We used computer simulations to create two lanes of traffic with identical characteristics, except that their congestion varied depending on random starting gaps. Vehicles then accelerated if no other vehicle was within the minimum headway distance, otherwise they decelerated. Minimum headway distances increased with higher velocities to prevent collisions.

We evaluated statistically the movements of an individual driver compared with all the others in the next lane. Under baseline conditions, many one-second epochs produced a change in the driver's relative position. Epochs in which the index vehicle was overtaken were more frequent than epochs in which the index vehicle was overtaking another vehicle (Fig. 1). The total number of vehicles that passed the driver was balanced by the total number of vehicles that were overtaken by the driver because of multiplicity in some epochs.

Figure 1: Relation of traffic density to the amount of time spent overtaking and being overtaken.
figure 1

All time estimates have standard errors of less than 5 seconds that decrease with traffic density. For example, a driver travelling for 10 minutes on a roadway with congestion of about 100 vehicles per kilometre would spend 47 seconds being overtaken by vehicles in the other lane and about 35 seconds overtaking vehicles in the other lane if both lanes have the same average speed. Details of simulation models are available from the authors.

Doubling the acceleration generated a larger difference, and halving the minimum headway distance, to represent ‘tailgating’, greatly increased the difference. Reducing the frequency with which the index driver glanced at the next lane from once every second to once every two seconds attenuated the difference. No combination yielded a slower apparent speed for the next lane.

We also videotaped traffic sequences by mounting a camera in a moving vehicle and filming the side-view perspective of the next lane on a congested road. When a section of videotape showing a slightly slower average speed in the next lane was screened to driving students (n=120), 70% stated that the next lane was moving faster and 65% said they would change lane if possible.

From these results, we suggest that drivers are responding to an illusion: namely, that the next lane on a congested road appears to be moving faster than the driver's present lane, even when both lanes have the same average speed. This occurs because vehicles spread out when moving quickly and pack together when moving slowly. A driver can therefore overtake many vehicles in a brief time interval, but it takes much longer for the driver to be overtaken by the same vehicles.

Other aspects of human perception may accentuate the impression that the next lane is moving faster. Differential surveillance can occur because drivers look forwards rather than backwards, so vehicles that are overtaken become invisible very quickly, whereas vehicles that overtake the index driver remain conspicuous for much longer4. Moreover, a driver is more likely to glance at the next lane for comparison when he is relatively idle5 while moving slowly.

Even if attention was not focused in particular directions and was evenly spaced in time, human psychology may make being overtaken (losing) seem more salient than the corresponding gains6. Furthermore, misconceptions about randomness can make runs of overtaking and being overtaken seem unduly prolonged7,8. Our study highlights the effects of congestion and the increasing importance of the illusion, given that the number of miles travelled by vehicles is increasing at a much faster rate than the amount of roadway9,10.