This work, first published in 1995, is primarily addressed to phonologists interested in speech and to speech engineers interested in phonology, two groups of people with very different expectations about what constitutes a convincing, rigorous study. The subject matter, the application of autosegmental theory for Markov modeling, is technical, but not really esoteric - autosegmental theory is at the core of contemporary phonology and Markov models are the main tool of speech recognition. Therefore, it is hoped that anyone interested in at least one of these two fields will be able to follow the presentation.