I present and discuss the SAPHO (Segmentation by Acoustico-Phonetic knowledge) model implemented in the AWk language under the Unix system on a MASSCOMP computer. The system is devised as a speaker independent ASS (automatic speech segmentation), by previous recognition of the phonetic articulation manner. In all the ASR systems the phonetic knowledge is at least implicitely used. It has to be referred to explicitely. Following the Level Building procedure SAPHO supplies a hierarchized set of acoustic properties and segments, and phonetic properties and segments which fit the phonetic parsing of the acoustic wave. The amenability of this system is entailed by its modularity which allows a possible further architecture as distributed tasks.