This study addresses the under-researched area of multimodal output enhancement in Content and Language Integrated Learning (CLIL) science education, specifically focusing on the impact of virtual reality (VR) games on fourth-grade students' English speaking and writing skills. While multimodal input and digital tools are increasingly recognized for their potential in boosting EFL learners' productive language skills, empirical research examining their effect on multimodal output, particularly writing and speaking, within CLIL remains limited. In Taiwan, the shift towards CLIL presents challenges, including an imbalance between receptive and productive skills, difficulties in teaching abstract scientific concepts in a non-native language, and increased cognitive load for students. This study aims to bridge this research gap by investigating whether VR games, rich in multimodal input, can effectively enhance fourth-grade CLIL students' productive language skills in English poster designs (writing) and oral presentations (speaking) compared to traditional PowerPoint (PPT)-led games. The study grounds its approach in self-regulated learning (SRL) theory, hypothesizing that the immersive and interactive nature of VR games will positively influence students' ability to convey scientific concepts through multimodal output, assessed using the 4Cs (Content, Communication, Cognition, and Culture) framework.

The study's theoretical framework is self-regulated learning (SRL), which emphasizes learners' active management of cognitive processes, social interactions, and motivational strategies. The literature review explores existing research on SRL's impact on English speaking and writing skills in EFL contexts, highlighting the benefits of SRL strategies in both digital and virtual learning environments. It also examines the potential of gamification and game-based virtual reality learning environments (VRLEs) for enhancing language learning, citing studies demonstrating VRLEs' effectiveness in improving writing and speaking skills. However, a critical gap exists in research on VRLEs' impact on multimodal output within CLIL science education. The literature review further examines multimodal assessments, emphasizing their potential to capture diverse aspects of student learning, especially in CLIL settings. The 4Cs framework (Content, Communication, Cognition, and Culture) guides the assessment design, ensuring holistic evaluation of students' multimodal productions.

This quasi-experimental study used a comparative case study design with 81 fourth-grade students from three Taiwanese public elementary schools. Two classes were randomly assigned to an experimental group (EG, n=40) using VR-based review games, and two to a control group (CG, n=41) using PPT-led review games. Both groups received identical instruction on five science units, followed by 20-25 minute review sessions using either VR or PPT games. The VR games, developed using CoSpaces Edu Pro, incorporated gamification elements, interactive features, and multimodal input (text, visuals, audio) and utilized a buddy system for collaborative learning. The PPT games also incorporated gamification and multimodal input, serving as a control condition. Multimodal assessments included English poster designs (writing) and oral presentations (speaking), evaluated using rubrics based on the 4Cs framework. Two instructors independently rated both posters and presentations. Interrater reliability was high. Independent-samples t-tests compared the EG and CG quantitatively. Post-intervention, the raters engaged in a reflective session to provide qualitative insights.

Quantitative results showed that the EG (VR games) significantly outperformed the CG (PPT games) in Content and Cognition for both poster designs and presentations. Specifically, EG students demonstrated greater depth, accuracy, and application of scientific concepts and higher-order cognitive skills. In terms of Communication, the EG showed higher target vocabulary usage and sentence complexity in presentations, but not in posters. Qualitative data from rater reflections corroborated the quantitative findings. Raters observed more innovative and integrated use of scientific content, critical thinking, and multimodal expressions in the EG's work, indicating deeper engagement with the material. The EG consistently demonstrated higher-order cognitive skills (e.g., evaluation, application, analysis) as opposed to the CG group which primarily demonstrated recall and basic understanding. Specifically, EG posters integrated graphical elements from the VR game directly into their designs, showcasing a direct link between the VR experience and their multimodal output. The EG presentations demonstrated a more comprehensive and detailed content knowledge, utilizing more precise and sophisticated language compared to the CG group.

The findings suggest that VR games effectively enhance multimodal output in CLIL science education, addressing the research questions. The superior performance of the EG in Content and Cognition can be attributed to the immersive and interactive nature of VR games, facilitating deeper engagement and promoting higher-order cognitive skills. The multimodal input in VR games, particularly the combination of graphic and textual information, interactive features (avatars, information boards), and the buddy system, contributed to enhanced vocabulary acquisition, sentence complexity, and overall understanding. The results support the effectiveness of multimodal assessments in capturing a holistic view of student learning, as evidenced by the convergence of quantitative and qualitative data. The significant differences in Content and Cognition are aligned with theoretical expectations related to SRL and the affordances of VRLEs. The lack of significant differences in Culture highlights a limitation – the need for future research to incorporate more explicit cultural content into VR environments to further enhance cultural awareness.

This study empirically demonstrates that game-based VRLEs significantly enhance students' multimodal output in content and cognitive skills within a CLIL context. The findings support the integration of VR games into CLIL classrooms to improve students' productive language skills and deep understanding of scientific concepts. Future research should focus on incorporating cultural content into VR environments, investigating long-term effects, and employing larger and more diverse samples to improve generalizability. The study also highlights the value of multimodal assessments in capturing the full range of student learning outcomes in CLIL.

The study's limitations include a relatively small sample size and specific context (fourth-grade CLIL science students in Taiwan), limiting generalizability. The focus on cognitive and content dimensions, with less emphasis on affective and cultural aspects, is another limitation. Potential observer bias in rater reflections and the absence of long-term follow-up are also noted. Finally, the lack of a publicly accessible project website for the VR games limits broader dissemination.