Injury tolerance numbers carry an authority they did not earn on their own. A criterion value in a federal standard looks like a bright line, but behind it sits a chain of cadaveric testing, sub-injurious volunteer work, animal scaling, statistical fitting and regulatory compromise, each step introducing assumptions the number no longer displays. Knowing where a tolerance value came from, and which body it describes, separates a defensible comparison from an arithmetic one.
A tolerance is a population statement
Human tolerance data describes how a sample responded to a loading condition. It is a distribution, summarised: a value tied to a stated injury probability, a curve of risk against a measured parameter, sometimes only a range within which no injury was observed. It is never a property of an individual the way yield strength is a property of steel.
Two people under identical loading can have different outcomes for reasons the literature does not resolve: age, bone density, muscle tone at impact, prior injury, posture, and whether the event was anticipated. The number does not fail then; it never claimed otherwise.
Where the underlying data comes from
Three sources, each limited. Post-mortem human subject testing produces injury-level data on real anatomy, but on tissue without muscle tone, generally from an older donor population, under research governed by institutional review and donation frameworks that constrain how the work is done.
Volunteer testing produces high-quality kinematic data on living, muscled subjects, but ethics review permits only sub-injurious loading, so such studies describe the approach to a threshold rather than the threshold. Animal work and surrogates bridge the gap by scaling, and scaling is where assumptions accumulate.
The dummy is an instrument, not a person
Anthropomorphic test devices are specified in regulation — the Part 572 family in the United States — and built to defined anthropometry: a mid-size male, a small female, a larger male, and child devices. Repeatability is their virtue. Biofidelity is conditional: they reproduce human response within the conditions they were validated for, not reliably outside them.
A device developed for frontal loading is not a general-purpose analogue in lateral or oblique impact, which is why separate side-impact devices exist and why ISO publishes biofidelity rating procedures. Instrumentation matters too: SAE J211 governs how impact-test channels are specified and filtered, and a value computed with the wrong filter class is not comparable to the literature.
HIC and what it does not say
The Head Injury Criterion condenses a resultant head acceleration time history into one number by integrating over the worst interval — commonly fifteen or thirty-six milliseconds — and weighting duration against magnitude. It descends from cadaveric skull-fracture work and the tolerance curve derived from it, and its regulatory use fixes a pass value for a specified device in a specified test.
What it does not describe is rotational loading, central to diffuse brain injury mechanisms and not captured by a linear acceleration measure. Angular kinematic metrics exist and are actively debated. Presenting HIC as a general index of brain injury risk claims more than the criterion was built to support.
Nij and the multi-axis problem
The neck injury criterion normalises simultaneous axial force and sagittal-plane moment against critical intercept values, producing separate terms for tension and compression combined with flexion and extension; in regulatory form a combined value of one is the pass boundary. Its merit is recognising what single-axis limits miss: a neck can sit below every individual threshold and still be loaded injuriously in combination. Its limitation is that the intercepts are themselves scaled quantities, derived rather than measured on the population discussed.
Rate dependence
Ligament, tendon and intervertebral tissue are viscoelastic. Measured strength and failure mode change with the rate at which load is applied; the same tissue behaves differently under a slow stretch and a rapid one. Every tolerance value therefore carries an implicit loading rate, and comparing a slow reconstructed loading against a value derived at high rate is a category error that looks like arithmetic.
Equipment standards are pass or fail, not prediction
Protective equipment standards define a test, not a tolerance. The federal motorcycle helmet standard, the Consumer Product Safety Commission's bicycle helmet rule, ASTM specifications for bicycle and snow sport helmets, NOCSAE's athletic headgear standards and the European motorcycle regulation each prescribe an impact onto a specified anvil at a specified energy, using a specified headform, against a criterion the sample must not exceed.
Certification means samples passed that test. It does not mean the product prevents injury in the field, and non-certification does not establish an injury would have been avoided. Treating a pass criterion as a human tolerance threshold is the commonest error here, made in both directions.
Probability rather than a switch
Injury assessment is increasingly expressed as risk functions: a measured value corresponds to a stated probability of injury at a stated severity, coded on a scale such as the Abbreviated Injury Scale. That is more honest than a threshold, since it makes explicit that injuries occur below the value and are sometimes absent above it. It also narrows the defensible finding: the reconstructed loading corresponds to a particular injury risk, with an interval around it, for a stated population.
Using tolerance data in a dispute
Three questions expose most misuse. Which population does the cited value describe, and does the person in question resemble it. At what loading rate and direction was it derived, and does the reconstruction match. And is the figure a research risk function or a regulatory pass criterion — different objects, quoted interchangeably far too often. An opinion answering all three unprompted reads as careful; a bare number invites the cross-examination it will get.
This article is general technical orientation, not a failure analysis, an engineering opinion, or advice on any specific matter. Determining the cause of a particular incident requires hands-on examination by a credentialed expert.