In 2012, the US Army Natick Soldier Research, Development, and Engineering Center completed the most comprehensive anthropometric study of a large population since the original ANSUR survey in 1988. They measured 6,068 active duty, Reserve, and National Guard soldiers across 93 body dimensions each. They collected 3D body scans of heads, feet, and complete bodies. They recorded demographic data including age, race, gender, and unit.
Then they released all of it to the public.
The Army Anthropometric Survey II — ANSUR II — is now the most widely used public anthropometric dataset in applied research, product design, and machine learning. It shaped the sizing systems in military uniforms, vehicle cockpits, protective equipment, and body armor worn by US soldiers today. And it became one of the primary training and validation datasets for the generation of body measurement APIs now being used by fashion companies, health apps, and ergonomic designers.
Why the Army cares so much about body measurements
The military has been measuring soldiers for longer and more systematically than almost any other institution. The reasons are practical and consequential.
Equipment fit affects performance. A body armor vest that doesn’t fit correctly provides less ballistic protection and restricts movement. A helmet that’s too loose or too tight impairs hearing, vision, and stability. A seat in an aircraft cockpit that’s sized for the average male pilot of 1960 may not accommodate the woman fighter pilot of 2025. These aren’t comfort issues — they’re operational readiness issues.
Procurement at scale. The US military procures clothing, armor, and equipment for hundreds of thousands of personnel. Getting the size distribution wrong creates inventory imbalances, wastes procurement budget, and leaves personnel without correctly fitting equipment.
Population change over time. The previous major survey (ANSUR I, 1988) was conducted 26 years earlier. In that time, the US adult population had grown significantly heavier and the military’s demographics had changed — more women, different racial composition. The 1988 data was no longer representative.
What was measured
Each of the 6,068 subjects contributed 93 directly measured dimensions:
Body lengths and heights: stature, sitting height, knee height, crotch height, ankle height, tibial height, trochanterion height.
Reach and span: arm span, overhead reach, functional reach.
Breadths and depths: biacromial, bideltoid, chest breadth and depth at three levels, waist, hip, elbow.
Circumferences: head, neck, chest at three levels, waist at three levels, hip, thigh, calf, ankle, knee, upper arm, forearm, wrist, hand.
Head and face: head length, head breadth, face width, chin-to-ceiling height, interpupillary distance, sagittal arc, bitragion arc.
Hand and foot: hand length, palm length, individual finger lengths and widths. Foot length, ball of foot, heel width, instep length.
Beyond measurements, each subject received full 3D body scans. The 3D data allowed researchers to extract additional derived measurements and create population-level body shape models — something not possible from tape measure data alone.
The 1988 comparison: what changed
Comparing ANSUR II to ANSUR I reveals 24 years of physical change in the US Army:
Weight increased significantly. Average male soldier mass increased by approximately 8–10kg between 1988 and 2012. This drove increases in circumference measurements while height remained relatively stable.
Body mass index distribution shifted. The proportion of higher-BMI soldiers increased substantially, reflecting broader American population trends.
Women were included in greater numbers. ANSUR I had a limited female sample. ANSUR II included a larger and more representative female cohort, reflecting the expansion of women’s roles in the Army.
Racial composition changed. The demographic composition of the Army evolved significantly between 1988 and 2012, affecting the population-level distributions of all anthropometric measurements.
These changes weren’t uniform across dimensions. Height changed little. Weight changed substantially. Circumference dimensions changed in proportion to weight. Skeletal dimensions (biacromial breadth, limb lengths) remained relatively stable.
Why the public release was unusual — and valuable
Military data tends to stay classified. ANSUR II is an exception. The Army released the full dataset, including individual subject measurements (with personal identifiers removed), 3D scan geometries, and summary statistics.
The decision to release publicly was deliberate. The stated rationale was that the data serves a “wide range of equipment design, sizing, and tariffing applications within the military and have many potential commercial, industrial, and academic applications.” Keeping it proprietary would have limited its use to Army contractors. Releasing it created a scientific commons.
The impact has been significant:
Body measurement API development. ANSUR II became one of the primary training datasets for machine learning models that predict body dimensions from height and weight. Its scale (6,000+ subjects), measurement completeness (93 dimensions each), and public availability made it suitable for training regression models where most anthropometric datasets are too small or too restricted.
Academic research. Hundreds of peer-reviewed papers have used ANSUR II data for research in ergonomics, public health, fashion technology, and human factors.
Industrial design. Product designers across the US used ANSUR II to establish 5th–95th percentile design ranges for consumer products — a practice previously limited to organizations with access to commissioned studies.
Sizing system research. ANSUR II enabled systematic comparison of different size chart construction methods, helping resolve debates about optimal size grading in clothing.
The limitations researchers acknowledge
ANSUR II is remarkable data, but it has documented limitations:
It’s a military population. Soldiers are screened for physical fitness. The ANSUR II population has a lower average BMI and higher fitness level than the civilian population. For civilian consumer products, applying ANSUR II distributions directly overestimates lean body mass and underestimates obesity-related body shape variation.
It’s American. The US Army is a US population. ANSUR II proportions reflect the demographic mix and lifetime nutritional history of Americans in 2012. They differ from European, Asian, or African populations in systematic ways.
It’s dated. Data from 2012 is now 14 years old. American population anthropometry has continued to shift toward higher average BMI. Some dimensions from ANSUR II no longer accurately represent the current population.
It’s cross-sectional. ANSUR II measured each person once. It tells you the distribution of body dimensions at one moment, not how they change over time within individuals.
These limitations don’t diminish the dataset’s value — they define its scope. Researchers who use ANSUR II appropriately treat it as a calibration input for models that are then validated against broader population data, not as a direct representation of any civilian group.
A note on “trained on ANSUR II”
If you encounter a body measurement product that claims to be “trained on ANSUR II” or “powered by ANSUR II,” the honest interpretation is: the model was calibrated using ANSUR II as one of its validation or training datasets. This is a meaningful claim about data quality — ANSUR II is a rigorous, well-documented study. But it’s also a population-specific claim: predictions calibrated primarily on ANSUR II may be less accurate for non-American, non-military populations.
The most robust implementations use ANSUR II alongside other national studies (SIZE KOREA, SIZE UK, NHANES, CAESAR) to build models that generalize better across populations. ANSUR II is a cornerstone, not a complete foundation.
The ANSUR II dataset is a remarkable example of public science generating private value. A government-funded study of soldiers created infrastructure that now helps civilians find better-fitting clothes, ergonomists design better workstations, and developers build sizing features for global apps. That’s an unusual and gratifying trajectory for anthropometric data.
The complete ANSUR II dataset is available at the Penn State Open Design Lab and through the US Army DEVCOM Soldier Center.