Reviews of:
A Comparative Analysis of Bicycle Lanes vs Wide Curb Lanes: Final Report, FHWA-RD-99-034

Bicycle Lanes vs Wide Curb Lanes: Operational & Safety Findings and Countermeasure Recommendations,

Evaluation of Shared-Use Facilities For Bicycles and Motor Vehicles: Final Report: Florida DOT

home                 facilities

1 Introduction

A study comparing the operational and safety aspects of bicycle lanes versus wide curb lanes! Just what we have been waiting all these years to read. So I set myself to make a thorough review of these apparently important documents. I had downloaded and printed out both, put them into a binder for easy, reliable, and permanent reference, got myself a pad of clean note sheet on a clipboard, equipped myself with black pen and very fine red pen, a small ruler for making neat underlines, and kept my calculator handy. I had plenty of time. The third document I reviewed much later. Its review is after the others.

2 Analysis of Deficiencies

Sure enough, I soon found some questionable sentences to underline for later reference and analysis. Pretty soon it became obvious that with the data the investigators were collecting, they couldn't reach reasonable and useful conclusions. Here are some examples.

2.1 Parked Car Space?

Some bicycle lanes contained lawful parking spaces, some did not. However, the additional average total width for those that did contain lawful parking spaces was given as only 0.2 meter greater than those bicycle lanes that did not. That distance, 0.2 meter, is less than 8 inches. Ever try to squeeze a parked car into less than 8 inches width of space?

2.2 Non-Comparable Sites

The investigation was conducted at 24 different intersections. The investigators divided these into intersections with bicycle lane and those with wide curb lane, and also into intersections with high traffic volume and those with low traffic volume. Here's the breakdown (Table 1):


Sites by Traffic Volume

Description of Facility

# of Sites

Bike-lane, Low Traffic


Bike-Lane, High Traffic


Wide Curb Lane, Low Traffic


Wide Curb Lane, High Traffic


The investigators also gave the average widths for these sites (Table 2):


Sites by Total Width

Description of Facility

Avg total Width meters

Bicycle Lanes


Wide Curb Lanes


In other words, the study compares wide-curb-lanes that are narrow and are on high-traffic streets against bicycle-lanes that are wide and are on low-traffic streets. Pitiful, isn't it?

2.3 Definition of Conflict

The data are grouped into "midblock" locations, between 90 and 150 meters from the intersection stop line, and "intersection" locations, from the 90-meter point through the intersection. The number of conflicts seen on the videotape records was recorded. The definition of "midblock conflict" is "A conflict was defined as an interaction between a bicycle and a motor vehicle, pedestrian, or other bicycle such that at least one of the other parties had to change speed or direction to avoid the other." Presumably, the same definition was used for "intersection conflict," although I could not find such a definition.

Consider the two following situations.

  1. 1: A cyclist is preparing for a left turn and seeks to merge left to the centerline. The motorist behind him slows a bit to let him do it. Maybe, the motorist sees a red light ahead and knows that he will have to stop in any case. Because the motorist slows, this is counted as a conflict, even though it represents the way that traffic ought to operate.
  2. 2: A motorist is preparing to turn right. The cyclist coming up behind him, rather than trying to overtake on the motorist's right, slows down, maybe even merges left around the back of the motorist. That would also be counted as a conflict, again even though it is representative of how traffic ought to operate.

2.4 Merging Across Lanes

One of the maneuvers recorded in the data is "Bike turned or swerved across a traffic lane." So far as one can tell, this is always coded as a bad action. There is no distinction between doing this properly and doing it improperly. Of course, the observers reading the videotapes probably can't tell the difference; the action could be up to 330 feet away, with possible intervening traffic. At that distance, and with that direction of view (backward from the sidewalk, along the traffic lane), one cannot tell whether or not the cyclist properly negotiated his merge. The evidence of the study is that the investigators didn't care tuppence for the skill, or its absence, used in merging.

2.5 Failure to Record Action Sequences

Yes, the videotapes presumably show each approaching cyclist from the start of the "midblock" area to when he crossed and exited the intersection. However, the data that is recorded and listed does not show the connection between the different elements of the cyclist's actions. In other words, you cannot connect the lane-crossing action with a left turn, proper or improper. Nobody can tell, without rerunning the videotapes, whether or not a cyclist made a left turn that was proper in all of its parts, or did something else, such as a chicken-left-turn. You know the numbers of each kind of letter, but you can't read the words.

2.6 Overtaking on the Right

The data processing apparently considers all cases of the cyclist overtaking on the right of a motorist to be bad. There is no consideration of whether or not the motorist is likely to turn right. At many major intersections (and many of those in the study were of this type), it is perfectly acceptable and safe for the cyclist to overtake on the right-hand-side of waiting motorists, because those motorists cannot turn right. The motorists may be waiting for the signal to change. The motorists may have already gone past the location for their right turn (right-turn-only lane, or pork-chop-island). Again, no consideration for how traffic normally operates.

2.7 Failure to Consider Right of Way

In many of these situations, the data coding fails to consider the normal operation of traffic as defined by the right-of-way rules. The subject just isn't mentioned very often, but it always applies to traffic movements.

2.8 What's a Stop?

Here are the coding categories for Stop Sign Errors:

  1. 1: {Cyclist or motorist] failed to stop at stop sign
  2. 2: [Cyclist or motorist] stopped at controlled intersection but failed to yield

This has been argued for decades, and the investigators still don't know how to code this situation. Stopping is one item, not so very important, while yielding is another item that is very important. All you have to do is to code them so the data processing can separate them, which their scheme failed to do.

2.9 Neglect of Skill

At no point do the investigators consider the effect of skill on cycling, or even of the traffic law. The cyclists are considered no more than animals observed; they do what they do and that's the end of it. The investigators might just as well have been counting the number of migrating geese that fly over a lake versus the number that land in it. The result, of course, are data that represent the collective ignorance and incompetence of typical American cyclists above the age of childhood (very few children were observed). There is no consideration for the types of training that would enhance the operation of the facilities being investigated. There is no investigation of the reasons for the defective maneuvers that were observed, and therefore none concerning the way to get cyclists to correct their behavior.

2.10 Is It Ignorance or Incompetence?

Most of the coding and observational defects that I noted above have been discussed in the professional literature for twenty years. The defective observational position was discussed at the presentation at TRB's Bicycling Committee of the analysis of the California Bike Safety Training about 1980, the analysis done by a female traffic expert (name forgotten now) from University of Souther California. The items to be observed and the data to be collected appeared in my Cyclist Proficiency Test observation sheet, used in my study of the operation of cyclists on bike-laned and non-bike-laned streets at about the same date.

Twenty years ago, I did much the same thing as this investigation, much better, for probably less than one-tenth the cost of this investigation (that is, if anybody had chosen to pay me for what I did for free).

3 Conclusions, Good and Bad

3.1 Meaningless Conclusions

I must say this for the investigators. Once they had conducted this investigation and collected all these data, they sat down to see what sense they could make of it. They reached the entirely justified conclusion that no useful conclusions could be reached from their data. They almost had the courage to say so, unequivocally.

What they did was to write up the investigation and the recommendations that they based upon it, just as if these were useful products, but when you come to read the "fine print" at the back you discover that there's nothing to use.

The investigators repeatedly say that the confounding variables (different volumes of traffic, different widths of lane, different amounts of sidewalk cycling, and the like) have so confounded the data that no conclusions or recommendations are justified.

3.2 Some Quoted Conclusions

"The debate over whether BLs or WCLs are preferable has been heated for many years. ... This comparative analysis of BL versus WCL sites utilized an extensive data base to examine many factors related to the operations and safety of these facilities. ... Across the board, these facilities work well, with the vast majority of identified conflicts in this study being minor in nature. ... in most cases the noted problems at the higher conflict rate sites could not be labeled as particular BL or WCL deficiencies. The destination patterns of bicyclists traveling through the project sites led to maneuvers and conflicts that in many cases would have occurred whether the bicycle facility present was either a BL or a WCL. ... The identified differences in operations and conflicts were related to the specific destination patterns of bicyclists riding through the intersection areas studied."

3.3 The Bad Conclusion

The investigators added one bad conclusion to their legitimate conclusion that their data provide no conclusions.

"Given the stated preferences of bicyclists for bicycle lanes in prior surveys, along with increased comfort level on bicycle lanes found in developing the Bicycle Compatibility Index, use of this facility [that is, bicycle lanes as opposed to wide curb lanes] is recommended where there is adequate width."

That is, because this study cannot demonstrate that bicycle lanes are worse than wide curb lanes, the nation should go for bicycle lanes.

At least that statement puts things in terms that can be quoted as being from the experts hired by FHWA. What they have written is that the only justification they can find for bike lanes is that the typical American cyclist thinks that bike-lane stripes protect him. Well, at least, they have officially stated that that false superstition is all the justification that they can find for bike lanes, while they carefully ignored the other reasons why wide curb lanes are better.

 4 The Florida Study

Wayne Pein reviewed this study in detail. That review is posted at
His review is sufficiently devastating as to details. I add the following comments based on the general position of such studies in reference to the full world of bicycle transportation.

The paper presents insignificant findings of an investigation that was not well designed. While the authors don't say it, the major conclusion is that when the available width is greater, the users tend to use "more" of it. So what? The real problem is at the heart of the study, which was erroneously conceived, inaccurately stated, and misleadingly described. The authors state their purpose: "The missing element from the FLDOT design guidelines [previously stated to be for wide curb lanes, bicycle lanes, and paved shoulders], as well as other documents suggesting specific geometric designs, are (sic) results indicating which types of facilities provide the greatest benefit to bicyclists and motorists by making it more comfortable for both modes, where comfort is defined as reducing unpredictable or potentially unsafe movements by either motorists or bicyclists and minimizing the risk of a conflict or crash. This research effort was undertaken to develop such results."

Which types of facilities are most comfortable? They never measure comfort, whatever that is. Instead, they redefine it as "reducing unpredictable or potentially unsafe movements." This bait-and-switch maneuver is rather forced on the investigators by the nature of the controversy. By reading the study one receives the implication that the authors prefer striped facilities, bike lanes and striped shoulders, to unstriped facilities, wide curb lanes, and they were undoubtedly hired by an organization, FLDOT, that has that preference. Were they measuring unpredictable movements? Doesn't look to me as though they were. When a motorist overtakes a cyclist, it is quite predictable that he moves a bit to the left, nothing unpredictable about that. How about potentially unsafe movements? Whatever they may have seen of these, such were so infrequent that the authors deliberately ignored them. "Also recorded were any erratic maneuvers or braking applications which took place during the passing maneuver; these events were extremely rare and were not included in the analysis due to the small number." In any case, we have no statistical data regarding collisions between motorists as the result of one of them overtaking cyclists; presumably these are very rare. We have more data concerning motorists sideswiping cyclists, which does indicate an insufficient clearance problem, but this study has no bearing on this condition. Therefore, this measurement of "comfort", whether of comfort that was not measured, or of "unpredictable or potentially unsafe movements", which were not measured either, means absolutely nothing at all.

The paradox that besets all these bicycle facilities studies is that the "comfort" factor, as it exists at all, occurs only where there is no crossing or turning traffic, while the collision frequency factor, which is real, occurs very largely where there is crossing or turning traffic and is entirely unrelated to the measurements made in this study. The two factors are entirely unrelated, except in the minds of the misinformed. In fact, none of the serious discussion of bikeways concerns the comfort factor, such as it may be, but concerns two other completely different subjects. The first is the effect of the bikeway upon the crossing and turning movements that must occur. The conclusion of this discussion is that all practical types of transportational bikeway system increase the number of traffic conflicts and increase the difficulty of making these conflicting movements safely; the greater the degree of practical separation, the greater are the difficulties. The second aspect of the bikeway discussion concerns the secondary effects of bikeways upon public policy. The conclusions of this discussion are that bikeways cause both bicyclists and motorists to make more dangerous movements, reduce the recognition that proper training is required (particularly for cyclists), and denigrate the legal and social statuses of bicyclists, all of which are detrimental.

The result of our knowledge about bikeways is that the only characteristic in their favor is that ill-informed people think that bikeways make them comfortable. Well, that's a subjective feeling, but the existence of the feeling represents nothing at all about the realities of the physical world. The result of this condition is that the bikeway advocates have no other characteristic with which to pursue their advocacy than that of comfort for ill-informed people. They have had to give up on the safety argument, and the motorist convenience argument is not very successful. The combination of this result and the fact that there is much money for advocating bikeways is a plethora of studies trying to demonstrate the value of bikeways by using the comfort argument. Often, as in this case, this is misstated in the bait-and-switch manner. And, since the purpose presents intellectual paradoxes that nobody has solved, or shows any sign of solving, the work is commonly of low intellectual caliber.

The Harkey, Stewart, and Rodgman study is just another of these intellectually vapid attempts to justify the intellectually decrepit bikeway hypothesis.

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