Recent developments in the design of human-machine interfaces have resulted in interfaces which make access to computer-based equipment more difficult for visually disabled people. The aim of this project was to explore whether it is possible to adapt such interfaces so as to make them usable by people who cannot see a screen. The approach adopted was based upon two principles: the replacement of visual interlace entities by auditory analogues and appropriately constraining the resultant interlace. Two forms of sound were used to embody the auditory interface: musical tones and synthetic speech. The 'auditory screen' so produced consists of a grid of 'auditory objects', each of which is associated with a spatial location, a tone, a name and an action. In order to test the principles a word processing program was designed and implemented to demonstrate how a visual program might be adapted to be accessed through such an interface. This program was evaluated with the assistance of a number of visually disabled testers. They were trained to use the word processor through completing a graded set of exercises. Data were obtained in a number of ways during and after the completion of the exercises. Since the auditory interface had some novel components, special evaluation methods were applied. The nature of the interaction was analyzed, using an approach based on work on keystroke models of visual interfaces. This yielded a simple model of the 'hand-ear' coordination employed, which forms a basis for prediction of user behaviour. It was also necessary to evaluate aspects of the program, such as ease-of-learning and these were investigated by interviewing the subjects. The results demonstrate that the approach is viable. The thesis also discusses a number of problems in using such an interface, most of which are related to the memory load on the user.