Authors investigated the accuracy of anaverageannual daily traffic volume obtained from the observation of an arbitrary number of days as samples. The statistical an alyses were carried out for the results of the long range traffic survey more than one year at three congested straits in Japan. The accuracy (standard deviation over sample average) was approximately inversely proportional to the square root of number of days observed. The following guides for the traffic survey were obtained; (1) Considering the traffic survey cost, the observation for successive seven days might be reasonable from the above relation and the weekly trend of traffic flow. (2) In view of the monthly variation, the traffic survey in October seemed to be suitable because of the relatively steady traffic flow.