Reaction time training systems are used to improve user performance. Until now, such setups use physical 2D flat surfaces, e.g., a 2D touch screen or buttons mounted on a wall. We designed and investigated a mid-air reaction time training system with an immersive virtual reality (VR) headset. 12 participants performed an eye-hand coordination reaction test in three conditions: both in mid-air with or without VR controller as well as with passive haptic feedback through hitting a soft-surface wall. We also altered target and cursor sizes and used a Fitts' law task to analyze user performance. According to the results, subjects were slower and their throughput was lower when they hit a solid surface to interact with virtual targets. Our results show that Fitts's model can be applied to these systems to measure and assess participant training.