A crash dummy, formally called an anthropomorphic test device (ATD), is a full-scale human replica built to measure the forces a person’s body would experience during a vehicle collision. These devices are packed with sensors that record data on impact forces, motion, and compression across the head, neck, chest, spine, and limbs. That data tells engineers whether a car’s safety systems, like seatbelts and airbags, actually protect people in a crash.
Modern crash dummies cost between $550,000 and $750,000 depending on their instrumentation level. They’re precision instruments defined by roughly 250 engineering drawings and specifications, regulated under federal standards that ensure every dummy of the same model responds identically under the same conditions.
What a Crash Dummy Is Made Of
A crash dummy isn’t a simple mannequin. It’s a layered assembly of materials chosen to respond to impact the way human tissue does. The outer skin is a PVC-based vinyl layer produced through injection molding. Beneath that sits a sponge-like filler made of polyurethane foam, with an adhesive layer bonding the two together. This sandwich of materials is designed to compress and flex similarly to human soft tissue when struck.
The internal structure uses metal components for the skeleton, including steel ribs in the chest. But the design is careful to prevent metal-on-metal contact during a crash, except where metal parts already touch when the dummy is sitting still. This matters because unintended contact between hard internal parts would create false readings, corrupting the data engineers rely on.
That said, current dummy skin doesn’t perfectly replicate how human tissue behaves. Real soft tissue responds differently depending on how fast it’s hit. Dummy skin materials are about two to three times less viscous than human tissue, meaning they hold their shape more like an elastic rubber band rather than flowing and deforming the way skin and muscle actually do. Researchers continue working to close that gap.
Sensors Inside the Body
The real value of a crash dummy is what’s inside it: dozens of sensors embedded throughout the body that record exactly what happens during a collision. These fall into three main categories. Accelerometers measure how quickly different body parts speed up or slow down on impact. Load cells measure the raw force applied to specific areas, like the thighbone hitting the dashboard. Torque sensors capture twisting forces, particularly in the neck during sudden deceleration.
Together, these instruments track things like how much the chest compresses against a seatbelt, how much force the knees absorb from the dash, and how violently the head snaps forward. The readings get mapped to injury thresholds that predict whether a real person would suffer a concussion, broken ribs, spinal damage, or worse.
The Nine Major Body Assemblies
A standard crash dummy is built from nine modular component assemblies: a head, neck, shoulder and thorax (upper torso), right arm, left arm, lumbar spine (lower back), pelvis and abdomen, right leg, and left leg. Each assembly can be removed, inspected, recalibrated, or replaced independently. This modularity is essential because crash testing is destructive. Parts wear out or break, and engineers need to swap components without rebuilding the entire dummy from scratch.
Different Sizes for Different People
Not everyone who rides in a car has the same body. Crash dummies come in multiple sizes to represent different portions of the population. The most commonly used model is the Hybrid III 50th percentile male, which represents a roughly average-sized adult man. There’s also a 5th percentile adult female, which represents a smaller-statured woman. The National Highway Traffic Safety Administration (NHTSA) has refined the female dummy’s specifications over time, including updates to chest dimensions based on body measurement data from the CDC to better reflect the average American woman’s proportions.
Pediatric dummies representing children of various ages also exist for testing child restraint systems like car seats and booster seats. Researchers at institutions like Children’s Hospital of Philadelphia have developed computational models of children’s bodies to supplement physical testing, since child-sized dummies are harder to validate against real injury data.
From Cadavers to Sierra Sam
Before crash dummies existed, the alternatives were grim. Starting in the late 1930s, researchers at Wayne State University in Detroit used human cadavers to study what happens to a body in a high-speed collision. Biomechanics barely existed as a field, and there was no reliable data on how crushing and tearing forces affected human anatomy. Cadavers provided the first real dataset.
By the mid-1950s, cadaver research had yielded most of the information it could. Researchers needed to study survivable crashes, something cadavers couldn’t help with, and donated bodies were in short supply. This led to animal testing. Chimpanzees rode rocket sleds. Bears were placed on impact swings. Anesthetized pigs were strapped into harnesses and crashed into steering wheels at 10 miles per hour.
The first actual crash dummy arrived in 1949. Samuel W. Alderson created “Sierra Sam” at his Alderson Research Labs in partnership with Sierra Engineering Co. Sierra Sam was originally built to test aircraft ejection seats and pilot restraint harnesses. By the early 1950s, Alderson and aircraft manufacturer Grumman had adapted the concept for motor vehicle crash testing. In 1973, NHTSA formalized crash dummy specifications under Part 572 of the Code of Federal Regulations and issued the first occupant crash protection regulation requiring dummy-based testing.
Hybrid III vs. THOR
The Hybrid III has been the standard crash dummy for decades, but a newer model called THOR (Test Device for Human Occupant Restraint) is gradually replacing it for frontal crash testing. Both represent a 50th percentile adult male, but they have different strengths.
In side-by-side testing, the Hybrid III produced more realistic responses in the chest and lower extremities, while THOR was more accurate in replicating head movement. Where THOR really pulled ahead was in predicting actual injuries. When researchers compared both dummies’ sensor data against injuries observed in cadaver tests under the same crash conditions, THOR’s injury predictions were more accurate overall, even though the Hybrid III’s moment-by-moment physical responses sometimes looked more lifelike.
THOR’s improved injury prediction is the main reason regulators are adopting it. A fully instrumented THOR-50M runs up to $750,000. NHTSA has estimated that switching from Hybrid III to THOR in regulatory testing adds roughly 16 euros (about $17) to the cost of manufacturing each vehicle that needs to be re-validated for compliance.
Virtual Crash Dummies
Physical dummies have a fundamental limitation: they measure forces on a rigid mechanical structure, not on actual human anatomy. Virtual human body models are computer simulations that replicate the bones, organs, muscles, and connective tissue of a real person. Engineers can run thousands of simulated crashes, varying the speed, angle, and occupant position, far faster and cheaper than physical tests allow.
These digital models let researchers “look inside” the body before, during, and after a simulated crash to see how internal organs shift, where bones flex, and how soft tissue deforms. This is especially valuable for populations that are hard to represent with physical dummies, like children of specific ages. Pediatric human body models have been used for over five years at major research centers to refine child safety technologies and car seat guidelines.
Virtual models don’t replace physical dummies. Federal regulations still require real crash tests with real dummies for vehicle certification. But the two approaches complement each other: virtual models explore a wider range of scenarios, and physical dummies provide the legally mandated, repeatable measurements that determine whether a car passes or fails.