If a dialog system can respond to a user as natural as a human, the interaction will be smoother. In this research, we aim to develop a dialog system which can make various behaviors appearing in chat-like dialogs. In this paper, we developed a dialog system which could generate chat-like responses and their timing using a decision tree. The system could make “synchronous utterances”, “aizuchi” and so on. The decision tree utiliezed the pitch and the power contours in the last 100 ms of user’s utterance, recognition hyposethes, and the prepared response contents at that time as features. This system also reacts user’s overlapping utterances and disfluencies robustly. 1 [1] [2] [3]