The development of a spatial audio ear trainer has been proposed and investigated. A review of the relevant literature has shown that although numerous researchers have simulated spatial characteristics of sound, the devised simulation algorithms were not (properly) verified with respect to their intended subjective effects, thus making them unsuitable for training purposes. Additionally, whilst various timbral ear-training systems have been set up, there is none for spatial quality. To address these shortcomings four spatial attributes of reproduced sound have been considered: source distance, source width, ensemble width and ensemble depth. For each attribute a processing algorithm was developed that allowed highly controlled changes in the respective percept. Using these algorithms four sets of stimuli were carefully generated with the aim of achieving unidimensional variation in terms of the qualities of interest. In order to allow detailed and reliable validation of the simulations' auditory effects, a sensory evaluation strategy was devised that relies on Multidimensional Scaling techniques and the elicitation of supplementary qualitative data as a means of obtaining a complete sensory profile of a group of stimuli. With the help of this methodology as well as critical listening panels the psychological structures of the sound excerpts were measured. Results showed that the envisaged unidimensionality of the source distance, ensemble width and ensemble depth samples was accomplished. As to the source width simulation, results were not as clear-cut, which was attributed to listener unfamiliarity with this perceptual construct and greater stimulus uncertainty. Nevertheless, the analyses unfolded only one major dimension, which is why all four attribute simulations were deemed suitable for the proposed training purposes.