Studying the behavior of people who read braille by tracking their hands typically requires fixtures on the fingers or complex laboratory setups. This paper investigates if and how a state-of-the-art virtual reality hand tracking system from Ultraleap (a stereo infrared camera designed to track hands in mid-air) is able to detect fingers on a refreshable braille display without such inconveniences. Different camera positions and predefined software modes were experimentally compared by observing (with a novice and an expert braille user) whether fingers moving on the braille cells were detected at all, whether they were consistently tracked after initial detection, and whether the visualization of the virtual hand model was true to reality. The most reliable results were obtained with a camera position above the braille display in combination with the “head mounted” mode, which is normally intended for use on a virtual reality headset. In 100 iterations, the system failed to detect the hand(s) only twice. However, the braille display in such a setting requires a surface that appears dark in infrared light to provide sufficient contrast. On a display without this feature, detection failed 29 times out of 80 iterations. The results provide evidence that the system can be used for braille finger tracking under certain conditions.