A robot control device includes a learning control unit for calculating a learning correction amount, a position storage unit for storing a position of a leading end of a robot mechanism part during the learning control, and a speed storage unit for storing a speed of the leading end of the robot mechanism part during the learning control. The robot control device determines, while the robot mechanism part is operated by a position command after the learning control, whether or not the position and the speed of the leading end are in an abnormal state based on errors with respect to the position and the speed of the leading end stored during the learning control. The robot control device switches a determination as to whether the learning correction amount is applied in accordance with this determination result.