In some applications, no real-estate is available for speech-specific displays. In some situations, such as telephone-based recognition, no display is possible.
Auditory feedback is often confusing for the novice or casual user, since there is no ready framework for interpreting auditory feedback. Thus is it best to restrict the variety of auditory feedbacks and to make their meaning easy to discern from theri context of occurence. It is also useful to borrow signals whose meaning is already established in other environments, say telephony.