J Forensic Sci, July 2009, Vol. 54, No.

4
doi: 10.1111/j.1556-4029.2009.01048.x
Available online at: www.blackwell-synergy.com

Bradley J. Adams,1 Ph.D. and Nicholas P. Herrmann,2 Ph.D.

Estimation of Living Stature from Selected

Anthropometric (Soft Tissue) Measurements:
Applications for Forensic Anthropology

ABSTRACT: Estimation of living stature has obvious utility in the identification process. Typically, anthropologists estimate stature from the
measurement of long bone length. This type of analysis is traditionally conducted on skeletonized or badly decomposed remains, so collection of the
necessary bone measurements is relatively simple. As the role of anthropologists expands into medical examiner offices and mass fatality incidents,
the analysis of fleshed bodies and body parts is a more common scenario. For stature estimation in these types of cases (e.g., analysis of body por-
tions recovered from an aircraft crash site or from intentional dismemberment), the presence of soft tissue on the human remains would usually
necessitate dissection to expose skeletal elements to derive metric data for stature estimation. In order to circumvent this step, this paper provides var-
ious formulae that allow for standard anthropometric (i.e., soft tissue) measurements to be used in place of skeletal measurements. Data were com-
piled from several anthropometric studies (National Health and Nutrition Examination Survey [NHANES] and U.S. Army Anthropometric Survey
[ANSUR]) and numerous regression models are presented. Results are compared between skeletal measurements and the anthropometric measure-
ments from each study. It was found that the ANSUR models are similar to the skeletal models, while the NHANES models exhibit weaker correla-
tion coefficients and higher standard errors. Overall, this study finds that stature estimates derived from anthropometric data provide good results and
remove the necessity for dissection when working with fleshed body portions.

KEYWORDS: forensic science, forensic anthropology, stature estimation, mass fatality, dismemberment, identification, NHANES,
ANSUR

Estimation of living stature is one of the core components in a technique uses a summation of measurements taken on the major
forensic anthropologist’s analysis of unidentified remains. This elements, along with a correction value to account for cartilage and
information may be useful for decedent identification, especially soft tissue. Stature estimates based on the anatomical method result
when individuals are observed to be particularly tall or short in in a single point estimate and they are not dependent on knowing
comparison with their associated population. As forensic anthropol- the ancestry or sex of the individual. When feasible, stature esti-
ogists routinely work with skeletonized or badly decomposed mates based on the anatomical method, especially Fully’s method,
bodies, the collection of osteometric data is generally very straight- have been shown to be the most reliable techniques available (3,4).
forward and stature estimation is often one of the easiest compo- Recently, Fully’s technique has been revised by Raxter and col-
nents to generate for the biological profile. When dealing with leagues (5,6) through better measurement definitions and an adjust-
fleshed remains, the collection of osteometric data can be more ment to the soft tissue correction value. The main problem with the
challenging as there is the need for soft tissue dissection. One way anatomical technique is that it necessitates the presence of a rela-
to bypass this requirement would be the use of anthropometric (i.e., tively complete and well-preserved skeleton, something that is not
soft tissue) measurements for the estimation of living stature. This always available in forensic contexts. Clearly its advantages include
paper attempts to provide numerous regression models based on the accuracy of the estimates and the lack of sex or population
anthropometric data which may be useful for the estimation of liv- effects.
ing stature from fleshed body parts. The mathematical method, on the other hand, is not reliant on
Historically, there have been two different approaches to the esti- access to a complete skeleton. It uses linear regression to estimate
mation of stature from skeletal remains: the anatomical method and stature from a single bone or multiple bones. Most frequently these
the mathematical method (1). The anatomical methods derive stat- estimates are derived from maximum length measurements of long
ure estimates from numerous skeletal elements. One of the earliest bones that are input into appropriate regression models. Maximum
proponents of the anatomical method was Dwight, who recom- length measurements of long bones have been shown to be straight-
mended that the skeleton should be meticulously rearticulated on a forward and easy to measure (7), making stature estimation from
table using clay to account for soft tissue, resulting in a close complete long bones a highly replicable process that is not subject
approximation of living stature (1). This is obviously a very tedious to significant interobserver error. One notable exception for skeletal
process that requires considerable effort on the part of the analyst measurements is the difficulty with measurements of the tibia (8,9).
and completeness of the skeleton. A more straightforward variant Some early studies examining the relationship between long
of the anatomical method is that of Fully (2). Fully’s anatomical bone length and stature can be traced back to the French studies of
Rollet and Manouvrier (1), but it was Pearson (10) who first
1
Office of Chief Medical Examiner, 520 First Ave., New York, NY applied linear regression to the estimation of stature from bone
10016.
2
Department of Anthropology and Middle Eastern Cultures, Mississippi
length. The relational tables of Manouvrier and the regression mod-
State University, Mississippi State, MS 39762. els of Pearson were derived from samples of French cadavers in
Received 25 April 2008; and in revised form 17 Aug. 2008; accepted 21 the late 1800s, and these were the common standards in anthropo-
Aug. 2008. logical analyses in the U.S. until the seminal works of Trotter and

 2009 American Academy of Forensic Sciences 753

754 JOURNAL OF FORENSIC SCIENCES

Gleser (11,12) provided more robust models applicable to the U.S. elderly and disabled) who are not able to stand erect. Regression
population. Since Trotter and Gleser’s publications, regression mod- models are available in the published literature for these scenarios,
els for stature estimation have been developed for a wide range of many with very high correlations cited (e.g., [28–33]). Most of
bones and bone fragments, thus providing anthropologists with these regression equations use age as a variable in the formulae
numerous options depending on the composition of their case (e.g., and are intended for use on the elderly. One notable exception is a
13–18). study by Chumlea and colleagues (29) who utilized data from a
Unlike the anatomical models, the regression (i.e., mathematical) series of national health surveys conducted between 1960 and
models for stature estimation are usually dependent on knowing the 1970. From these data, they present regression models for estimat-
sex and ancestry of the unidentified individual. Studies have shown ing stature of white and black men and women (18–60 years) and
that models developed for one population may not provide reliable boys and girls (6–18 years) from knee height. The models pre-
estimates when applied to another population (e.g., 19,20). Input- sented in this study for the men and children could be applied to
ting measurement data into the regression models will produce a forensic contexts, but the models for women incorporate age as a
point estimate (the predicted stature). In order to account for nor- variable in the models. For other clinical models that incorporate
mal variation in the population, the associated standard error (SE) age into their regression models, they are not readily applicable to
of the estimate should be used to report a prediction interval. most forensic scenarios. Obviously, patients in the clinical setting
Proper calculation of a prediction interval is addressed by Giles are of known age and the intent of the stature estimate is far differ-
and Klepinger (21). Commonly, prediction intervals are reported at ent from the forensic context. For forensic investigations in which
either the 90% or 95% level. stature would need to be calculated, the identity of the decedent is
Most of the stature models (anatomical and mathematical) esti- unknown or tentative. To this end, it is not appropriate to use
mate living, or measured, stature. For many of the most frequently known age as a variable in the regression model because it may
utilized models, the known stature of the subjects was based on only be possible to provide a very general age estimate (especially
cadaver height (e.g., Terry and Hamann-Todd Collections) or data in circumstances of body fragmentation).
derived from military records (e.g., WWII and Korean War dead). Additional anthropometric studies have been conducted in order
Studies have shown that people tend to misrepresent their true stat- to evaluate the health and nutrition status of populations. One of
ure when self-reporting (22,23). For many forensic scenarios, ante- the larger studies in the U.S. is the National Health and Nutrition
mortem stature is based on self-reported information (e.g., driver’s Examination Survey which has been conducted periodically since
license) or information provided by a friend or relative. There is the 1960s. Data, including anthropometric measurements, have
clearly more error involved in these types of reference data as com- been routinely collected from a representative sample of the popu-
pared with measured data. For this reason, Ousley (24) differenti- lation in order to track the country’s health and nutrition status.
ates ‘‘forensic’’ stature from ‘‘measured’’ stature. He suggests that Specifically, the National Health and Nutrition Examination Survey
appropriate models should be utilized that are dependent on the (NHANES) body measurement data are used to track the preva-
scenario at hand. For example, a missing U.S. soldier would likely lence of obesity and to examine associations between body weight
have a measured stature in his or her medical records, while a rou- and the health and nutritional status of the U.S. population. These
tine medical examiner case might rely on less reliable antemortem data are available via the internet from the Centers for Disease
information, such as a driver’s license or missing person report. For Control and Prevention and provide a valuable dataset for
the models presented in this paper, which are based on anthropo- research.
metric measurements, the data are derived from ‘‘measured’’ stature Anthropometric studies have also been conducted in order to
of living subjects and this should be considered during their appli- understand body proportions and dimensions. Some of the most
cation to forensic casework. ambitious anthropometric studies for this purpose have been under-
taken on military personnel in order to fully understand the body
sizes and proportions of active duty personnel. The anthropometric
Overview of Soft-Tissue Data
data are critical as a guide for the design and sizing of clothing,
While most forensic anthropologists focus their analyses on skel- personal protective equipment, military weapon systems, and work
etal data when deriving stature estimates, the incorporation of stations (34). One of the most extensive studies was the U.S. Army
anthropometric (i.e., soft tissue) data into these estimates is a logi- Anthropometric Survey (ANSUR) conducted in 1987 and 1988.
cal step for forensic applications and one that has been underuti- Close to 9000 subjects were measured and over 130 measurements
lized. Today, the role of the forensic anthropologist is not solely were collected (34).
restricted to dry bones. For example, anthropologists play integral As noted previously, few studies have looked at the use of
roles in the resolution of mass fatality events and medical examiner anthropometric data for forensic purposes. A few notable examples
casework. More frequently than not, the decedents in these scenar- include Ozaslan and colleagues (35,36) and Attallah and Marshall
ios will still have soft tissue present. Certainly many types of (37). The studies of Ozaslan and colleagues were both based on a
anthropological analyses, such as an assessment of skeletal trauma, reference population of 202 males and 108 females from Istanbul,
require dissection and maceration of soft tissue in order to allow a Turkey. One study focused on the measurements of the upper
thorough evaluation, but this may not be necessary for the estima- extremity, while the other dealt with several leg and foot measure-
tion of living stature. ments. In these studies, the authors found that living stature could
Anthropology has a long history relating to the collection of metric be accurately estimated from body segments and they present
data from living subjects (e.g., [25–27]), but its incorporation into the regression models for estimation. The study of Attallah and Mar-
forensic arena has been limited. Generally, anthropometric data has shall focused on stature estimation from limb segments of children
been utilized as a gauge of a population’s health and nutrition status, between the ages of 4 and 18 years. Their anthropometric sample
to understand body proportions for the design of clothing and equip- was compiled in London and consists of 514 boys and 680 girls
ment, and by medical personnel to estimate stature of the disabled. described as ‘‘western European origin.’’ Numerous univariate and
Medical personnel may use anthropometric measurements in a multivariate regression models are presented by the authors to esti-
clinical setting to estimate stature of the living (particularly, the mate stature from subadult body measurements. Finally, there have
 ADAMS AND HERRMANN • STATURE ESTIMATION FROM ANTHROPOMETRIC MEASUREMENTS 755

been numerous studies that have explored the relationship between and can be located by placing sliding calipers against the distal
living statue and hand and foot length (e.g., [38–43]). end of the femur, as though you were measuring the breadth of
the patella. When positioning the calipers in this manner, a
mark can be placed on the anterior surface of the thigh, along
Materials and Methods
the horizontal bar of the calipers. This delineates the measure-
The goal of this paper is to provide numerous stature regression ment point associated with the distal femur ⁄ proximal patella.
models derived from large samples of anthropometric data which are
suitable for application within the U.S. Data from two large anthropo-
metric studies have been published and are available for research. ANSUR Natick Data 1987–1988
These include the NHANES study conducted by the National Center
In order to complement the NHANES data and account for addi-
for Health Statistics and the ANSUR conducted by the U.S. Army
tional scenarios involving fragmentary ⁄ dismembered body portions,
Natick Research, Development, and Engineering Center.
additional models based on anthropometric studies of active duty
U.S. Army personnel are reproduced from the ANSUR report con-
NHANES Data 1999–2006 ducted by the U.S. Army Natick Research, Development, and Engi-
neering Center (Technical Report Natick ⁄ TR-90 ⁄ 035). This study
The NHANES has been an on-going study conducted by the
was based on data compiled from active duty Army personnel in
National Center for Health Statistics (part of the Centers for Dis-
1987 and 1988 (34). A follow-up study in 1996 showed that this data
ease Control and Prevention) since the 1960s. The study group is
was still representative of the U.S. Army population in 1996 (45).
composed of a large, representative sample of civilians (adults and
Although the ANSUR anthropometric survey consisted of nearly
children) from across the continental U.S. There are many compo-
9000 individuals, the published results are based on the ‘‘working
nents to the study, only one of which is anthropometric data collec-
database,’’ which was selected to represent the various age and race
tion. In order to develop regression models for the estimation of
groups representative of the U.S. Army in 1988 (34). The working
stature, a subsample of the NHANES data was extracted from stud-
database consisted of 3982 individuals between 17 and 51 years of
ies conducted between 1999 and 2006. Only data pertaining to
age. Table 2 presents the sample composition by sex and ancestry.
adults between the ages of 18 and 50 years were selected
The summary statistics and published regression models are divided
(n = 14,548). Table 1 presents the sample composition by sex and
by sex, but all of the racial groups have been pooled as part of the
ancestry. Anthropometric variables utilized from the NHANES
original study (46). Formulae for the ANSUR data were published
study were Standing Height, Upper Arm Length (UPARMLTH),
in the Natick summary reports and several are reproduced here for
and Upper Leg Length (UPLEGLTH). For the NHANES data,
comparative purposes. The variables selected for possible forensic
measurement values for the limb portions were recorded to the
application include: Stature (STATURE), Buttock-Knee Length
nearest 0.1 cm using a measuring tape. The NHANES measure-
(BUTTKLTH), Forearm-Hand Length (FORHDLG), Knee Height-
ments are described as follows (44):
Sitting (KNEEHTSI), Shoulder-Elbow Length (SHOUELLT), Lat-
1. Standing Height (i.e., STATURE) is the maximum vertical size eral Femoral Epicondyle Height (LATFEMEP), Span (SPAN), Foot
from the bottom of the feet to the top of the head. The mea- Length (FOOTLGTH), and Hand Length (HANDLGTH). For the
surement is taken with a fixed stadiometer with a vertical back- ANSUR data, the measurements are all taken with calipers or an-
board and a movable headboard. thropometers and are described as follows (34):
2. UPARMLTH is measured with a tape measure. The upper most
1. STATURE is the vertical distance from a standing surface to
edge of the posterior border of the acromion process of the
the top of the head. The subject stands erect with the head in
scapula is marked, and the tape measure is extended down the
the Frankfort plane and the measurement is taken with an
posterior surface of the arm to the tip of the olecranon process
anthropometer.
of the ulna. The measurement is taken with the arm flexed 90
2. BUTTKLTH is the horizontal distance between the most poster-
(Fig. 1).
ior point on the buttock and the anterior point of the knee mea-
3. UPLEGLTH is taken while the subject is in a sitting position
sured while the knee is flexed 90 (Fig. 1). This measurement
with the knee bent at 90. The measurement is taken with a
is taken along the lateral thigh with an anthropometer.
tape measure from the inguinal crease (just below the anterior
3. FOOTLGTH is the maximum length of the foot from the heel
superior iliac spine), along the anterior midline of the thigh, to
to the tip of the longest toe measured with calipers (Fig. 1).
the proximal patella (Fig. 1). The mark on the proximal patella
4. FORHDLG is the horizontal distance between the posterior sur-
is perhaps the most complicated landmark for this measurement
face of the elbow and the tip of the middle finger taken while
the elbow is flexed 90 (Fig. 1). It should be measured with a
caliper.
TABLE 1—Population samples from the NHANES data (18–50 years of 5. HANDLGTH is the distance from the stylion landmark (the tip
age).
of the styloid process of the radius) to the tip of the middle fin-
NHANES
ger, measured with a caliper (Fig. 1).
Samples BM BF WM WF HM HF Total 6. KNEEHTSI is the vertical distance between the bottom of the
foot and the suprapatellar landmark (Fig. 1). The suprapatellar
1999–2000 607 688 1080 1279 954 1132 5740
landmark is the superior surface of the patella and should be
2001–2002 343 371 619 735 475 524 3067
2003–2004 350 380 605 683 359 397 2774 marked while the leg is in a standing position. The overall Knee
2005–2006 379 420 571 665 425 507 2967 Height measurement, however, is taken with the knee flexed
Total 1679 1859 2875 3362 2213 2560 14548 90, measured with an anthropometer.
NHANES, National Health and Nutrition Examination Survey; BM, 7. SHOUELLT is the distance between the acromion landmark
Black male; BF, Black female; WM, White male; WF, White female; HM, (superior tip of the acromion process) of the shoulder and the
Hispanic male; HF, Hispanic female. olecranon landmark on the bottom of the elbow (Fig. 1). The
756 JOURNAL OF FORENSIC SCIENCES

SPAN

HANDLGTH

FORHDLG
UPARMLTH
&
SHOUELLT

UPLEGLTH

LATFEMEP BUTTKLTH
KNEEHTSI

FOOTLGTH

FIG. 1—Anthropmetric measurements examined from U.S. Army Anthropometric Survey (ANSUR) and National Health and Nutrition Examination Survey
(NHANES).

TABLE 2—Population samples from the ANSUR data. TABLE 3—Regression coefficients for stature estimation from UPARMLTH
and UPLEGLTH measurements from the NHANES data (1999–2006).
ANSUR
Samples n Intercept Slope Error r
1987– Asian ⁄ American-
1988 White Black Hispanic Pacific Indian Other Total UPARMLTH
All (unknown sex 14,540 61.27 2.88 6.04 0.80
Male 1172 458 68 28 12 36 1774 and group)
Female 1140 922 58 32 14 42 2208 Sex (group unknown)
Total 2312 1380 126 60 26 78 3982 Male 6763 83.25 2.38 5.56 0.72
Female 7777 88.80 2.05 5.20 0.66
Group (sex unknown)
measurement should be taken while the elbow is flexed 90 and White 6235 65.68 2.81 5.84 0.79
measured with a caliper. Note that UPARMLTH in the Black 3535 66.84 2.72 6.36 0.76
NHANES study utilizes the same landmarks, but the NHANES Hispanic 4770 60.26 2.87 5.53 0.80
Group and sex known
measurement is taken with a measuring tape. White male 2873 95.87 2.10 5.30 0.66
8. LATFEMEP is the vertical distance between the bottom of the White female 3362 97.02 1.88 4.91 0.62
foot and the lateral femoral epicondyle (most laterally projecting Black male 1677 98.48 2.00 5.34 0.67
point on the lateral femoral condyle at the knee pivot point). Black female 1858 100.65 1.72 5.01 0.61
The measurement is taken with an anthropometer while the leg Hispanic male 2213 83.71 2.31 5.13 0.71
Hispanic female 2557 84.91 2.11 4.74 0.66
is in a standing position (Fig. 1). UPLEGLTH
9. SPAN is the distance between the tips of the middle fingers All (unknown sex 14,350 85.56 2.02 6.51 0.76
when the arms are horizontally outstretched (Fig. 1). For living and group)
subjects, this is taken while the subject is standing against a Sex (group unknown)
Male 6680 103.77 1.67 5.77 0.69
measured wall chart. Although not formally tested, it is hypothe- Female 7670 109.85 1.34 5.45 0.61
sized that an anthropometer may provide a comparable Group (sex unknown)
measurement. White 6169 86.03 2.05 6.37 0.74
Black 3483 84.18 2.02 6.58 0.74
Hispanic 4698 87.74 1.94 6.04 0.75
Group and sex known
Results White male 2849 115.26 1.45 5.55 0.61
Using the NHANES data, it was possible to construct anthropo- White female 3320 114.02 1.28 5.19 0.56
Black male 1656 105.83 1.61 5.35 0.67
metric regression equations for stature estimation. Models were Black female 1827 118.24 1.11 5.26 0.55
derived for White, Black, and Hispanic groups by sex. In addition, Hispanic male 2175 105.58 1.57 5.41 0.67
models are presented for circumstances when ancestry and ⁄ or sex Hispanic female 2523 109.88 1.28 5.03 0.61
are unknown. The regression models are presented in Table 3. In UPLEGLTH, upper leg length; UPARMLTH, upper arm length;
addition, numerous regression models using the selected measures NHANES, National Health and Nutrition Examination Survey.
derived from the ANSUR study are presented in Table 4. Note that All measurements in cm.
 ADAMS AND HERRMANN • STATURE ESTIMATION FROM ANTHROPOMETRIC MEASUREMENTS 757

TABLE 4—Regression coefficients for stature estimation from various anthropometric data has a higher r-value and a lower SE than the
measurements from the U.S. Army ANSUR data (1987–1988). osteometric data. For the ANSUR anthropometric data, the correla-
tion for SHOUELLT = 0.82 and the SE is 3.84. This would sug-
Intercept Slope Error r
gest that the ANSUR model outperforms the osteometric model of
Males, n = 1774 Trotter and Gleser.
BUTTKLTH 64.94 1.80 3.99 0.80 When comparing the data for the upper leg, the results are com-
FOOTLGTH 79.24 3.57 4.77 0.70
FORHDLG 71.70 2.15 4.43 0.75
parable with the arm. Comparison of the NHANES model for UP-
HANDLGTH 89.58 4.44 5.08 0.65 LEGLTH (Table 3) reveals an r-value of 0.61 for WM and an
KNEEHTSI 57.17 2.12 3.11 0.89 r-value of 0.67 for BM. The NHANES values for the SE are 5.55
LATFEMEP 64.69 2.21 3.30 0.87 for WM and 5.35 for BM. The Trotter and Gleser correlation
SHOUELLT 63.10 3.05 3.84 0.82 values for Femur Length are considerably higher (Table 5), with
SPAN 54.45 0.66 3.87 0.83
Females, n = 2208 values of 0.80 for WM and 0.81 for BM. Trotter and Gleser’s error
BUTTKLTH 68.91 1.60 4.25 0.74 values are also lower, with 4.04 reported for WM and 3.83 reported
FOOTLGTH 77.40 3.50 4.71 0.67 for BM. The ANSUR data for pooled males and BUTTKLTH
FORHDLG 77.14 1.94 4.47 0.71 (Table 4) reveals an r-value of 0.80 and a SE of 3.99, both of
HANDLGTH 87.58 4.18 4.91 0.64
KNEEHTSI 56.23 2.07 3.28 0.86
which are nearly identical to the osteometric values reported by
LATFEMEP 62.36 2.18 3.37 0.85 Trotter and Gleser.
SHOUELLT 64.97 2.92 3.83 0.80 To illustrate the correlation differential between osteological and
SPAN 59.92 0.62 3.92 0.79 soft tissue measures, plots of the NHANES data and a combined
ANSUR, Anthropometric Survey; BUTTKLTH, Buttock-Knee Length; skeletal sample from Trotter’s Terry and World War II data are
FOOTLGTH, foot length; FORHDLG, forearm-hand length; HANDLGTH, presented in Figs. 2 and 3. Figure 2 pertains to UPLEGLTH and
hand length; KNEEHTSI, Knee Height-Sitting; LATFEMEP, Lateral Femo- shows the relationship of WM from NHANES and Trotter’s bone
ral Epicondyle Height; SHOUELLT, Shoulder-Elbow Length. samples. In this example, bone length relative to stature produces a
All measurements in cm.
stronger correlation. The distribution of the NHANES UPLEGLTH
relative to stature is far more dispersed. Similarly, the comparison
the raw ANSUR data were not available for reanalysis and the of Trotter’s humerus length data to NHANES UPARMLTH data
models presented require that sex is known. show the same relationship (Fig. 3). It is also apparent that the
In order to observe the difference between models derived from NHANES UPARMLTH correlation with stature is stronger than
anthropometric measurement data and models derived from long the UPLEGLTH. In both cases, the NHANES fit lines roughly par-
bone length, it was possible to compare the results from the allel the bone models and suggest a similar relationship to stature
NHANES and ANSUR data with data presented by Trotter and albeit shifted up or down along the y-axis. Although not graphi-
Gleser (12). Table 5 presents the correlation values between stature cally plotted, the ANSUR anthropometric data are more consistent
and maximum lengths of the humerus and femur from Trotter and with the skeletal data and provide tighter prediction models than
Gleser’s data. Although the anthropometric data does not corre- seen with the NHANES sample.
spond precisely to these bones, the models for the upper arm and Initially, it may seem unusual that the anthropometric studies
leg are comparable. For the NHANES data, the correlations and provide such divergent results for very comparable measurements.
SE of UPARMLTH and UPLEGLTH can be compared. For the This could be an indication that there are population differences
ANSUR data, these can be compared with BUTTKLTH and between the NHANES (civilian) sample and the ANSUR (military)
SHOUELLT. sample, even though the demographic composition is similar. A
When comparing Trotter and Gleser’s 1958 Korean War data for comparison of the ANSUR and NHANES distributions by sex
the humerus (Table 5) to the NHANES model for UPARMLTH finds that the means are very similar, although there are differences
(Table 3), it is clear that the long bone measurements are more between the samples. The basic statistics by sex for both samples
strongly correlated with stature than the corresponding soft tissue are presented in Table 6. Overall, male mean stature in the two
models. For the NHANES anthropometric data, the correlation (r-
values) for White male (WM) = 0.66 and for Black male
(BM) = 0.67. For Trotter and Gleser’s osteometric data, the correla-
tion for WM = 0.73 and for BM = 0.76. When comparing the SE
of the estimate values, the NHANES figures for WM = 5.30 and
BM = 5.34, while the Trotter and Gleser values are lower with
WM = 4.61 and BM = 4.26.
Interestingly, when the most comparable data is observed from
the ANSUR sample (pooled males) for SHOUELLT (Table 4), the

TABLE 5—Correlation (r) and SE values of humerus and femur with

stature in 18- to 46-year-old males from the Korean War (12: Table 3).

White Black

r SE r SE
Humerus* 0.73 4.61 0.76 4.26
Femur* 0.80 4.04 0.81 3.83
SE, standard error. FIG. 2—Comparison of upper leg and femur length from NHANES and
*Values represent the right side only. Trotter’s World War II and Terry Collection measures for White males.
758 JOURNAL OF FORENSIC SCIENCES

will have more adipose tissue, which may artificially inflate the
length measurement if a tape is extended along the soft tissue of
the arm or leg.
BMI is a variable that is present within the NHANES dataset,
which allowed for the opportunity to test this theory. The strength
of correlation (r-values) was explored for UPLEGLTH and UP-
ARMLTH according to the participants’ BMI. Table 7 shows that
the heaviest individuals (BMI > 30) have the lowest correlation
with Standing Height. One of the NHANES manuals provides con-
firmation of this supposition. It states, ‘‘Analysts should examine
the distributions of the body measurements carefully. In particular,
the upper arm length…and upper leg length…values are affected
by extreme amounts of adipose tissue’’ ([47], p. 3).

Recommendations and Conclusions

For this paper, large datasets of anthropometric data were ana-
lyzed or reviewed in order to present regression models that could
FIG. 3—Comparison of upper arm and humerus length from NHANES be useful for the forensic scientist in the estimation of living stature
and Trotter's World War II and Terry Collection measures for White males.
from fragmentary bodies. It is anticipated that these formulae may
samples is not significantly different (t = 0.118, p = 0.906, be applicable to mass fatality incidents resulting in body fragmenta-
d.f. = 3228). The female means differ by less than one centimeter; tion (e.g., aircraft crashes) or in cases of intentional body dismem-
however, this difference is statistically significant (t = 4.955, berment. These models are presented in Tables 3 and 4 and they
p < 0.0001, d.f. = 3801). These distributions of the two samples contain the appropriate values to calculate stature estimates from
are plotted in Fig. 4. The ANSUR sample for both males and various body segments. For the NHANES data, numerous options
females does appear more tightly clustered around the mean and are made available depending on the case parameters, such as
this observation is supported by the smaller standard deviations knowledge of the decedent’s sex or ancestry. For the ANSUR data,
(SD) in the ANSUR samples as compared with NHANES. It is only sex-specific models are available. Because of the very large
suspected that lower variation in overall stature in the ANSUR sample sizes in both the NHANES and the ANSUR studies, it is
sample may be one contributing factor to the higher correlation very straightforward to calculate a point estimate and the associated
reported in the study. confidence intervals from the presented models.
While there is validity to this assertion, another likely hypothesis For example, take a scenario where portions of a dismembered
for the difference may simply be related to the methods of mea- male body are discovered. There is a right leg present from the
surement involved in each study. As pointed out by Gordon and proximal femur to the foot. An appropriate measurement to use
colleagues ([34], p. 623) in their discussion of the ANSUR study, would be the LATFEMEP from the ANSUR data (note that this
‘‘Differences in landmark definitions, subject positioning, instru- measurement should be taken with calipers). The regression model
ments and their techniques of use can and do lead to significantly is:
different results.’’ For the NHANES study, the measurements were STATURE ¼ LATFEMEPð2:21Þ þ 64:69
taken with a tape measure extended between specific landmarks.
The ANSUR data, on the other hand, were all collected with an- Assuming that the measurement is 56.0 cm, the resulting point
thropometers or calipers. It is suspected that the use of calipers in estimate for the individual’s stature would be 188.45 cm, or 74.2
the data collection process provides more precise values that are inches. Because of the large sample sizes associated with both the
more closely related to actual skeletal dimensions. When tape mea- NHANES and ANSUR data, it is possible to derive a prediction
sures are utilized, Body Mass Index (BMI) may have an effect on interval directly from the SE of the estimate. For a 90% prediction
the overall measurement. For example, a person with a high BMI interval, simply multiply the SE of the estimate, in this case 3.30,
by 1.645. For a 95% prediction interval, the error value should be
TABLE 6—NHANES and ANSUR sample statistics by sex.
multiplied by 1.96. Adding and subtracting the result to the point
Samples estimate will provide the desired prediction interval. In this exam-
ple, the 90% prediction interval would be 183.05–193.85 cm
Sex NHANES ANSUR (72.1–76.3 inches).
Males As shown in this paper, there appears to be a marked difference
n 6767 1774 in the anthropometric regression models which we believe is based
Mean 175.60 175.58 largely on the type of measurement equipment utilized in the data
SD 7.98 6.68 collection process. The NHANES data were collected with a tape
Kurtosis 0.02 3.09
CV 4.54 3.80 measure that was extended along the body, while the ANSUR data
Females was collected with calipers. It is apparent that BMI does have an
n 7781 2208 effect on the NHANES data, likely skewing measurements when
Mean 162.17 162.94 significant amounts of adipose tissue are present. Another contribut-
SD 6.89 6.36
ing factor may be the sample compositions of the NHANES and
Kurtosis 0.08 3.01
CV 4.25 3.90 ANSUR studies. For the NHANES group, there appears to be
greater dispersion around the sample’s mean stature (indicated by
NHANES, National Health and Nutrition Examination Survey; ANSUR, the higher SD and coefficient of variation [CV] values) than
Anthropometric Survey; SD, standard deviation; CV, coefficient of variation.
 ADAMS AND HERRMANN • STATURE ESTIMATION FROM ANTHROPOMETRIC MEASUREMENTS 759

20
ANSUR
NHANES
18

16

14

12

Percent
10

0
144

147

150

153

156

159

162

165

168

171

174

177

180

183

186

189

192

195

198

201

204
Male Stature (cm)

20
ANSUR
NHANES
18

16

14

12
Percent

10

0
129

132

135

138

141

144

147

150

153

156

159

162

165

168

171

174

177

180

183

186

189

Female Stature (cm)

FIG. 4—Distribution of NHANES and ANSUR statures for males and females.

TABLE 7—BMI correlations for UPARMLTH and UPLEGLTH from the observed with the ANSUR group. This suggests that there is
NHANES data. greater variability in stature among the NHANES participants.
Based on the results of this study, anthropometric measurements
BMI Correlations (n) of certain body portions are suitable for the estimation of living
stature and are applicable to forensic contexts within the U.S. Use
Limb <20 20–25 25–30 >30
of these regression models removes the need for soft tissue dissec-
UPARMLTH tion in order to obtain skeletal measurements. While radiographic
Male 0.749 (456) 0.741 (2249) 0.748 (2451) 0.713 (1700) alternatives may also be feasible for stature estimation of fleshed
Female 0.713 (674) 0.722 (2336) 0.713 (2198) 0.691 (2662)
UPLEGLTH remains, the use of the anthropometric techniques presented in this
Male 0.722 (426) 0.716 (2213) 0.689 (2410) 0.650 (1652) paper is an expedient option (simply relies on the use of calipers or
Female 0.606 (652) 0.663 (2302) 0.671 (2163) 0.539 (2586) a tape measure) and provides results of comparable accuracy to
BMI, Body Mass Index; NHANES, National Health and Nutrition Exami-
skeletal techniques. Numerous stature estimation formulae are
nation Survey; UPARMLTH, upper arm length; UPLEGLTH, upper leg presented based on the NHANES and ANSUR anthropometric
length. data. Although the NHANES models do not perform as well as
760 JOURNAL OF FORENSIC SCIENCES

comparable osteometric equations, the ANSUR models meet or 24. Ousley SD. Should we estimate biological or forensic stature? J Forensic
exceed the skeletal models. For this reason, it is encouraged that Sci 1995;40(5):768–73.
25. Hrdlicka A. Anthropometry. Philadelphia, PA: Wistar Institute of Anta-
anthropometers and calipers should be used along with the ANSUR tomy and Biology, 1920.
models if feasible. The NHANES models will still provide useful 26. Boas F. Remarks on the theory of anthropometry. Publications of the
information if access to the proper instrumentation is not possible American Statistical Association 1893;3(24):569–75.
or for situations in which sex is unknown. If the NHANES models 27. Boas F. Report on an anthropometric investigation of the population of
the United States. Race, Language and Culture. New York, NY: Mac-
are to be utilized, the measurements should be taken with a tape
millan, 1940. 28–59.
measure. 28. Chumlea WC, Guo S. Equations for predicting stature in white and
black elderly individuals. J Gerontol 1992;47(6):197–203.
29. Chumlea WC, Guo SS, Steinbaugh ML. Prediction of stature from knee
Acknowledgments height for black and white adults and children with application to
mobility-impaired or handicapped persons. J Am Diet Assoc 1994;
The authors would like to thank the Ellis R. Kerley Forensic 94(12):1385–91.
Sciences Foundation for their recognition of the importance of 30. Chumlea WC, Roche AF, Steinbaugh ML. Estimating stature from knee
this research. The authors would also like to thank the numer- height for persons 60 to 90 years of age. J Am Geriatr Soc 1985;
ous individuals who took an interest in this project, including 33(2):116–20.
Frank and Julie Saul, P. Willey, and John Byrd. The authors 31. Han TS, Lean ME. Lower leg length as an index of stature in adults. Int
J Obes Relat Metab Disord 1996;20(1):21–7.
would like to acknowledge Chris Rainwater for creating the 32. Johnson RK, Ferrara MS. Estimating stature from knee height for per-
anatomical figures. sons with cerebral palsy: an evaluation of estimation equations. J Am
Diet Assoc 1991;91(10):1283–4.
References 33. Prothro JW, Rosenbloom CA. Physical measurements in an elderly black
population: knee height as the dominant indicator of stature. J Gerontol
1. Stewart TD. Essentials of forensic anthropology. Springfield, IL: Charles 1993;48(1):15–8.
C. Thomas, 1979. 34. Gordon CC, Churchill T, Clauser CE, Bradtmiller B, McConville JT,
2. Fully G. Une nouvelle mthode de dtermination de la. taille. Ann Med Tebbetts I, et al. 1988 anthropometric survey of U.S. army personnel:
Legale 1956;35:266–73. methods and summary statistics. Technical Report NATICK ⁄ TR-
3. Lundy JK. The mathematical versus anatomical methods of stature esti- 89 ⁄ 044. Natick, MA: U.S. Army Natick Research, Development, and
mate from long bones. Am J Forensic Med Pathol 1985;6(1):73–6. Engineering Center, 1989.
4. Lundy JK. A report on the use of Fully’s anatomical method to estimate 35. Ozaslan A, Iscan MY, Ozaslan I, Tugcu H, Koc S. Estimation of stature
stature in military skeletal remains. J Forensic Sci 1988;33(2):534–9. from body parts. Forensic Sci Int 2003;132(1):40–5.
5. Raxter MH, Auerbach BM, Ruff CB. Revision of the Fully technique 36. Ozaslan A, Koc S, Ozaslan I, Tugcu H. Estimation of stature from upper
for estimating statures. Am J Phys Anthropol 2006;130(3):374–84. extremity. Mil Med 2006;171(4):288–91.
6. Raxter MH, Ruff CB, Auerbach BM. Technical note: revised fully stat- 37. Attallah NL, Marshall WA. The estimation of stature from anthropomet-
ure estimation technique. Am J Phys Anthropol 2007;133(2):817–8. ric and photogrammetric measurements of the limbs. Med Sci Law
7. Adams BJ, Byrd JE. Interobserver variation of selected postcranial skel- 1986;26(1):53–9.
etal measurements. J Forensic Sci 2002;47(6):1193–202. 38. Agnihotri AK, Purwar B, Googoolye K, Agnihotri S, Jeebun N. Estima-
8. Jantz RL, Hunt DR, Meadows L. Maximum length of the tibia: how did tion of stature by foot length. J Forensic Leg Med 2007;14(5):279–83.
Trotter measure it? Am J Phys Anthropol 1994;93(4):525–8. 39. Giles E, Vallandigham PH. Height estimation from foot and shoeprint
9. Jantz RL, Hunt DR, Meadows L. The measure and mismeasure of the tibia: length. J Forensic Sci 1991;36(4):1134–51.
implications for stature estimation. J Forensic Sci 1995;40(5):758–61. 40. Gordon CC, Buikstra JE. Linear models for the prediction of stature
10. Pearson K. Mathematical contributions to the theory of evolution. V. On from foot and boot dimensions. J Forensic Sci 1992;37(3):771–82.
the reconstruction of the stature of prehistoric races. Philos Trans R Soc 41. Krishan K, Sharma A. Estimation of stature from dimensions of hands
Lond 1899;192:169–244. and feet in a North Indian population. J Forensic Leg Med 2007;
11. Trotter M, Gleser GC. Estimation of stature from long bones of Ameri- 14(6):327–32.
can whites and Negroes. Am J Phys Anthropol 1952;10:463–514. 42. Ozden H, Balci Y, Demirustu C, Turgut A, Ertugrul M. Stature and sex
12. Trotter M, Gleser GC. A re-evaluation of estimation of stature based on estimate using foot and shoe dimensions. Forensic Sci Int 2005;147(2-
measurements of stature taken during life and of long bones after death. 3):181–4.
Am J Phys Anthropol 1958;16:79–123. 43. Sanli SG, Kizilkanat ED, Boyan N, Ozsahin ET, Bozkir MG, Soames R,
13. Byers S, Akoshima K, Curran B. Determination of adult stature from et al. Stature estimation based on hand length and foot length. Clin Anat
metatarsal length. Am J Phys Anthropol 1989;79(3):275–9. 2005;18(8):589–96.
14. Holland TD. Estimation of adult stature from fragmentary tibias. 44. Anthropometry and physical activity monitor procedures manual:
J Forensic Sci 1992;37(5):1223–9. National Health and Nutrition Examination Survey, 2005. Available at:
15. Meadows L, Jantz RL. Estimation of stature from metacarpal lengths. http://www.cdc.gov/nchs/data/nhanes/nhanes_05_06/BM.pdf.
J Forensic Sci 1992;37(1):147–54. 45. Gordon CC. U.S. Army anthropometric survey database: downsizing,
16. Simmons T, Jantz RL, Bass WM. Stature estimation from fragmentary demographic change, and validity of the 1988 data in 1996. Natick, MA:
femora: a revision of the Steele method. J Forensic Sci 1990;35(3):628– Army Natick Research, Development, and Engineering Center, 1996.
36. 46. Cheverud J, Gordon CC, Walker RA, Jacquish C, Kohn L, Moore A,
17. Steele D. Estimation of stature from fragments of long limb bones. In: et al. 1988 anthropometric survey of US army personnel: correlation
Stewart T, editor. Personal identification in mass disasters. Washington, coefficients and regression equations. Part 4. Bivariate regression tables.
DC: National Museum of Natural History, 1970;85–97. Technical Report NATICK ⁄ TR-90 ⁄ 035. Natick, MA: United States
18. Steele DG, Bramblett CA. The anatomy and biology of the human skel- Army Natick Research, Development, and Engineering Center, 1990.
eton. College Station, TX: Texas A&M University Press, 1998. 47. Documentation, codebook, and frequencies: body measurements—
19. Ross AH, Konigsberg LW. New formulae for estimating stature in the National Health and Nutrition Examination Survey 2005–2006,
Balkans. J Forensic Sci 2002;47(1):165–7. 2007. Available at: http://www.cdc.gov/nchs/data/nhanes/nhanes_05_06/
20. Konigsberg LW, Hens SM, Jantz LM, Jungers WL. Stature estimation bmx_d.pdf
and calibration: Bayesian and maximum likelihood perspectives in phys-
ical anthropology. Am J Phys Anthropol 1998;27(Suppl. 27): 65–92. Additional information:
21. Giles E, Klepinger LL. Confidence intervals for estimates based on linear Bradley J. Adams, Ph.D.
regression in forensic anthropology. J Forensic Sci 1988;33(5):1218–22. Forensic Anthropologist
22. Giles E, Hutchinson DL. Stature- and age-related bias in self-reported Office of Chief Medical Examiner
stature. J Forensic Sci 1991;36(3):765–80. 520 First Ave.
23. Willey P, Falsetti T. Inaccuracy of height information on driver’s New York, NY 10016
licenses. J Forensic Sci 1991;36(3):813–9. E-mail: [email protected]