When the visual channel of communication is unavailable because the user is blind, non-visual user interfaces must be developed. The proposed methodology consists of three interrelated specification levels. Information and supported tasks are specified in abstract terms at the conceptual level, taking into account requirements imposed by manipulation of interaction devices and information provided by analysis of the visual representation. The perceptual structure of the auditory scene is specified next at the structural level and then the physical dimensions of sound are defined at the implementation level. The methodology is applied to the specification of a simple listbox widget.