PURPOSE: To facilitate an operation when training for language study is performed.

CONSTITUTION: A recording/reproduction processing part 10 paragraphs the voice of a teacher that is a sample by a silent part for several seconds or more when it is inputted, and records it on a RAM 12 at every phrase. When the training is performed, the voice that is the sample of one phrase is read out from the RAM 12, and it is reproduced, then, a mode is transferred to a recording mode, and the voice of a pupil of one phrase is recorded on the RAM 12 similarly. After that, the voice of the teacher and that of the pupil are reproduced by automatically reading out from the RAM 12. Therefore, the operation can be remarkably simplified without issuing an instruction from which position the training should be started, etc. Also, recording/reproduction with length required for the training can be automatically performed. Furthermore, since a training voice can be automatically recorded in phrase unit, the operation can be remarkably simplified.