Visual-to-tactile sensory substitution devices are designed to assist visually impaired people by converting visual stimuli into tactile stimuli. The important claim has been made that, after training with these devices, the tactile stimuli can be moved from one body surface to another without any decrease in performance. This claim, although recurrent, has never been empirically investigated. Moreover, studies in the field of tactile perceptual learning suggest that performance improvement transfers only to body surfaces that are closely represented in the somatosensory cortex, i.e. adjacent or homologous contralateral body surfaces. These studies have however mainly used discrimination tasks of stimuli varying along only one feature (e.g., orientation of gratings) whereas, in sensory substitution, tactile information consists of more complex stimuli. The present study investigated the extent to which there is a transfer of tactile letter learning. Participants first underwent a baseline session in which the letters were presented on their belly, thigh, and shin. They were subsequently trained on only one of these body surfaces, and then re-tested on all of them, as a post-training session. The results revealed that performance improvement was the same for both the trained and the untrained surfaces. Moreover, this transfer of perceptual learning was equivalent for adjacent and non-adjacent body surfaces, suggesting that tactile learning transfer occurs independently of the distance on the body. A control study consisting of the same baseline and post-training sessions, without training in between, revealed weaker improvement between the two sessions. The obtained results support the claim that training with sensory substitution devices results in a relative independence from the stimulated body surface.