Abstract: In this paper, automobile drivers are divided into two groups; one is called accident group every member of which has ever had more than 2 accidents; and the other one is called non-accident group the member of which has never had an accident. The testing method for drivers’night visions is studied, and then a comparison of night vision between 213 drivers in accident group and 652 drivers in non-accident group is carried out. The tested results show that there exists a appreciable difference in night vision between 142 drivers of accident group and 459 drivers of non-accident group (p<0.05). It appears that drivers’night vision disorders are the main cause of traffic accident happenings. Meanwhile, a safety reliability assessment method is proposed for driving at night.