A pronunciation training system and methods are provided as a series of programmed routines stored on an item of removable storage media, and select information generated by a speech analysis engine to compute and display graphical representations of metrics useful to a student. The student selects from among a plurality of pre-recorded utterances spoken by a native speaker, and the student then records his or her pronunciation of the utterance. The software computes and displays graphical metrics for the native speaker's utterance and the student's utterance, in any of a variety of formats, on a side-by-side basis. The system also permits the student to repeat selected phrases and to monitor improvement by similarity between the graphical metrics.