The first practical application of systematic body measurement wasn’t fashion. It wasn’t ergonomics. It wasn’t medicine. It was catching repeat criminals.
In 1879, Alphonse Bertillon, a 26-year-old records clerk at the Paris Prefecture of Police, proposed a system for identifying recidivist criminals based on a standardized set of body measurements. The system he invented — now called bertillonage or the Bertillon system — was the first practical biometric identification database in history. It worked, it spread globally, and it ultimately failed — for reasons that illuminate why body measurement science is harder than it looks.
The problem Bertillon solved
In 1879, the Paris police had a filing system for criminal records organized by name. This created an obvious loophole: a repeat offender could provide a false name and his prior record would never be found. The police knew this happened constantly. They had no solution.
Bertillon’s insight was deceptively simple: while people can change their names, they cannot change their bodies. A person’s physical measurements are unique to them (approximately) and don’t change significantly in adulthood. If you could measure a person on first arrest and then identify them with certainty on any subsequent arrest, you could track criminal recidivism across aliases.
The challenge was creating a measurement system that was:
- Precise enough to be unique per individual
- Reproducible across different operators and different times
- Fast enough to be practical for routine police work
- Based on measurements that didn’t change with age, weight fluctuation, or disguise
The eleven measurements
Bertillon settled on eleven primary measurements, selected to maximize discriminating power while remaining practical to collect in a police station:
- Stature — standing height
- Sitting height — height to top of head while seated
- Arm span — horizontal reach from fingertip to fingertip
- Head length — front to back
- Head breadth — widest point
- Right ear length — from top to lowest point of the lobe
- Left middle finger length — from base to tip
- Left little finger length
- Left foot length
- Left forearm (cubitus) — elbow to tip of middle finger
- Trunk circumference
These were supplemented by photographs (two standardized portraits: full face and profile — the “mug shot” format Bertillon invented), eye color, hair color, and a written description of distinctive marks.
The eleven measurements were selected through what we would now call feature selection: Bertillon determined, empirically, which combinations of measurements would divide the criminal population into the largest number of distinguishable groups. He calculated that eleven measurements, each divided into three categories (small, medium, large), produced 3^11 = 177,147 possible combinations — enough to make random matches vanishingly rare.
The mathematics of the system
Bertillon’s statistical reasoning was sound for its time. He correctly understood that the discriminating power of the system depended on measurement independence — each measurement contributing different information about the individual.
The eleven measurements he chose were selected partly because they are relatively independent of each other: head breadth and foot length are not highly correlated, so both contribute independent information. This is the same statistical principle that drives feature selection in modern anthropometric machine learning models.
The filing system organized records by a hierarchy of these measurements. To find a matching record, an operator would take the measurements of a new arrest, calculate which of the 177,147 cells they fell into, and search that cell — a dramatically smaller set than the full archive.
The system’s success
The Bertillon system was adopted by police departments across France within a few years, and then exported internationally. By the 1890s, police departments in the UK, Germany, Belgium, Austria-Hungary, and the United States were operating Bertillon identification bureaus. The Paris bureau alone was processing thousands of identifications per year.
Cases that would previously have been impossible to solve — a suspect claiming a false identity with no paper trail — could now be solved by walking them to the measurement station and comparing results against the archive.
The system’s public success created the first popular awareness of biometric identification — and the first public discussion of whether systematic physical measurement of citizens by the state was appropriate. These debates are recognizable as ancestors of the contemporary arguments about facial recognition, biometric databases, and surveillance.
The collapse: the West problem
The Bertillon system’s end is as instructive as its beginning. In 1903, the Leavenworth Federal Prison in Kansas encountered a paradox.
A prisoner arrived named Will West. The intake officer measured him using the Bertillon system and searched the records. They found an identical match — not a close match, an essentially identical set of measurements — for a man named William West already serving a life sentence in the same prison.
The two men were examined. They were different people, with no known family relationship. Their eleven Bertillon measurements were virtually identical. The Bertillon system could not distinguish them.
The resolution came through a new technology: fingerprints. Their fingerprint patterns were entirely different. Will West was fingerprinted, his record was created, and within a decade, fingerprinting had replaced bertillonage as the standard criminal identification technology in most of the world.
What the failure reveals about anthropometry
The Will West case is usually told as a story about fingerprints beating body measurement. It’s also a story about the fundamental statistical limitation of anthropometric identification.
Body measurements are distributed continuously. Two unrelated people can have similar measurements by chance. For personal identification — where you need near-certainty about specific individuals — measurements are not unique enough.
For population-level statistical work — understanding average dimensions, designing for a range of body sizes, calibrating ergonomic standards — measurements are exactly the right tool. They describe distributions, support design decisions, and enable products that fit populations.
The distinction between “identify a specific individual” (where measurements fail) and “predict population-level dimensions” (where measurements succeed) is the key to understanding what body measurement technology can and cannot do.
The modern inheritance
The Bertillon system left several legacies:
The mug shot: The standardized two-photograph format (full face and profile) that Bertillon invented for his system is still used by police departments worldwide.
The anthropometric survey tradition: Bertillon’s careful, standardized measurement methodology influenced the scientific tradition of anthropometric surveying — the same tradition that produced ANSUR II, SIZE UK, and the other population datasets that body measurement prediction systems draw on today.
The biometric privacy debate: Bertillon’s system provoked the first large-scale public discussion about whether systematic physical measurement of individuals by institutions is appropriate. The arguments made in 1890s France about police measurement of suspects resonate strongly with contemporary debates about facial recognition and biometric databases.
The principle of standardization: Bertillon’s insistence on absolute measurement standardization — specific equipment, specific postures, trained operators following the same protocol — was an early instance of the principle that comparison of measurements across time and across observers requires rigorous methodology. ISO 7250-1 exists because Bertillon was right about this.
A note on the field’s moral history
Anthropometry’s history is not uniformly benign. In the late 19th and early 20th centuries, body measurement was enlisted in support of racist pseudoscience — the claim that body measurements could reveal racial hierarchy, intelligence, or criminal tendency. These applications were scientifically wrong and morally catastrophic. They were also scientifically influential in their time, which delayed recognition of how wrong they were.
The scientific body measurement tradition that emerged from this history — particularly after World War II and the Nazi use of race anthropometry — deliberately separated physical measurement from any claims about human hierarchy or intrinsic worth. Modern anthropometric science measures dimensions, not people’s value.
This history is worth knowing for anyone working with body measurement data: the field’s applications have ranged from genuinely useful (ergonomic design, clinical growth monitoring, sizing) to deeply harmful (racial categorization, criminal typing). The technology is a tool; its moral valence depends entirely on how it’s used and what claims are made from it.
Bertillon himself died in 1914, before fingerprinting had completely replaced his system. He spent his last years working on the forensic application of handwriting analysis, convinced it would prove more reliable than fingerprints. He was wrong about that too.
His broader contribution — that systematic, standardized body measurement could create actionable information at scale — has proven more durable than any of his specific methods.