure 8).

4.4. Comparison of Feedback Preferences

Regarding feedback preferences between humans and AI, there was an increase in the preference for practicing with machines to reduce stress (from two in the pre-test to seven in the post-test). However, McNemar’s test yielded a p-value of 0.500, indicating no significant shift in the feedback preference (Figure 9).

Based on the correlation analysis and McNemar’s test results, we concluded that there were no significant differences in perceptions and experiences among the English proficiency groups (A2, B1, B2, and C1). Additionally, no significant shifts in preferences for scaffolding strategies, VR learning styles, or feedback preferences were observed from pre- to post-tests. These findings suggest that although VR learning interventions positively influence learners’ overall perceptions and experiences, individual preferences and experiences with VR learning tools remain relatively stable across proficiency levels and over time. Individual learning preferences and styles are typically long-established and are not easily altered in the short term. Although VR learning interventions provide new experiences, students might still rely on familiar learning methods and tools. Additionally, students lacking experience with VR technology might require more time to adapt to and accept this new technology. Short-term interventions might be insufficient to fully shift students’ preferences regarding learning tools. Finally, the limited duration of the intervention might have prevented the students from fully experiencing all the benefits of VR learning. Long-term use and extensive practice might lead to significant changes.

5. Conclusions

Several educational and research applications were derived from the findings of this study. Differentiated instruction could be highly effective in educational settings. Given that initial perceptions moderately influence post-intervention outcomes, educators could adjust their teaching strategies according to students’ initial reactions and feelings, thereby providing personalized guidance. For example, educators should examine students’ previous VR experience to avoid dual anxiety (language and technology). In addition, emphasizing user guides is crucial, because positive early user experiences predict better perceptions and experiences. When designing user guides, it might also be beneficial to consider the different learning styles of students. Teachers could consider using the Visual, Auditory, Read/Write, and Kinesthetic (VARK) models to identify four types of learners: visual, auditory, kinesthetic, and reading or writing [37]. In addition, different VR learning software, learning paths, and assessment methods should be recommended for students with different learning styles. We also observed the importance of diversity in assessment, which is consistent with the previous literature on the significance of cognitive scaffolding in VR learning. This scaffolding helps students monitor and reflect on their learning processes [8].

Furthermore, reducing anxiety should be a priority. Although initial anxiety had a minor impact on overall perception, it is important to address student anxiety levels. Designing user-friendly VR environments and providing appropriate scaffolding support could reduce anxiety and improve learning outcomes. This study found that the primary source of student anxiety was uncertainty regarding assessment effectiveness, as VR environments do not have a common outcome orientation or standardized answers, unlike traditional teaching methods. This finding aligns with those of previous studies [12,15]. Regularly monitoring student learning and training students to self-regulate and report their learning processes are necessary. Teachers should continuously interact with students and scaffolding, adjusting or removing it flexibly to maximize student engagement and learning outcomes.

This study investigated the application of virtual reality (VR) technology in English language education, specifically examining the effectiveness of VR scaffolding. Results confirm that VR scaffolding can promote English language learning. However, further research is needed to explore the impact of various VR content types and activities on learning outcomes. Given the study’s focus on short-term effects, further research should investigate the long-term impact of VR interventions on language skills and learner attitudes through longitudinal studies. Additionally, future research should examine the performance differences among learners with varying English proficiency levels, particularly within more complex learning tasks.

The study holds valuable implications for both educators and policymakers. For educators, it underscores the need for differentiated instruction, advocating for personalized guidance tailored to individual VR experiences and learning styles. For policymakers, this study emphasizes the importance of supporting VR technology integration within educational settings. This support can manifest through funding for VR infrastructure, developing relevant curricula, and implementing teacher training programs focused on effective VR implementation. The study acknowledges limitations, including the lack of standardized assessment methods in VR environments. Future research should address this gap by developing and validating assessment tools for VR learning contexts. Addressing these limitations will provide a more comprehensive understanding of VR’s potential in language education and guide the development of effective VR-based learning interventions.