Some problems and solutions related to automated error detection/correction as they pertain to the improvement of foreign language pronunciation are discussed as well as their underlying premises. A model is proposed as a functional solution which is based on pedagogical framing and hierarchical error processing. The application of contrastive/applied phonetic principles both in the predicting of errors and the pedagogical framing of sounds allows the development of efficient analysis routines which do not rely upon full-spectrum matching and maximizes the chances for success.