Background: This study investigates the effect of immersion levels in virtual reality (VR) learning environments on the acquisition of declarative and procedural knowledge. Prior research indicates that immersion affects cognitive load, but its impact on declarative and procedural knowledge outcomes remains unclear. This study utilises a multimodal approach, combining self-reports with data from psychophysiological sensors. Objectives: The study aims to evaluate how different levels of immersion in VR--high-immersive VR (head-mounted display, HMD-VR) and low-immersive VR (Desktop-VR)--affect cognitive load and learning outcomes, focusing on procedural and declarative knowledge. Methods: A prospective experimental study was conducted with 74 sophomore nursing students, assigned to either the HMD-VR group or Desktop-VR. Eye-tracking and electrodermal activity (EDA) were used to assess cognitive load during learning. A pre-test/post-test design measured declarative and procedural knowledge using a Medication Administration Test (MAT), whilst a presence questionnaire evaluated user experience. Results and Conclusions: Both study groups demonstrated significant improvements in declarative and procedural knowledge. However, the low-immersive Desktop-VR group exhibited significantly greater pre- to post-test gains in procedural knowledge compared to the high-immersive HMD-VR group. The HMD-VR group exhibited higher cognitive load during procedural tasks, indicated by lower blink rates and a higher rate of EDA peaks. No significant differences were found in the sense of presence between the two groups. This study contributes to understanding learning with immersive VR, showing that high-immersive VR may require careful instructional design to mitigate cognitive overload, especially for procedural tasks. Low-immersive VR presents a cost-effective alternative for immediate knowledge gains.