The invention provides a technique for learning a stability factor of a vehicle by which the stability factor of the vehicle can be calculated in response to a state of the vehicle. To this end, a state of the vehicle during traveling is detected, and it is decided based on the detection information whether or not the current traveling state of the vehicle is a stable turning state. If it is decided that the current traveling state of the vehicle is a stable turning state, then the stability factor of the vehicle is calculated in accordance with a predetermined arithmetic operation expression based on a vehicle speed, a steering angle and a yaw rate detected during traveling, and the calculated value is determined as a learned value of the stability factor of the vehicle thereby to determine the stability factor in the current state of the vehicle.