American Scientist

To the Editors:

I was excited to see your recent coverage of the traffic-signal change interval, the dilemma zone, and other aspects of traffic-signal timing (“Traffic Signals, Dilemma Zones, and Red-Light Cameras,” Engineering, May–June). Thirty years ago, I was right in the middle of these controversies, having written a peer-reviewed paper on the subject of vehicle and pedestrian signal timing and being appointed as chair of the technical committee working on these topics while I was a traffic-signal engineer in Athens, Georgia. I had moved there to attend the University of Georgia and earn a bachelor’s degree in what we now call “Big Data,” which was substantially based on using probabilities to estimate human behavior.

I immediately recognized that all traffic-signal timing should be based on probabilities, but soon found that engineers greatly preferred deterministic solutions. As a result, we were living with rules of thumb, engineering judgment, and fixed formulae rather than looking directly at the problems of human behavior. Traffic engineers accepted early the role of probability in the mechanism for determining speed limits, which utilizes the 85th-percentile free-flow speed observed on the roadway. This approach is based on the premise that widespread flaunting of the law begins when more than 15 percent of the public violates it. Most states codified this means for determining traffic speed limits in their statutes.

Although the yellow interval timing approach similarly uses the 85th-percentile speed of approaching traffic, codification of this element of the calculation has not occurred. In fact, as Dr. Petroski found, the states do not even agree on what the yellow interval is supposed to accomplish or what driver behavior, upon seeing the yellow indication, should be. Implicitly, the methodology accepts the possibility that up to 15 percent of vehicles will enter the intersection after the red indication has begun; however, I don’t feel that anyone actually believes such a result is acceptable.

In 1982, the deceleration rate used in the calculation of the yellow change interval was 15 feet per second squared, and the reaction time was 1 second. The committee changed the deceleration rate and reaction time to 10 feet per second squared and 1.5 seconds, respectively. We felt these values were more reasonable because they were based on actual driver observation; however, the behavior on which we based these choices was limited to the vehicles that chose to stop. There was no way to judge the reaction time of drivers who chose not to stop simply by observing traffic.

Jack A. Butler
Orlando, FL