Abstract
This study investigated on the dialogues between four students from Japanese junior class and three Japanese TAs, covering six topics for 472 minutes, and analyzed error-corrections and non-error-corrections for 971 errors. Analysis results revealed that overall error-correction ratio was low, and the highest one is about lexical errors. Explicit correction was used commonly, while guiding students for self-correction was few. Most of the error-corrections didn't break down the communication, while when pronunciation was incorrect, the communication was easily broken down. In addition, when the meaning of students was ambiguous, error-correction ratio was relatively high, and repetition and clarification request were usually employed.
