Driver states adaptive drive supporting system is highly expected in use for the reduction of the number of traffic accidents. This study aimed at creating a constituent technology for detecting driver's cognitive distraction, which may be one of major factors of driver's psychosomatic states just before a traffic accident. We reproduced driver's cognitive distraction by means of imposing cognitive loads such as arithmetic and conversation to a subject on a driving simulator. Besides gaze angle, head rotation angle, and interval between heart R-waves (hereafter, heart rate RRI) from an ECG (electrocardiogram), we added pupil diameter of a subject as recognition features for pattern recognition. We established high accuracy and rapid detection methodology for driver's cognitive distraction by adopting the AdaBoost.