M. Kary’s review of the Lusk et al. Montreal bikeway study — A compendium of errors and omissions, or: What is not in this article

Montreal resident, cyclist and mathematician M. Kary has reviewed the study by Lusk et al. of Montreal’s two-way one-side of-the-street barrier-separated on-street bikeways.

The study by Lusk et al. purports to show that the bikeways are 28% safer than riding on comparable streets without the bikeways. Kary points out a number of serious methodological and factual errors which, he contends, invalidate that conclusion.

Kary’s review is available in two formats: the PDF version looks nicer, especially when printed on paper. The HTML version is better for viewing in an Internet browser.

The review is heavily annotated and is supported by a large number of photographs. Each photo caption includes a link to a Google satellite view or street view, to pinpoint the location.

Here are a few quotes to convey the flavor of the review.

Rachel between St Urbain and Marquette

The authors list this path segment as being 3.5 km long. In fact it is approximately 1.7 – 1.8 km long, and thus has approximately twice the rates of injuries and crashes per kilometre given by Lusk et al. (The authors need to explain how they obtained the lengths they give for all their path segments.)

Construction of the reference and intervention samples

…Since presumably the authors do not count incidents occurring to cyclists travelling along the path but in the terminating intersections, using short path segments as the authors do also lowers the overall number of intersections per kilometre, making for a comparison more favourable to the paths. (Example: for block lengths of 200 m, a 1 km path without the terminating intersections would have incidents resulting from 4 intersections per kilometre, while a 5 km stretch of the same path would have incidents resulting from 4.8 intersections per kilometre— a 20% increase.)

Crash rates described as injury rates

The “What this study adds” box incorrectly compares the (incorrectly calculated, see sections 1.1 and 1.3) injury rate of 8.5 per km reported by the authors with a range of 3.75 to 67 supposedly reported elsewhere. In fact the latter figures are crash rates, not injury rates.

Figure 13: René Lévesque at Maison Radio-Canada

In 1990. renowned CBC journalist and producer Joan Donaldson attempted to cross this two-way path to get a taxi, and was struck down by a cyclist coming from her right. She remained in a coma for at least six months, could not speak at all for three years, and was left permanently brain-damaged and quadriplegic. The cyclist was uninjured. Donaldson died in 2006 from extended complications of the accident. Authors’ method does not count such injuries, for not being to a cyclist.

Enough. Please read the review.

13 responses to “M. Kary’s review of the Lusk et al. Montreal bikeway study — A compendium of errors and omissions, or: What is not in this article

  1. Thanks for the heads-up, John.

  2. A question, John. When rebutting or critiquing a published, peer-reviewed paper, it is normal to send in the critique to the original publication (in this case Injury Prevention) for a Comment and Reply. Do you know if this critique was sent to the Editor of Injury Prevention? Generally, that gives the original authors a chance to respond to the criticism, and have the general journal audience able to see the original, the critique, and any response to the critique. Been there, done that.

  3. Thanks for sharing John. Quite interesting, IMO.

  4. > The authors list this path segment as being 3.5 km long. In fact it is approximately 1.7 – 1.8 km long,

    Hmm. 1.75km (the average of 1.7 and 1.8 km) x 2 = 3.5 km. What a strange coincidence …

    Could it be that this section has “path segments” on both sides, one for each direction of travel?

    (Of course, you probably caught that too, and intentionally highlighted that section without explaining it so that others would catch it as well.)

  5. OK, but still … it would seem that counts as two path segments rather than one.

    Sure, perhaps the author should “explain how they obtained the lengths they give for all their path segments” — or the readers could go “oh, that’s what they meant” rather than screaming “wrong!”

  6. If the authors counted a path segment twice, ie. once for each direction of travel, it is important to make sure that is consistent with established crash methodology if the numbers are to be directly comparible. Counting each direction (i.e., doubling the length) is not consistent with the following guidance.


    If bicycling crash rates are calculated similarly to MV crash rates, than a 1.7 mile length of bidirectional bike path should be counted as 1.7 miles. It would be important to know whether different papers measuring different cities are all using the same criteria, or you end up with numbers that do not directly compare.

  7. There is a bicycle and pedestrian crash analysis tool here. I have no experience with it.

  8. The authors counted the other segments once over their two-way length, more or less, with this exception. Please read their paper, and Kary’s review, to clear up the confusion!

  9. This is a very much appreciated info, for the others who is still in doubt, we bikers has the same risk as the motorcycle/car drivers, you can never choose where,when and how will it occur, atleast there is some one who is caring to post about it,My arizona personal injury lawyer once told me a quote which i really appreciate,”Did you ever observe to whom the accidents happen? Chance favors only the prepared mind”. a line by Louis Pasteur, which has a very deep meaning.

  10. Pingback: North-south, east-west: is that best? | John S. Allen's Bicycle Blog

  11. Pingback: China Cups and Butterflies; Options and Ethics | i am traffic

Leave a Reply

Your email address will not be published. Required fields are marked *

7 + = sixteen

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>