Last year I wrote an article about widespread criticisms of the BMI measurement, and in that piece I included the following paragraph:
Another criticism is that the BMI was developed over 150 years ago, and its longevity somehow discredits it. What would become the BMI was developed around 1850 by a Belgian statistician named Adolphe Quatelet, and has been used more or less since then. This is, of course, flawed logic: if anything the fact that the BMI has been widely used for so long is actually evidence that it works-not that it doesn’t work. To use only one example of many, germ theory has been around since at least 1815 (when Agostino Bassi did experiments showing that the etiology of disease could be traced to germs), and no one suggests that germ theory is “outmoded” or incorrect merely because it’s been around for nearly 200 years.
In response, I recently got an e-mail from a reader:
“It’s a good article and one I agree with; however I feel compelled to point out a logical flaw. You state: ‘if anything the fact that the BMI has been widely used for so long is actually evidence that it works-not that it doesn’t work.’ Data shows something works and is useful, not time.”
I gave his comment some thought, and replied:
Thanks for writing… I thought someone might take me to task for that point, but I think it remains valid. You are correct that on the surface it seems like an “argument from antiquity” logical fallacy, but the context is important: the BMI is not used among soothsayers, acupuncturists, or homeopaths; it is used in the scientific and medical establishment.
It is true that data, not time, determines whether something is useful, and in this case both data and time show that the BMI is useful-furthermore, data and time are not independent factors but inextricably linked. Scientific and medical data (especially long-term longitudinal efficacy studies) take time (years or even decades). In fact, because short-term studies will reflect random variation (and create false-positive results), it is the bulk of studies, the long-term overall body of evidence, that determines the scientific consensus: Data, yes, but also, by definition over long periods of time. You can’t really separate out the two elements in this context.
Because science is a self-correcting enterprise (as is its medical applications), I think it’s reasonable to link time and data. It is true that now and then a drug or medical procedure is, in retrospect, found to be ineffective. But when it is, medicine addresses it, and the drugs and treatments fall by the wayside and are discredited-usually in a few years but in rare cases a decade or more. Because of this, things that work (such as the BMI) are likely to be valid if used for over a century. I did not mean to suggest that the BMI’s longevity proves its efficacy, just that it lends some support and credibility.
Conversely, if the BMI had only been around for a few years, critics would complain that it had not been around long enough to have been adequately tested and proven its efficacy-which would also smack of an “argument from antiquity” fallacy, but in reverse.
You can’t have it both ways: either a widely-used medical measurement has a better likelihood of being true because doctors have (apparently successfully) used it for generations (and in the process gathered data in support of its usefulness, as they have with x-rays, antihistamines, statins, heart drugs, etc. etc.) or its longevity is irrelevant. Data and time are not the same measurement, but they are closely linked, and I think we can validly suggest that something that has been used for generations in the context of a science-based enterprise is more likely to be valid than something introduced yesterday or last year.
What do you think?