CPSC Reflector Research Goes Nowhere,
While 3M Co. Threatens Great Changes:

Report of the 16 July, 1997, Meeting
John Forester, 1997

home                   lights

1 Background

The Consumer Product Safety Commission of the United States has, since 1974, controlled the safety aspects of bicycle design through its standard for bicycles. One very important part of that standard is the requirement for nighttime protective equipment, 22 reflective surfaces, but that says nothing about lights at night. The theory of the 22-reflector system is that at any angle there is at least one that faces the motorist. Therefore, simplistically speaking, a motorist at night will always be able to see and avoid the bicycle.

The error in that simplistic reasoning is that the motorist will see a reflector only if there is both a direct line of sight between his eye and the reflector and the strong part of his own headlamp beams illuminates the reflector. In many accident situations the motorist's headlamp beams don't illuminate the bicycle's reflector until it is too late to avoid the collision. Because a bicycle headlamp produces its own light, it can be seen even when the motorist's own headlamp beams are not shining on the bicycle. The CPSC was informed of the inability of any reflector system to perform the required function before it issued its regulation in 1974. It was sued over this issue, but lied to the court and won.

In 1994, the CPSC finally discovered (what everybody else had recognized for decades) that nighttime cycling was considerably more dangerous than daytime cycling. In 1995, the CPSC initiated a Bicycle Reflector Project. In 1996 it publicly described its proposed research plan, which was strongly criticized as having little relationship to real collision situations and none to the known problem accident types, which were largely at intersections. In response to this criticism, the CPSC's engineers showed no understanding of the problem, which had to be explained to them in exquisite detail. The CPSC added a little bit to its plan, and completed its research in December, 1996. In July, 1997, the CPSC held this meeting to discuss its conclusions and recommendations.

2 Research

2.1 Objective

The CPSC says that its intent was to evaluate the adequacy of the current reflector requirements and to investigate the possible improvement to reflectors to increase nighttime conspicuity.

2.2 Method

2.2.1 Literature Research

This purpose required looking at the data on nighttime car-bike collisions to learn the typical accident mechanisms. The CPSC did no original research on this subject; it just used data from others.

Well, what data did it use? It ignored the classic, detailed, and engineeringly and statistically robust study of car-bike collisions that was done by Ken Cross using 1974 data. Instead, the CPSC used the data from the Fatal Accident Reporting System for some recent year(s). This has two systematic errors. The first is that it considers only fatal car-bike collisions, which have a different mix of types and frequencies than do all car-bike collisions. Second, it is merely a compilation of police traffic accident reports without any additional investigation. Cross and his staff had performed detailed investigations into each accident in their study, and these revealed much useful information. The CPSC also used one fact from the University of North Carolina national compilation of police accident reports, that 53% of car-bike collisions involved crossing path, 40% parallel path. To show how useless this information is, it doesn't divide the parallel-path collisions into those where the cyclist swerved in front of the motorist, those where the motorist drove into the rear of a lawful cyclist, and those in which the motorist turned left into a cyclist coming from the other direction, all of which are extremely important distinctions when analyzing the effectiveness of nighttime protective equipment. For that matter, the CPSC could have examined my analysis of Cross's data that shows the proportions of car-bike collisions that are probably caused by darkness.

One might argue that Cross's data are old. However, the mechanisms of collisions have not changed much since 1974, and it is far better to have good data to analyze, of any reasonable date, then to try to get facts from data that don't contain many useful ones at all.

2.2.2 Engineering Analysis

The CPSC presented an analysis of the sequence of mental events in a collision between a cyclist and a motorist. The sequence of events for the motorist was presented as Search, Detection, Evaluation, Decision, Action, Vehicle Response, Collision either incurred or avoided. The cyclist's chain of events was the same, except that the CPSC forgot to include Vehicle Response, as if the cyclist were a pedestrian. The CPSC said that only if both parties failed along the line would there be a collision. I said that that was not so, that one had to consider the right-of-way. As I explained it, if a cyclist swerves in front of a motorist and gets hit, that is not a failure in perception on the motorist's part, and if a motorist restarts from a stop sign and hits a cyclist on the protected arterial, that is not a failure of perception on the part of the cyclist. There was loud disagreement. If the motorist fails to anticipate the cyclist's swerve, that is a failure on the part of the motorist, so they said. There were statements that the motorist should slow down and leave swerving room. Such statements have two repercussions.

The first concerns the effectiveness in accident reduction. The CPSC standard covers only the conspicuity of the bicycle. Changes in that conspicuity will most affect accident types in which the motorist has to observe the cyclist and yield to him. In that situation, failure of the bicycle's conspicuity will cause an accident; in those types of collision, bicycle conspicuity is vital. However, changes in the bicycle's conspicuity are not likely to have much effect in accident types where the cyclist has to observe and yield to the motorist. The motorist often has insufficient time to avoid the accident even in daylight; in those types of collision, bicycle conspicuity is of little importance. Whenever evaluating the effectiveness of an accident reduction system, much more attention should be paid to those accident types in which changes to the system might have large consequences, rather than to those accident types in which the system itself is rather inconsequential.

The second repercussion is that this is another example of the social image of the cyclist as irresponsible child, which has consequences for vehicular cyclists.

All in all, the degree of analysis was shallow.

2.2.3 Field Measurements

The CPSC planned to measure the conspicuity of different reflector systems to motorists at night, and use this data to develop appropriate recommendations. For this test plan to produce reasonable results, the conspicuity measurements must be made in situations that simulate impending car-bike collisions that the motorist would be able to avoid if he sees the cyclist, and it would be beneficial if the tests simulated the known difficult accident situations. Whether the CPSC so planned it or was the beneficiary of unplanned stupidity, the tests failed to demonstrate any inadequacy in the CPSC's all-reflector system. The Motorist-Overtaking Tests

For this test the CPSC had motorists drive around a private road system and speak out whenever they saw something that might be in the roadway. They used a car with modern headlamps operating on low beam. Along the roadways were bicyclist dummies, pedestrian dummies, construction barricades, barrels, and a few unplanned deer. The bicyclist dummies were set up on bicycles with the wheels and cranks turned by electric motors to represent 10 mph pedaling. (Gear unknown) The bicyclist dummies differed in their equipment. Some had standard CPSC equipment. Some had white instead of amber pedal reflectors (1.6X brighter). Some had amber instead of red rear reflectors (3X). Some had a larger red reflector (3.5X). Some had very large lime-green reflectors on rear and pedals. Some had blinking rear lamps. The sequences were randomized and the usual precautions were taken.

The distance at which the item was detected, and the distance at which it was recognized, were measured. Except for the red light, all items were detected at 700 feet or a bit more, and all items were recognized at 650 feet or a bit more. The red light was detected at about 840 feet. The Intersection Collision Test

The intersection collision test was said to measure the effectiveness of side reflectors to prevent collisions. The car traveled toward an intersection at about 20 to 25 mph, while the bicycle with dummy rider traveled from the motorist's right, at 7 mph, also toward the intersection. The timing was set so that the two would arrive at the intersection simultaneously. The cyclist was stopped just short of the intersection, so that no actual collision was possible. The dummy cyclists had different items of equipment. Some had standard CPSC spoke reflectors, some had two on each wheel. Some had ring-shaped reflectors. Some had the large lime-yellow reflectors and reflective tires. Some had a 2.4 watt halogen head lamp, some had a blinking white LED headlamp. The test results all showed that detection and recognition occurred simultaneously for each type of equipment, and that the differences between the types of equipment were not significant.

2.3 CPSC's Conclusions

The CPSC reached the following significant conclusions.

1) "Significant improvement to CPSC reflectors is not practical."

2) "Red blinking tail light could increase rear detection significantly." By this the CPSC meant that detection would occur earlier, at a greater distance, but recognition would not occur earlier.

3) The originally offered conclusion was: "Reflectors or lights tested were not effective for side detection or recognition." It was recognized that these words did not express the intent. The intended thought is that because the additional reflectors or lights that were tested did not improve the performance of the CPSC wheel reflectors, it is impractical to improve on existing wheel reflectors.

4) Not formally listed in the conclusions was the CPSC's expression that it thought that the bicycle light manufacturers should cooperate in producing a voluntary standard for front, side, and rear lights.

3 3M Company's Threatening Proposal

The lime-green reflectors that were used in the CPSC's tests were samples submitted by 3M. The material used is a 3M reflective sheet product that is colored with lime-green fluorescent dye. The result is an item that both shows brightly in sunlight, when the fluorescent dye is activated by UV light from the sun, and gives good retroreflectivity, when illuminated by cars' headlamps, which contain substantially no UV light. The reflectors submitted by 3M were very large. The front and rear reflectors were about the size of a box 6 inches on a side. Each reflector was made of three panels, the center one across the front and the side panels bent back but not quite at a right angle. The pedal reflectors were bands that looked more than an inch in height, running around the perimeter of the pedal except adjacent to the crank.

3M showed graphs comparing the retroreflectivity of bicycles equipped with the standard CPSC reflectors against bicycles equipped with this system plus reflective tires. The performance of the CPSC reflector system will be described below. The performance of the 3M system was both stronger and more uniform around all angles, producing an elliptical plot, with the greatest reflectivity to the sides, just as with the CPSC system. Despite having two to three greater reflectivity than the CPSC system, it showed no greater detection and recognition distances than the CPSC system in the CPSC tests.

The 3M Co. says that if the CPSC does not revise its standard to require the performance of the 3M system, it will make a formal legal petition to have the standard changed in that way. One can certainly say that, as far as retroreflective optical performance is concerned, the 3M system is superior to the CPSC system. The question is whether retroreflective performance is the true measure of the effectiveness of nighttime protective equipment.

4 Discussions About Bicycles and Cycling

There were three sets of discussions about reflector systems, but they really discussed much the same subjects. The first discussion was stimulated by the CPSC's analysis of the mental part of the collision avoidance process, the second by consideration of the proper rear reflector, the third one by 3M's presentation.

The first discussion concerned the question of who avoids a collision. The general consensus was that both parties must avoid the collision, that if either failed in this duty a collision would occur. I stated that this failed to consider the question of right of way. I pointed out that in a collision avoidance situation, the first responsibility, and the greatest and easiest ability to prevent the collision, lay with the driver who was obliged to yield the right of way. If that driver fails to yield the right of way, it is much more difficult for the driver who expects the right of way to avoid the collision. If the conspicuity of bicycles is to be the variable factor, then we need to examine the situations in which the motorist must yield the right of way to the cyclist, because these are the situations in which an increase in the conspicuity of the bicycle might produce the greatest reduction in accidents.

Many voices objected to this point of view. They thought that both parties to a collision should be considered equal in responsibility and in ability to avoid the collision, and therefore the bicycle's equipment should be evaluated without regard to right of way. This is contrary to both fact and to traffic law, and nobody offered any argument supporting their side of the controversy. I think that they want a physical system, preferably reflective, that is intended to protect cyclists who violate the traffic laws. Whether such a system can be devised is unknown, but we do not have such a system today and we don't know how to design one.

The second discussion concerned the recognizability of bicycles. The theory is that if a motorist sees something that he recognizes as a bicycle he will take special care to protect it. This discussion arose with respect to the rear reflector. As I said, all that the motorist has to do is to steer clear of whatever it is that he sees ahead, and all the cyclist has to do is to continue cycling along the road and it doesn't matter whether or not he sees the motorist. Oh, no, said many voices: the motorist has to be able to know that what he sees ahead is a bicycle. "What," I asked, "Will the motorist do that is different from what he would do if he thought that this was a parked car?" Somebody suggested that if the motorist knew that this was a bicycle ahead instead of a parked car, he should slow down and give it more room. "Why should he do that?" I asked. No real answer. In the end, although they would not say so, it was obvious that the only thought that they had in mind was that the cyclist might swerve in front of the motorist and get hit, and that the motorist wanted some warning that this might happen.

3M's presentation stimulated more discussion. 3M said that the object of the fluorescent color in the reflective material was to make the bicycle more conspicuous in daylight, and to do so with an unmistakable pattern, or signature. That meant that the equipment would legally have to be used even in daylight, no matter how much it interfered with the operation of the bicycle. The front and rear reflectors would interfere with handlebar bags, saddlebags, and rack loads, all of which would also obscure the reflectors. The high pedal reflectors wrapping around the pedals would require that pedals be one inch or more higher than at present to keep the same ground cornering clearance. The wind resistance of the front and rear reflectors would be considerable. If reflective tires became a legal part of the 3M system, as 3M would like, then cyclists would have to put up with reflective tires for all purposes, although reflective tires have not been developed for all cycling purposes, and reflective materials might significantly detract from the performance of the better tires.

In support of their proposal, 3M made several serious errors. They asserted that lack of daytime conspicuity was the cause of a large proportion of car-bike collisions. I broke in, loudly, to say that Cross's data showed that the proportion was only six percent. 3M asserted that the problem was during twilight, stating that there was a large peak of car-bike collisions over the hours of 3 and 6 pm. That data item, of course, did not demonstrate anything of the kind. That peak is far more the result of the amount of cycling going on than a result of twilight, which is only a short time, and in summer is after that timespan.

3M asserted that its fluorescent and retroreflective material worked equally well in daylight and in darkness. I doubt that assertion, because the fact that near twilight the light reddens indicates that the blue has been lost, and if the blue has been lost the UV, which is beyond the blue, surely has had an even greater reduction. Without the UV, the 3M material will not fluoresce. 3M said that no objections will be effective because this material will be required by law, supporting this by presenting a pedestrian jacket made of this material with the assertion that Minnesota law requires such jackets. It does, but only for road workers who have to walk about in traffic.

3M asserted that bicycles had to look unlike motor vehicles so that they could be easily distinguished from motor vehicles, which is what their proposed equipment does. 3M argued that all-around reflectivity was very effective in reducing nighttime car-bike collisions and didn't require the cyclist to realize that his nighttime equipment had to be turned on. 3M remarked that whether bicycles are regarded as toys or as vehicles, their all-reflector system is the equipment of choice, and made several remarks about protection against erratic operation by child cyclists.

In response to much of this, I told the CPSC that if it chose to adopt the 3M system for the reasons that 3M gave, then it would have to go back to its original legal authority and original proposal for this standard, and limit this standard to bicycles actually used by children, and produce a different standard for bicycles that are intended to be used by adults.

5 Presentation of Various Lights

In the afternoon there was a discussion of bicycle lights. Cateye made the major presentation, showing the slight differences that were required to produce lights for different nations. Cateye did not emphasize any proprietary information, and openly wished that some others in the lights business had been present. The CPSC obviously doesn't want to get into the business of establishing a standard for bicycle lights, but it offered to provide information to any industry standards group, such as ANSI, that decided to establish a voluntary standard. for front, rear, and side lighting.

6 Criticism of the CPSC's Research and Conclusions

6.1 Are The CPSC's Conclusions Valid?

The CPSC's conclusion that it is impractical to improve on its reflectors should be divided into two parts, consideration of the rear reflector and consideration of the side reflectors (wheel reflectors).

6.1.1 The Rear Reflector Conclusions

The CPSC found that adding brightness to the rear reflector did not increase either the detection or recognition distances, which were about 700 feet with low beams. That is about 16 seconds at 30 mph, which is an appropriate speed for low beam driving. That in itself is fine. However, the conditions of the test were ideal. Good headlamps, clean windshield, alerted driver, no other traffic, etc. Several experts said that the results should be considered only comparative, not absolute. However, there is one other consideration. It may be that with that car the 700 feet was as far as the low headlamp beams would illuminate a reflector on the rear of a bicycle. Therefore, there was little difference in distance because no reflector, however reasonably bright, would shine at all at any greater distance.

This test procedure eliminates the problems of both poor visibility and traffic confusion, and the fact that we know that sometimes an occasional cyclist is not seen. It is reasonable to conclude that a brighter image at the same distances would be more likely to be seen. So the appropriate test would have been more like the real-world situation in which the relative brightnesses of reflectors at ranges from far distant to quite close-up were observed under more typical conditions. I have done that, and the 3-inch amber SAE reflector wins hands down. That reflector is easily available in auto parts stores, and the CPSC knows that. The CPSC engineer said that since they couldn't test everything they chose not to test that one. Why not test that which has been proclaimed far and wide as the best rear reflector that is easily obtainable? One explanation is that it would show up the CPSC reflectors as bad choices. What other reasonable explanation is there?

6.1.2 The Front and Side Reflector Conclusions

The intersection test of front and side reflectors is designed to be the easiest for the side reflectors to pass. The motorist is traveling fast, the cyclist is traveling slowly, from the motorist's right, toward an intersection at which they will arrive simultaneously. This puts the cyclist only a slight distance to the right of the motorist's path, and at a slight angle, said by the CPSC engineer to be from 15 to 20 degrees. That is well within the headlamp beams of the motorist's car. Naturally, the wheel reflectors of the bicycle showed up well as soon as they entered the headlamp beams, which spread usefully for about 20 degrees to the right, giving the motorist about 3 seconds to do something about the situation.The test basically could not fail, at least given its presumptions. Example 1: Uncontrolled Intersection

Let's examine a more real-world example. If the intersection is an urban uncontrolled intersection, at which right-of-way is determined by first reaching it, the appropriate speed for the motorist is 15 mph. The speed of the cyclist might also be 15 mph. Then the cyclist will always be at an angle of 45 degrees to the motorist's path until the collision, far out of the headlamp beams until very late. Example 2: Cyclist Required To Stop At Intersection

At intersections at which the cyclist is required to stop, the motorist is allowed to be driving through intersections at 25 mph. The intersection is then a controlled intersection, with either a traffic signal or a stop sign in the motorist's favor. The cyclist then is obliged to stop. The motorist will not change course or speed until he observes the cyclist not stopping. That will be when the cyclist is very close to the intersection, and so is the motorist, so that there is very little time for the motorist to avoid the collision. Sure, the motorist may see the cyclist running the stop sign at full speed, but, just as in daylight, if the motorist is so close to the intersection that he might hit the cyclist, he will do so because there is insufficient time to avoid the collision. Example 3: Motorist Required To Stop At Intersection

At other intersections, the motorist might be required to stop at a stop sign. The motorist is entitled to restart whenever he sees that no traffic is coming. His headlamp beams are shining across the intersecting road, not along it, with a bit of a rightward bias because they are on low beam. If the cyclist is coming from the motorist's right, the motorist won't see him until the cyclist enters the headlamp beams. If the road is 40 feet wide, that won't be until the cyclist is less than 15 feet from the motorist's path. The motorist starts out because he doesn't see the cyclist, then sees him, if at all, immediately before the collision. If the cyclist hits the side of the car, the motorist will never have seen him at all. If the cyclist had been approaching from the motorist's left, which is the dim side of low-beam headlamps, the case against reflectors is even stronger because the motorist can never see the cyclist at all. Example 4: Motorist Left Turn

When a motorist intends to turn left, he is obliged to yield to traffic coming toward him. If he sees nothing he turns left and finds a cyclist crashing through his windshield. That is because the motorist's headlamps, on low beam, do not illuminate the front reflector of cyclists approaching from the other direction. Such cyclists are in the very dim, leftward, part of the low headlamp beams.

6.1.3 The All-Around Reflector System

The CPSC's analysis and test program never questioned the actual need for the all-around reflector system. The rear reflector, in particular, is weakened by the requirement to reflect at angles far away from directly behind. The design to produce reflectivity in directions from which overtaking motorists don't come weakens reflectivity in the direction from which overtaking vehicles do come. The 3M Company presented polar graphs of the degree of reflectivity all around the circle. The graphs look rather like a duck flying in the same direction as the bicycle. To each side there are long, sharply pointed wings, indicating that the wheel reflectors have high reflectivity coefficients. To the front is a shorter neck, indicating that the front reflector is about as bright as each of the two wheel reflectors, while to the rear is a tiny tail, showing how dim the rear reflector is. In the vees between each wing and the neck are small bumps, indicating some reflectivity at 45 degrees from straight ahead, while the similar ones to the rear are practically invisible. In short, the CPSC's requirements hardly carry out the all-around reflectivity model that they were intended to carry out.

6.2 Did the CPSC Conduct a Reasonable Research Plan?

The CPSC's all-reflector system has been criticized for decades. The examples of the motorist-at-stop-sign and the motorist-turning-left types of car-bike collision were submitted to the CPSC at the time of its initial proposals about 1972. While the existence of these types of car-bike collision was recognized at that time, their proportions among nighttime car-bike collisions was not known. My analysis of car-bike collisions probably caused by darkness, drawn from Cross's data sheets, was published in 1977 in Cycling Transportation Engineering, and reprinted in Bicycle Transportation in 1983 and 1994. This analysis is given in Tables 1 and 2.:


Table 1: Relative Proportions of Car-Bike Collisions Due to Darkness

Type of Collision  %
Motorist exiting from side street 47.2
Motorist turning left 22.3
Motorist overtaking cyclist 21.0
Wrong-way cyclist head-on 9.5

Any research program that is intended to improve an existing safety system should be directed at the system's known problems and failures. Depending on the results of the research, they can then be used to either improve the system or to demonstrate that this particular system is not suited for this application.

Table 2: Relative Importance of Nighttime Protective Devices

Type of Device  % of Collisions
Due to Dark  All 
 Headlamp 79 2.8
Rear reflector 21 0.9
Front reflector substituting for headlamp 9.5 0.39
Side reflectors 0 0

The CPSC did not follow this procedure. Instead, it originally designed a test procedure that its system was practically certain to pass. If that test disclosed possible improvements, those improvements could be easily implemented without disturbing either the physical or the intellectual bases of the system. When the original test procedure was strongly criticized for not addressing the known defects of the system, which were at intersections, the CPSC designed an added intersection test procedure that simulated the easiest to pass of all the possible intersection accident situations. Then, when the expected satisfactory results were observed, the CPCS was able to announce that its regulation did not need any improvement.

It is reasonable to inquire whether the research program was consciously designed to achieve this result, or whether the CPSC just happened to find data to support its happy conclusion through plain dumb luck. There are few enough data items on which to base a conclusion. On one hand, one can argue that given the possible tests that would have shown the CPSC's all-reflector system up for what it is, dangerous and seductive, and given the knowledge of the problems had been published in volumes that the CPSC's engineers should have read, it took conscious design to steer clear of the problems and devise a test program that best avoided the possibility of any adverse results. On the other hand, there is the continuing evidence of the incompetence of the CPSC's engineers, both in the past and in the present, who showed no understanding of the problems, either initially or after these problems had been explained to them in exquisite detail.

There is the possibility of a compromising position. The engineers may be so imbued with the popular view of cycling's dangers (the cyclist-inferiority superstition and phobia) that, despite being normally competent engineers in other fields, they find it very difficult to understand the real dangers of cycling. Their views may also be limited by the typical view of cyclists as children, given that the CPSC defines bicycles as "toys or other articles intended for use by children," and the standard legally covers only bicycles that are intended for use by children (even though it also covers all bicycles, on the theory that any bicycle can be used by a child).

Lacking any definitive documents or other direct evidence, I cannot decide one way or the other. Either the CPSC's engineers are very incompetent, or they are very, and unethically, devious, or some combination of the two.

6.3 Reasonable Conclusions About the CPSC's All-Reflector System

6.3.1 General

The reasonable conclusion about the CPSC's all-reflector system is that it is just as dangerous as we always said that it was. The test program has done nothing to demonstrate that the dangers are either less than we had predicted, or have been or will be in any way ameliorated as a result of the tests.

6.3.2 Rear Reflector

The CPSC's conclusion that no practical improvement in rear reflectors exists applies only to ideal viewing conditions. Under these conditions, the CPSC rear reflector is visible at the maximum range of low-beam headlamps and no greater brightness is required. However, that says nothing about whether a brighter reflector would be more likely than the CPSC reflector to be observed under less than optimum conditions, which are frequent in the traffic environment. The SAE amber reflector is about ten times brighter than the CPSC rear reflector, and this additional brightness might well be desirable. This reasonable conclusion is not affected by the results of the CPSC's tests.

6.3.3 Front and Side Reflectors

The CPSC concluded that additional lights and reflectors did not increase the detection distance over wheel reflectors in those situations in which the wheel reflectors were illuminated by the motorist's headlamp beams. In short, one reflector is as good as another in cases of collisions on intersecting paths. However, the test did not compare reflectors against bicycle headlamps in the frequent collision situations in which we know that reflectors cannot work. The reasonable conclusion that bicycle headlamps are required to provide conspicuity in cases of potential collisions on intersecting paths, or when the motorist is turning left, is unaffected by the results of the CPSC's tests.

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