Scaffolding is a mechanism for helping novice learners to solve a problem or complete a task that is beyond their current skill set. Prior research on immersive virtual reality (IVR) environments with scaffolding has mainly focused on the effectiveness of scaffolding for conceptual understanding and information processing. However, research on defining guidelines for designing computer based scaffolds for IVR is scarce. To fill this gap, some recommendations are provided in this paper on how to design scaffolds for IVR environments for teaching English as a Foreign Language (EFL). An eye tracking study was conducted with N = 41 university students from three EFL courses that used two versions of an IVR application for learning and practicing prepositions of place in English with different scaffolding mechanisms and in different contexts: a context in which the participant can move through the virtual space and a context in which the participant remains seated. A comparative analysis was conducted to determine which scaffolds were more appropriate for learning in IVR. This study extends previous research on scaffolding in IVR environments by suggesting some empirical evidence, collected from eye tracking, of the effectiveness of different types of scaffolds in IVR environments and some design guidelines are provided. The results showed that scaffolds that are naturally embedded in the IVR environment and are closer to the objects that students need to interact with, are more effective for increasing their learning performance. Practical implications of this study for instructional designers and developers include: scaffolds should be designed considering the spatial contiguity principle and should be naturally blended with the VR environment. Scaffolds should direct students attention to the most important information and scaffolds can be provided in different forms (text, images and on site indicators).