Skip to main content
Scientific Principles

Bertillonage

A late-19th-century system of criminal identification using a set of anatomical measurements, based on the assumption that each person's body dimensions are unique and stable over time. Galton's analysis of correlated measurements revealed diminishing returns from adding more measurements, leading to the development of correlation and the eventual replacement by fingerprinting.

THE IDEA

BERTILLONAGE

Bertillonage, also known as the Bertillon system, was a method of biometric identification based on anthropometry—the precise measurement of the human body. Used widely in the late 19th and early 20th centuries, it combined detailed numerical data (such as head length and breadth, finger and foot lengths, and arm span) with photographs and written descriptions to create a unique profile of each individual. The system was the brainchild of Alphonse Bertillon, a French criminologist devoted to a rigorously quantitative approach. Appalled by the unsystematic way police identified suspects, he reasoned that a set of unchanging physical measurements would be far more reliable than names or aliases: “You can replace your name by an alias, but there’s no alias for the shape of your head.” Adopted by the Paris Prefecture of Police in 1883, bertillonage quickly spread across the globe; at its peak, departments from Bucharest to Buenos Aires maintained “Bertillon cabinets,” which became, in the words of the criminologist Raymond Fosdick, “the distinguishing mark of the modern police organization.”

The Victorian statistician Francis Galton, who had done extensive work with anthropometric data, saw an opportunity to refine the system. He asked whether Bertillon’s choice of measurements was optimal, or if taking more measurements would improve accuracy. Galton realized that bodily measurements are not independent: for example, hand size and foot size tend to vary together. Adding measurements that are highly correlated with ones already recorded yields diminishing returns, because the new numbers add little extra information. To study this, Galton plotted height against cubit (elbow-to-middle-finger) and found the same elliptical scatter pattern he had observed in other correlated pairs. He looked for measurements that added information beyond those already recorded, a question he could investigate using his collection of anatomical data.

In the end, Galton’s statistical critique did not lead to an improved Bertillon system. Instead, he championed a competing method: dactyloscopy, or fingerprinting. Like bertillonage, it reduced a suspect to a set of recordable symbols, but it had the crucial advantage that fingerprints are often left behind at crime scenes, available even when the criminal is not. The case of Vincenzo Peruggia—who stole the Mona Lisa in 1911 and left a print on the discarded frame—made this point vividly. Had the police files contained dactyloscopic records instead of Bertillon cards, Peruggia would have been identified immediately. Thus, bertillonage gave way to fingerprinting, which in turn prefigured modern biometric tools like DNA profiling. Indeed, as Justice Anthony Kennedy noted in the 2013 U.S. Supreme Court case Maryland v. King, a DNA sequence functions much like a twenty-first-century Bertillon card—a set of uniquely identifying data points attached to an individual.

When to use it

  • Criminal identification before the adoption of fingerprinting

  • Archival studies of historical police records using Bertillon cards

  • Teaching the evolution of biometric identification methods

  • Designing a biometric identification system: assessing how much identifying information a measurement adds and how reliably it can be recorded.

  • Historical analysis of forensic methods: examining the limits of adding measurements that largely repeat information already collected.

  • Evaluating the efficiency of diagnostic tests: when multiple tests are highly correlated, adding more tests yields little new information.

  • Teaching statistical concepts: using Bertillonage as a real-world example of correlation and diminishing returns.

FROM THEORY TO EVERYDAY LIFE

In practice

01

Police identification in the 1880s

Authorities used Bertillonage to track repeat offenders by recording detailed body measurements on index cards.

02

Fall from use

The system was eventually replaced by fingerprinting, which proved more reliable and easier to implement.

03

The Mona Lisa Thief

Vincenzo Peruggia, who stole the Mona Lisa in 1911, had been previously arrested in Paris. His Bertillon card was filed according to his body measurements, but because the card did not include fingerprints, the print he left on the discarded frame could not be matched to him. This case highlighted the limitation of Bertillonage compared to fingerprinting.

04

Maryland v. King (2013)

In this Supreme Court case, Justice Kennedy compared the collection of DNA from arrestees to the Bertillon system, arguing that a DNA sequence is simply another set of data points used for identification, much like the anatomical measurements of Bertillonage.

✳

A MOMENT TO REFLECT

Put it to work.

  1. How did Bertillonage reflect the scientific approach to law enforcement in its time?

  2. What were the practical limitations of using body measurements for identification?

  3. How does the principle of diminishing returns from correlated measurements apply to modern data collection and feature selection in machine learning?

  4. What ethical concerns arise from using biometric data for identification, as seen in Bertillonage and modern DNA databases?

  5. Why did fingerprinting ultimately replace Bertillonage, and what does this tell us about the evolution of forensic science?

Connected ideas

Another lens can bring the same problem into focus.

Often confused with

References

  1. Signaletic Instructions — Alphonse Bertillon (1886)
  2. European Police Systems — Raymond B. Fosdick (1915)
  3. Maryland v. King — Supreme Court of the United States (Justice Anthony Kennedy) (2013)

CONTINUE THE CONVERSATION

Source texts

  • How Not to Be Wrong The Power of Mathematical Thinking (Ellenberg, Jordan)
  • How Not to Be Wrong The Power of Mathematical Thinking (Ellenberg, Jordan)
← More scientific principles