Stature Estimation from Long Bones: Regression Formulae and Anatomical Methods

Every skeleton tells forensic investigators three basic facts about a person: their sex, their ancestry, and their stature. Of the three, stature estimation sounds deceptively simple, since it is really just about turning bone length into height. In practice, forensic anthropologists rarely find a complete, well-preserved skeleton. They usually work with fragments, weathered long bones, or remains missing key elements, which makes the choice of method just as important as the math behind it.

Two broad ways to estimate height from bone 🔗

Forensic anthropologists rely on two categories of methods. The first is the anatomical method, which reconstructs stature by adding up the height of every skeletal element that contributes directly to standing height, from skull to heel. The second is the mathematical or regression method, which uses statistical relationships between a single bone’s length and known stature in a reference population. Forensic standards from the Scientific Working Group for Forensic Anthropology note that the anatomical method gives the most accurate estimate of a person’s actual height at the time of death, while regression methods estimate a maximum possible living height and can overestimate stature, especially in older individuals whose height has decreased with age.

Fully’s anatomical method 🔗

The anatomical method used across forensic anthropology today traces back to Georges Fully, who developed it in 1956 while examining remains of concentration camp victims where identification mattered urgently. Fully’s method sums several measurements: basion-bregma height on the cranium, the height of the entire vertebral column, the height of the first sacral segment, oblique femoral length, tibial length, and the height of the talus and calcaneus at the ankle. Once these figures are added together, the formula for stature is:

Stature = 0.98 × (total skeletal height) + 14.63 ± 2.05 cm

This approach has held up well over decades of testing. A study revisiting Fully’s original technique on the Terry Collection, a well-documented set of skeletons with known cadaveric statures, found the method needed clearer measurement instructions but remained fundamentally sound once those were standardised, as described in this re-evaluation of the Fully technique. A later revision by Raxter and colleagues improved on the original further, producing a stature estimate closer to the true height in over 80 percent of individuals tested, according to a comparative accuracy study presented by the American Academy of Forensic Sciences. This Fully-Raxter approach is now considered the international benchmark for accuracy in stature estimation, as summarised in a recent review of stature estimation practice published on MDPI’s forensic anthropology journal.

Why the anatomical method has limits 🔗

The catch is right there in the formula. Fully’s method needs every single one of those skeletal elements present and measurable. If even one bone is missing, damaged, or too decomposed to measure, the entire calculation breaks down. This is a serious constraint in real forensic casework, where remains are commonly incomplete due to decomposition, animal activity, fire damage, or the sheer passage of time. This is exactly where mathematical, or regression-based, methods take over.

Regression analysis: predicting height from a single bone 🔗

Regression methods work on a straightforward statistical principle. Researchers measure a large sample of skeletons with documented stature, calculate the relationship between a specific bone’s length and that stature, and derive an equation that predicts height from bone length alone. Because this relationship differs across populations, sexes, and age groups, these formulae are never universal. A regression equation built on one population’s proportions applied to a different population can introduce meaningful error, which is why establishing sex, age, and ancestry is treated as a mandatory first step before any stature calculation is attempted.

This population-specificity is not a minor academic detail. A recent review of the field points out that genetic and environmental factors shape limb-to-height proportions differently across populations, and that using an outdated or mismatched reference sample is one of the biggest sources of error in modern stature estimation, according to this current status review of stature estimation methods. The most widely used sets of regression equations were developed by Mildred Trotter and Goldine Gleser in the 1950s using American war-dead skeletons, and although these remain a standard reference, later researchers have shown that secular changes in body proportions since then mean the equations sometimes need adjustment for modern populations, as discussed in this comparison of Trotter and Gleser’s stature equations.

Rebuilding a broken bone before estimating height 🔗

Regression methods assume a complete bone is available to measure. But long bones recovered from trauma, fire, or scavenging are frequently broken into pieces. To handle this, researchers including Steele developed techniques for reconstructing the full length of a bone from its fragments before applying a stature formula. For the humerus specifically, the bone is conceptually divided into four segments based on anatomical landmarks: from the top of the head down to the distal circumference point, from there to the near edge of the olecranon fossa, across the olecranon fossa itself, and finally from the far edge of the olecranon fossa down to the trochlea. Depending on which of these segments survive intact, a specific regression formula reconstructs the bone’s original maximum length.

This kind of fragment-based reconstruction has been tested and refined for various populations, including within India. A study on south Indian humeri divided the bone into five landmark-based fragments and derived population-specific regression equations, concluding that the longest surviving fragment gave the most reliable reconstruction of total humeral length, as reported in this study on humerus fragment reconstruction. Once the complete bone length is estimated this way, a second regression equation converts that length into an estimated stature.

Population-specific formulae in practice 🔗

Because bone-to-height proportions vary by ancestry and sex, forensic anthropologists maintain separate regression equations for different reference populations. Using the humerus as an example, commonly cited formulae include the following, all producing stature in centimetres from humerus length in centimetres:

• White males: stature = 3.08 (humerus) + 70.45 ± 4.05 cm
• Black males: stature = 3.26 (humerus) + 62.10 ± 4.43 cm
• White females: stature = 3.36 (humerus) + 57.97 ± 4.45 cm
• Mongoloid males: stature = 2.68 (humerus) + 83.19 ± 4.25 cm
• Mexican males: stature = 2.92 (humerus) + 73.94 ± 4.24 cm

Notice that the coefficients and constants shift noticeably between groups. Applying a White male formula to a skeleton of Mongoloid ancestry, for instance, would introduce systematic error because the underlying limb-to-trunk proportions differ. Similar population and sex-specific formulae exist for the radius and ulna, giving forensic anthropologists options depending on which bones survive in a given case. None of these equations produce an exact figure; they generate a range within which the person’s living stature likely falls, along with a margin of error expressed as a plus-or-minus value.

What Indian forensic research adds 🔗

Indian researchers have built on this framework using regional skeletal samples rather than relying only on formulae derived from American or European collections. A study on a north Indian population measured humerus and tibia length against known stature and found that combining both bones in a multiple regression model improved accuracy over using either bone alone, producing separate equations for males and females, as detailed in this gender-based stature estimation study from north India. This kind of population-specific calibration matters because India’s population is genetically and geographically diverse, and a single formula developed elsewhere is unlikely to fit every regional group equally well.

Reading the limitations honestly 🔗

Regression-based stature estimates come with built-in uncertainty that forensic anthropologists must communicate clearly in their reports. These formulae estimate a maximum living height rather than the person’s actual height at death, and they do not automatically account for the height loss that occurs naturally with advancing age due to spinal compression and posture changes. For older individuals, an uncorrected regression estimate can therefore run higher than the true stature at time of death. This is one more reason the anatomical method, when a complete skeleton is available, remains the preferred choice whenever circumstances allow it.

Long bones do more than hint at height. Their thickness and cross-sectional shape can also indicate a person’s general build and even handedness, since the dominant arm typically shows slightly greater bone density and robusticity from repeated mechanical loading. Combined with sex, age, and ancestry estimation, stature analysis becomes one more piece connecting a set of bones back to an individual identity.

What do you think? If forensic anthropologists know that regression formulae only give a height range rather than an exact figure, how much weight should that range carry in a legal identification compared to DNA or dental records? And given how different population-specific formulae can be, should India invest more heavily in building its own large-scale skeletal reference database rather than adapting formulae developed abroad?

References

1. https://www.nist.gov/system/files/documents/2018/03/13/swganth_stature_estimation.pdf
2. https://pubmed.ncbi.nlm.nih.gov/16425177/
3. https://www.aafs.org/sites/default/files/media/documents/AAFS-2019-A11.pdf
4. https://www.mdpi.com/2673-6756/6/1/23
5. https://academic.oup.com/fsr/article/8/1/16/7084858
6. https://pubmed.ncbi.nlm.nih.gov/22847047/
7. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12818479/