How accurate are witness distance estimates given in car lengths?
Robins [1] observed that “in spite of the many converging cues available to humans to judge the separation distances and speeds of other objects, we perform poorly at such tasks.” Similarly, Olson [2] observed that “witnesses’ judgements of distance in the usual measuring units are likely to be unreliable.” Despite the limitations of humans to judge distance, witnesses and drivers involved in car crashes are often asked to use the unit of a car length to estimate how far they were from another vehicle at some point prior to a collision. There seems to be a presumption that reformulating the question in car lengths somehow improves a persons ability to judge distance. Accident reconstructionists sometimes use such car length distance estimates to perform calculations related to following distance or accident avoidance.
In 2016, Neal Carter and I published a study in Collision (Volume 11, Issue 1) examining the validity of using distance estimates given in car lengths to perform accident reconstruction calculations. This study utilized 20 volunteers who were asked to drive their own vehicle through a standard and predetermined route on a mixture of highway and city streets. In the course of the drive, subjects were asked to estimate their distance from a vehicle driving ahead of them in an adjacent lane. One of us rode along with them and posed a series of standard and predetermined questions to each subject. A second vehicle, which was driven by another one of us, was utilized as the lead vehcile to which to the test subjects would estimate their distance. While driving the route, each test subject was asked to make 5 estimates of their distance behind the lead vehicle. Four of these estimates occurred on the highway and one on a surface street. A sixth estimate was elicited at the end of the route, when the subject was still seated in the car, but the car was stationary.
At the beginning of testing with each subject, the subject was asked for their current age, the age at which they obtained their driver’s license, and their occupation. During the highway driving with each test subject, the lead vehicle maintained a relatively constant speed (57 to 60 mph) and the test subject drove behind the lead vehicle, in the adjacent lane to the left. This lane offset was chosen so that a range of distances could be evaluated, without having the test subject directly behind a vehicle at close distances. At four different points along the highway driving portion, the researcher who was riding along with the test subject asked them to verbalize their current speed and then to estimate their distance behind the lead vehicle, in units of car lengths. As they gave their distance estimate, the researcher documented their estimate and also recorded the actual distance to the lead vehicle, in units of feet. The actual distance was measured using a handheld laser rangefinder with an optic sight (TruPulse 200). The manufacturer of this rangefinder reported an accuracy of plus or minus 1 foot and a range of 3,280 feet.
Test subjects were not offered any definition of the term “car lengths” nor were they provided any instructions around how they should define that term. Also, the subjects were not given any feedback about their estimates and they were never told the actual following distances measured by the rangefinder. This was to ensure that their estimates would not evolve due to feedback during the testing. Once the driving was completed, each subject was asked to define a car length in whatever units they preferred.
Car Length Defined as 16.5 feet
For the purpose of analyzing the test results, we initially chose to define a car length as 16.5 feet. Our results and ultimate conclusions are not sensitive to this definition, however. In only 3 instances did the subject overestimate the distance. In 97 out of 100 instances, the subjects underestimated the distance. The actual distances were as much as 6 times longer than what the subjects estimated. The average ratio of actual distances to estimated distance was 2.27, with a standard deviation of 0.96. The most frequent estimation error (35 instances) was for the actual distance to be between 1.5 and 2 times larger than what the subjects estimated. More than half of the distance estimates (51 instances) were underestimated by more than a factor of 2. Based on this data, if a witness estimates they were 3 car lengths behind another vehicle, there is a more than 50 percent probability that there were actually more than 6 car lengths between their vehicle and the other.
Car Length Defined by Subjects
One could reasonably ask the degree to which this outcome is dependent on our definition of “car length.” Test subjects’ estimates of of a car length ranged between 5 feet and 45 feet, with an average of 14.5 feet and a standard deviation of 9.1 feet. We found that even when the subjects set their own definition of car length, in the majority of cases they still significantly underestimated the actual distances. More than half the distance estimates (62%) were underestimated by more than a factor of 2.5. 100% of the estimates were shorter than the actual distance.
Conclusion
The findings in this study suggest that when witnesses and drivers involved in car crashes are asked to recall and estimate how far they were (in car lengths) from another vehicle, accident reconstructionists and lawyers can reasonably assume that the provided estimate is inaccurate. This finding should not be interpreted as justification for dismissing a witness’s description of an accident altogether. A distinction should be drawn between the story a witness tells and the quantitative information they give. Even within the category of quantitative information, reconstructionists should consider what quantitative information may be accurate. For instance, if a witness estimates the speed of another vehicle that they were following, this estimate may be very accurate, since the witness has a device in their car for measuring their own speed (i.e., a speedometer).
The drivers in this study were allowed to focus on the lead vehcile and evaluate the distance before responding with an estimate of the following distance. Witnesses to a crash, on the other hand, are often required to make these estimates based on their recollection of vehicle positions before the crash. One could reasonably expect such estimates to be even worse than those that were recorded in this study.
References
Robins, Patrick J., Eyewitness Reliability in Motor Vehicle Accident Reconstruction and Litigation, Lawyers and Judges Publishing Company, 2001.
Olson, Paul L., Farber, Eugene, Forensic Aspects of Driver Perception and Response, Second Edition, Lawyers and Judges Publishing Company, 2003.
Photo by Alexander Popov on Unsplash