For a natural communication robot cooperating with human, an adequate control mechanism of motion and utterance is required. This paper presents a robot motion planning method which considers utterance timing by utilizing Self-Organizing Map (SOM). Adequate target position of the robot motion and the utterance timing for an autonomous robot are decided by searching the best-matching-node on the SOM which was trained using normative data of human behavior. Applying the presented method to cooperative carrying situation in a virtual cooperative simulator, human impression to the robot behavior was investigated. As a result, it was confirmed by the Tukey-Kramer test that several human could recognize an effect of the utterance timing and motion planning method.