Maryland (v. King) Corrections Department: David Kaye on Bertillonage

In a recent post on the Maryland v. King case, I criticized the Court’s reference to Bertillon measurements, or Bertillonage, as “a thoroughly debunked form of proto-eugenics.” In correspondance, however, David Kaye, an expert on forensic science at Penn State Law (and co-author, with Hank Greely, on an amicus brief in the King case) has suggested that my criticism may be too harsh.

In the late nineteenth-century, Alphonse Bertillon, a French mathematician, developed what became known as Bertillonage, a form of  identification based on measurements of various aspects of the human body, such as the length of the left foot and the width of the head. The principle behind Bertillonage was that, with enough variables, exceedingly few individuals should have all of the same measurements. What are the chances, for example, that two people would be 5 ‘ 9 ” and have  a 40-inch chest and a size 10.5 shoe? Eventually Bertillonage fell out of favor due to the widespread adoption of fingerprint identification, but one of its vestiges remains: the mugshot.

“thoroughly debunked”
I based my claim that Bertillonage was thoroughly debunked on two factors: (1) that the practice suffered from what is known in statistics as “endogeneity;” and (2) that it’s no longer used anywhere. Endogeneity is a relatively simple problem: you may underestimate the likelihood of two “chance” occurrences happening together if they are, in fact, correlated. The chance of any given person being a Stanford alumnus is relatively small; and the chances of any given person owning a red T-shirt may also be small, but the number of Stanford alumni who own red T-shirts may be much higher than “chance”: Stanford’s official school color, as emblazoned on many T-shirts, is cardinal.

So, for Bertillon measurements, let’s take my example above. What are the chances that two people would be 5 ‘ 9 ” and have  a 40-inch chest and a size 10.5 shoe? Answer: probably a lot higher than pure chance. Why? For two reasons: First, each of the measurements, at least in the U.S., is the mode–more people are 5 ‘ 9″, for example, than any other height. And second, humans are relatively proportional. Taller men generally have bigger chests and bigger than shorter ones. Thus, it’s much, much more common for an average-height man to have average sized feet than for an average-height man to have teensy weensy feet.

But does this mean that Bertillon measurements are totally bunk? No, says, David Kaye. Just because two variables are moderately correlated does not mean that they’re not discriminating whatsoever. While it’s true that it’s pretty common for someone to be 5 ‘ 9″, have a 40 inch chest, and size 10.5 feet, that still will narrow down a suspect pool, if necessary. And furthermore, if we do have a suspect who’s 6 ‘ 2 ” and has, say, a size 5 shoe, that will really narrow down our culprits. As the Hankbook of Multibiometrics states, Bertillonage, though not perfect, is “reasonably accurate.”

So much for endogeneity. What about the fact that no one uses Bertillonage anymore? Well, as I mentioned above, this is mainly a product of historical artifact. Fingerprint technology was coming “on line,” just after Bertillonage, and was thought of as the superior technology. Bertillonage was replaced rather than discredited.

There’s a darker side to Bertillonage, however. Alphonse Bertillon testified on behalf of the prosecution during the Dreyfus Affair, and presented a mathematical analysis of handwriting so shoddy, that almost every famous French mathematician of the time, including Henri Poincaré, disputed it. In 1955, the famous historian Nicholas Halasz, catalogued Bertillon’s ulterior motive: virulent anti-Semitism. Later, Francis Galton–the founder of the eugenics movement–incorporated Bertillonage into much of his earlier work. And at least one book links these two concepts–eugenics and Bertillonage–together. But, as David Kaye points out, this may not be necessarily so. Bertillonage, despite its provenance, was used purely for identification purposes, and while Galton may have used Bertillon’s measurements for his later work on eugenics, this doesn’t necessarily mean that Bertillonage had the same twinge of racism to it as did Bertillon’s testimony at Alfred Dreyfus’s trial. The history of eugenics is, in fact, relatively complicated and had (and unfortunately continues to have) many flagbearers.

Thanks, David Kaye, for pointing this nuances out! I suppose I wouldn’t have expected less from an expert on forensics. (And for those of you interested in more about Bertillon and the Dreyfus Affair, check out David Kaye’s 2007 essay in the Minnesota Law Review. A great read.)