Child Cyclists in Action?

Review of: Children's Perception of Gap Affordances: Bicycling Across Traffic-Filled Intersections in an Immersive Virtual Environment

By: Jodie M. Plumert, Joseph K. Kearney, James F. Cremer; Child Development, July/August 2004, Volume 75, Number 4, Pages 1243-1253

Reviewed by John Forester

home                  social


This study purports to detect some age-related developmental characteristics as exhibited by comparing one traffic behavior of adults and children. aged 10 and 12. However, the measurements made do not demonstrate such, because the investigators failed to consider both elementary principles of physics and standard traffic-cycling behavior.

The authors do admit that "the results of such experiments [virtual reality] are of questionable value if virtual environments lack ecological validity." That is the first factor on which I fault the investigators; the virtual environment did not match the real environment.


The investigators measured the performance of cyclists, of different ages, yielding to traffic at stop-signed intersections. The subjects rode a bicycle mounted on a training stand, instrumented for steering angle, hand brake application, and apparent road speed as shown by rear-wheel-tire speed. The test bicycle was mounted inside a rectangle composed of three screens at right angles. On these screens were projected the traffic scene, composed of 6 equivalent stop-signed intersections, spaced at an apparent block apart. At each intersection, a continuous stream of car traffic came from the left, at random between-car intervals ranging from 1.5 to 4.0 secs (in 0.5 sec modules). The speed of traffic was either 25 or 35 mph, in two groups, and some subjects had 25mph traffic first, others the opposite. The scenery picture moved toward the cyclist according to the speed of the bicycle's rear tire, as generated by the pedaling speed, and as directed by the bicycle's steering angle.

The test equipment had several defects. Since the test bicycle had hand brakes, one has to assume that it had a freewheel, although no such item is listed. (Hand-braked bicycles with either fixed gear or coaster brake are uncommon.) There is no listing of inertia loading of the test bicycle, although there is one intriguing comment. "Relative to adults, children had more difficulty in getting the bike started (despite the fact that the bicycle offered little resistance)." This statement suggests that there was no inertial loading. But maybe there was, and since there is no statement about adjusting the inertia load to match the mass of the cyclist, we must assume that if there were any inertial load, that load was constant for all masses of cyclist. Also, the test bicycle provided no inertial response to the subject as the subject applied the brakes, so that the braking action had to be controlled only by the changes in the apparent movement of the scenery, as produced by some calculation between the force at the brake levers and apparent deceleration. Since the bicycle was mounted on the training stand, there was no real stopping action, in which the cyclist would have to slide off the saddle and put foot on ground as the bicycle came to a halt, and no real starting action, in which the cyclist would rotate the pedals to starting position, stand on the proper pedal, slide back onto the saddle, and resume pedaling with both feet. The consequences of these defects will be discussed at the appropriate points.


The experimental procedure was also defective. The actions of the cyclist were filmed and combined with the projected scenery. These images were then viewed and scored. However, the only data that were recorded for evaluation were time data, no distance data. There was some attempt to score cyclist behavior, because coming to a full stop for at least two seconds scored 1 point. However, it appears that the investigators realized that no such scoring system would work, for the point score probably was used only as a condition datum to be used in ANOVA calculations. The subjects were instructed to ride along the virtual street, to stop at each intersection, then choose to move across at the first gap in traffic they thought suitable.

The subjects refused to obey the instructions given them. Instead, they probably reverted to their normal behavior under the supposed traffic conditions depicted. That is, the better informed or more experienced subjects slowed when approaching the stop sign, delaying until either they had to stop to avoid a collision or a usable gap approached to enable them to cross. In this respect the more skillful cyclists were given a demerit point for not stopping, relative to the less skillful cyclists.


All data items were times, no distances. One data item was the duration of the stopping time before choosing to cross. Another data item was the relationship between the length of the time gap between following vehicles and the decision to cross. No significant age differences. All age groups started to move after stop, or started to increase speed if they didn't stop, with the same total safe crossing time. All age groups crossed the traffic lane in the same time (the authors don't say this, but subtracting the average safe time after crossing from the safe time before starting to cross produces that value). Therefore, the average speed while actually crossing the traffic lane was the same for all age groups. Another data item was the clearance time between the rear wheel leaving the traffic lane and the front of the car reaching that point. Younger children had shorter average clearance times, which surely would be a safety consideration.

The authors investigated the difference in waiting time before moving to cross the intersection. Fortunately, they make no claim about age differences, but it should have been obvious that this time depends largely on the probable time before a usable gap in traffic approaches, which is entirely outside the cyclist's control, being determined by the combination of gap duration probabilities chosen by the investigators with the computer randomization that selected gaps according to those probabilities.

The authors investigated the cause of shorter average clearance times for younger children, but the data do not make sense. So here is the peculiarity. The younger subjects take longer time to accelerate to the same traffic-lane-crossing speed as the older subjects. If they took longer time to accelerate to the same top speed, then they took more distance. But the distances are not recorded. Does this mean that the younger subjects stopped further behind the traffic line? Quite possibly, but the authors don't mention this possibility. Instead they claim that the younger subjects had difficulty starting. That this is a false claim is demonstrated by the time interval taken from starting to move to entering the traffic lane; surely difficulty in starting would have delayed the start of movement and been outside this interval. Instead, the authors claim that their investigation reveals some characteristic of mental development that occurs after the ages of 10 to 12 years. That claim is not demonstrated by the data.

The authors also devote discussion to "time to get started." They state: "Clearly, the answer lies in the additive effects of taking longer to get started and taking longer to reach the roadway." However, the duration required to start in motion cannot be defined. The end, obviously, is the first visible sign of motion; but when was the start of this interval? In the typical race, say in swimming or track, the start of this interval is the crack of the starter's gun. But in this investigation, there is no such event. The cyclist sees a usable gap in the approaching traffic and either starts in motion or speeds up in order to use that gap. But there is no means of telling how far away that gap was when it became apparent that it would be usable by a "standard" cyclist, and it is equally undefinable for this particular subject cyclist. The claim that evaluation of time to get started is a developmental characteristic within the age range tested is not demonstrated by the data because that duration cannot be defined.

Defects of Study

 I offer some suggestions as to the defects that render this study useless.

The first defect is the disjunction between the investigators' assumptions about cycling and actual cycling practices. If the investigators wished to determine the behavior after stopping, then they should have insisted that all cyclists stop. That might have been an easier investigation, also, for there would be no necessity for the preceding apparent approach to the intersection; just repeat the action with the same scenery but a different mix of gaps. If the investigators had wanted to see how cyclists actually behave at stop signs, then they should have just let them act normally without issuing demerit scores for not stopping.

The second defect is the failure to adjust the inertial loading to suit the moving mass. The investigators show no awareness of the effect of inertia upon acceleration. It is possible that the subject bicycle had no inertial loading. In which case, if I had been one of the subjects, I would have had that scenery flowing past me at 25 mph in half a second from a dead stop. It is equally possible that the bicycle did have inertial loading, but that loading was not adjusted for the magnitude of the moving mass, that is largely determined by the mass of the cyclist. That would explain the longer time that young cyclists took to accelerate to crossing speed, for they would exert less force relative to mass than would stronger adults.

The third defect is the failure to consider, or even to list, the gear(s) used. It is reasonable to assume that, when crossing one lane of traffic, the cyclist is unlikely to change gear. However, the relationship between the gear used and the inertial load is vital for the cyclist in evaluating the performance of him on this test bicycle. If this relationship differs too far from that of the subject's own bicycle and cycling practices, the result is going to be unreliable.

The fourth defect is the failure to investigate and record the past cycling practices of the subjects. An item that I think would likely be very significant is the type of braking system of the bicycle typically used by the subject, be it coaster-brake or hand-brake. The braking action was not recorded, but differences in habit might affect the probability of stopping. Certainly, the differences in habit would affect starting, which was an action that was recorded and considered. Cyclists with hand-braked bicycles reposition the pedals for an easy start immediately after stopping; cyclists with coaster-braked bicycles cannot do this.  A subject with coaster-brake habits would be adversely affected in this investigation relative to a person with hand-brake habits.

Some persons who have read only the abstract, with its emphasis on time to get started, have commented that the study has the value that it informs us of the necessity of teaching child cyclists the proper methods of starting and stopping. That's irrelevant, because we already know that, when starting a cycling class for adults from the general public, we have to start by teaching starting and stopping. If we have to teach adults the skill, then surely we have to teach children the same skill.

In short, the study does not demonstrate the conclusions made by its authors. This is because of poor understanding of concepts of elementary physics and of traffic cycling procedures.

Return to: John Forester's Home Page                                 Up: Social