To provide a pronunciation evaluation system grading whether pronunciation is good or bad with a simple configuration by utilizing a speech teaching material which is heretofore in widespread use.

The speech of the teaching material is inputted from an MD player 2 and is outputted from a speaker 11 and its stress accent, tonic accent, intonation, etc., are analyzed in a DSP 13. The results of the analysis are stored into a memory 14. Next, the speech of a learner's practice is inputted from a microphone 3 and the pronunciation information, such as its stress accent, tonic accent, and intonation, are analyzed by the DSP 13 like with the speech of the teaching material. If the pronunciation information of the learner's pronunciation is the same as that of the teaching material, the speech is considered to be correctly pronounced and therefore both pronunciation information are compared and the grading result based on the similarity thereof is displayed on a display 22. As a result, the grading of the learner's pronunciation is carried out by using the conventional speech teaching material not having the information for the purpose of the grading.