This method distributes the counting days uniformly over the entire year, but evidently the expense would be considerable unless it can be done by the regular patrolman. For the purpose of laying out a system of highways quicker results might be desired. In Massachusetts, Connecticut and other states two counts were made about two or three months apart. If one were taken in July, say, and another in October and the results averaged it is thought they might fairly represent the year. Each period could cover an entire week, thus giving an average of 14 days during those seasons of the year when traffic might most nearly represent the normal. Blanchard’s method[177] contemplates taking the census in four periods of three days each, one in April, May, or June, one in July, one in August, and one in September or October, as local conditions may dictate. The days selected should include Sunday in order to include the heavy automobile traffic on that day. If further information is desired additional three-day periods could be taken during the winter season. Most of the authorities seem to think that the daylight hours from 5 A.M. to 9 P.M. are sufficient. But in the light of the tests being conducted by the Illinois Highway Commission which show alternate convex and concave warping of pavements from day to night it might be well to take account of night traffic. Some of the latest censuses, namely Iowa and Connecticut, placed scales on the highways and actually weighed the vehicles. In Connecticut road scales were used which weighed the individual wheel loads. The observers are supplied with cards on which is printed the classified list of vehicles and animals likely to pass with columns for tallying them during the separate hours. The in-and-out-of-town vehicles are recorded separately, and, if actual weights are not taken, whether loaded or unloaded. Information relative to the weather and condition of the roads is also noted, and there are blank spaces for the station, the date, and the signature of the observer.