This article, presented in three parts, is intended to be a tutorial on an analysis/synthesis system that I have been using to facilitate the performance of digital Musique Concr~ete in a MUSIC4 (FORTRAN) environment. Part 1 described several pieces that have been realized using this system, and introduced signal modeling in the context of human speech. Part 2 described (in nonmathematical terms) the analysis of speech based upon this model. This installment will present the mathematics, including an introduction to the concept of linear prediction (those who are already familiar with the theory of digital filters can skip directly to "General Inverse Filtering"). An "appendix"-a listing of FORTRAN programs and a user's manual, as well as copies of the pieces discussed are available from the author (Box 329, Route 13, Stillman, New Jersey 08558).