According to the modality principle of the Cognitive Theory of Multimedia Learning, learning is facilitated when visual content is accompanied by vocal narration compared to text explanation. However, the application of this principle in immersive virtual reality (IVR) remains unclear. In the present experiment involving 101 secondary school students, was compared the application of the modality principle in IVR and on desktop. Measurements of learning outcomes, eye movements behaviour, presence and cognitive load were carried out during a lesson about forest, in IVR or on desktop, with vocal narration or with text, involving four experimental groups. A modality effect was found for the desktop but not for the IVR condition, suggesting a reversed modality effect. The desktop condition was associated with less visual distraction, while the IVR condition led to higher presence. This study showed that the content of lessons in IVR should be adapted to avoid distractions.