The need to transcribe infant sound productions from birth onwards by using universally applicable coding tools has been basic to the development of our AMSTIVOC classification system. In this system early infant vocalizations are described by means of a sensori-motor approach based on the source-filter model for speech production. We applied the AMSTIVOC classification system, among other things, to early vocalizations of 6 deaf and 6 hearing infants in order to answer the question whether and where the lack of auditory perception can be traced in the early sound productions of deaf infants. By using this classification system it can be demonstrated that auditory feedback is needed to coordinate the movements of the phonatory and the articulatory system. This coordination capacity is likely to be a prerequisite for the development of normal speech production.