AMERICAN JOURNAL OF PHYSICAL ANTHROPOLOGY 122:2327 (2003)
Body Height Estimation Based on Tibia Length in
Different Stature Groups
Izzet Duyar1* and Can Pelin2
1
Department of Physical Anthropology, Ankara University, 06100 Sihhiye, Ankara, Turkey
2
Department of Anatomy, Faculty of Medicine, Baskent University, 06530 Baglica, Ankara, Turkey
KEY WORDS tibia length; stature estimation; body proportions; forensic anthropology
ABSTRACT Long bone length is one of the best-known
indicators of human stature. Although the long bone
length/height ratio differs in tall and short individuals, no
detailed study has investigated whether specic formulae
should be used to calculate height in different stature
groups. This study proposes a new height estimation
method. Body height and tibia length were measured in
121 male subjects aged 18.0 34.3 years. Three subgroups
were established according to body height (short, medium,
or tall), using the 15th and 85th percentiles as cutoff
levels. The general formula and a group-specic regres-
sion formula were used to estimate height in each sub-
Estimating stature from dry bone length or from
the size of other body parts continues to be of par-
ticular interest to forensic and biological anthropol-
ogists, and to geriatricians. Several authors studied
this subject in different populations, and observed
signicant interpopulation differences in body pro-
portions (Eveleth and Tanner, 1990; Cheng et al.,
1996).
Long bone length is known to be the best indicator
of stature. However, it is known that the long bone
length/height ratio does vary to some degree with
differences in stature (Olivier, 1969; Sjvold, 1990;
Meadows and Jantz, 1995; Konigsberg et al., 1998).
The tibia is one of the commonly used long bones for
stature estimation. The tibia length/body height ra-
tio has been shown to vary among populations, and
even among individuals. In fact, Meadows Jantz and
Jantz (1999) noted that such proportional changes
are more pronounced in distal bones.
Many anthropologists and forensic scientists use a
single formula devised by Trotter and Gleser (1952,
1958) to estimate height. No detailed study has in-
vestigated whether different height calculations
should be used for different stature groups (tall,
medium, and short). The aim of this study was to
test the hypothesis that estimations of stature are
more accurate if different regression formulae are
used for specic stature groups.
SUBJECTS AND METHODS
The study involved 231 randomly selected Turk-
ish males, ranging in age from 18.0 34.3 years. The
2003 WILEY-LISS, INC.
group. A control group with the same properties as the
study group was analyzed in the same manner. Particu-
larly with short and tall subjects, the difference be-
tween true height and the height predicted by the group-
specic formulae was smaller than the difference observed
when the general formula was used. These discrepancies
were statistically signicant. When estimating height
based on tibia length, the individuals general stature
category should be taken into consideration, and group-
specic formulae should be used for short and tall sub-
jects. Am J Phys Anthropol 122:2327, 2003.
2003 Wiley-Liss, Inc.
subjects were from various cities and socioeconomic
backgrounds, living in the city of Ankara. A sample
of 121 individuals was randomly selected from this
larger group, and was identied as the study group.
The other 110 subjects formed the control group.
When stature was analyzed in each group, both
showed a Gaussian distribution (Figs. 1, 2). The
respective coefcients of skewness and kurtosis
were 0.084 and 0.029 for the study group, and
0.085 and 0.327 for the control group. The Kol-
mogorov-Smirnov test of normality also veried
these ndings (P 0.001 for both groups).
Each subjects body height and tibia length were
measured using a Martin anthropometer. Body
height was measured with the subject standing in
bare feet with his back to anthropometer. The per-
sons head was adjusted to the Frankfurt horizontal,
and then the head was tilted slightly upwards by
applying gentle force to the mastoid processes and
zygomatic bones (Cameron et al., 1981). For tibia
length, the distance between the medial condyle and
the most distal point of the medial malleous was
*Correspondence to: Izzet Duyar, Department of Physical Anthro-
pology, Ankara University, 06100 Sihhiye, Ankara, Turkey.
E-mail: izduyar@yahoo.com
Received 14 December 1999; accepted 18 December 2002.
DOI 10.1002/ajpa.10257
24 I. DUYAR AND C. PELIN
Fig. 1. Distribution of individuals in study group.
Fig. 2. Distribution of individuals in control group.
measured (Martin et al., 1988). The authors took all
measurements and recorded to the nearest millime-
ter. The mean age, stature, and tibia length in the
study and control groups are listed in Table 1. As
Table 1 shows, there were no signicant differences
between the two groups.
The study group was divided into three subgroups
according to body height, with the 15th and 85th
percentiles used as cutoff levels. A height of 1,652
mm or less was identied as short, 1,6531,840
mm as medium, and 1,841 mm or more as tall.
Different linear regression formulae were generated
for each of these height categories. The general for-
mula and the group-specic formulae derived from
the study group were used in the control group in
order to test the accuracy of the formulae.
RESULTS
The mean stature and tibia length values for the
short, medium, and tall subjects in the study group
are presented in Table 2. Tibia length did not scale
isometrically with stature, and the tibia length/body
height ratio increased as stature increased (Figs. 3,
4). In other words, tall subjects had longer tibias
relative to their height than medium and short sub-
jects. The tibia length/body height ratios for the
three study-group stature categories were statisti-
cally different, according to one-way analysis of vari-
ance (F 22.98, P 0.0001).
These differences suggested the need for different
regression formulae to estimate stature. Table 3
lists the general formula, and the group-specic for-
mulae that were derived from the study group based
on tibia length in the short, medium, and tall cate-
gories. Table 3 also shows the standard errors of the
estimates, the R2 values, and the F values for the
general and group-specic formulae.
Table 4 shows the difference between stature pre-
dicted by the general formula (stature 678.68
2.738 * tibia length) and true stature for each height
category in the control group. In medium subjects,
the difference between estimated and true height
was relatively small, whereas in tall and short sub-
jects, the means for under- and overestimation were
approximately 35 mm. When the general formula
was used, the estimated stature of tall subjects was
less than the true height, and the opposite applied to
short subjects. In other words, estimations by the
general formula for subjects in the extreme height
categories were closer to average height than then
they actually were.
Table 5 shows the differences between the specic
formula-predicted and true stature in the three
height categories in the control group. The mean
values for the two estimating procedures (general
formula vs. specic formulae) indicated that the
group-specic formulae yielded more accurate re-
sults. The difference between the general formula
and group-specic formula ndings in short subjects
was approximately 25 mm. In tall subjects, this dif-
ference was approximately 30 mm. In both cases,
analysis by paired-sample t-test showed that the
general formula-specic formula differences were
signicant (t 9.18, P 0.001 for short subjects; t
4.88, P 0.001 for tall subjects). In medium sub-
jects, there was no statistical difference between the
results with the two methods (t 0.30, P 0.05).On
the other hand, from the viewpoint of general sam-
ple means, there was no signicant difference be-
tween the two procedures (t 0.33, P 0.05). In
summary, the ndings indicated that group-specic
formulae were more accurate for subjects in the
extreme height categories.
 STATURE ESTIMATION BASED ON TIBIA LENGTH 25
TABLE 1. General characteristics of study and control groups
Study group (n 121) Control group (n 110) t P
Variable Mean SD Mean SD
Age (years) 20.87 4.01 21.72 2.58 1.90 0.059
Stature (mm) 1,746.41 91.39 1,746.69 88.30 0.02 0.981
Tibia length (cm) 389.85 30.15 390.43 28.14 0.12 0.901

TABLE 2. Comparison of anthropometric measurements and indices in study group (n 121)

Short subjects (n 18) Medium (n 86) Tall subjects (n 17)
Variable Mean SD Mean SD Mean SD
Stature (mm) 1,608.90 41.58 1,746.47 53.35 1,891.71 44.46
Tibia length (mm) 347.61 14.66 398.67 19.17 436.18 18.74
(Tibia length * 100)/stature 21.60 0.68 22.31 0.75 23.06 0.76

Fig. 4. Scatterplot (with 95% condence interval) for tibia

Fig. 3. Scatterplot (with 95% condence interval) for tibia
length and stature in control group.
length and stature in study group.

DISCUSSION modied because of the secular changes in propor-

Populations show considerable differences in body
proportions. This has important implications for
stature estimations based on body remains. To ad-
dress these differences, specic formulae have been
generated for certain populations (Breitinger, 1937;
Trotter and Gleser, 1952, 1958; Olivier, 1969; For-
micola and Franceschi, 1996; de Mendonca, 2000).
However, body proportions vary not only with ethnic
origin, but also with time and geographical context.
In their comparison of anthropometric data collected
from Japanese children in 1957, 1967, and 1977,
Tanner et al. (1982) found that the lower limb
length/stature ratio increased over time. Other au-
thors reported similar ndings (Greulich, 1957;
Zhang and Huang, 1988; Duyar, 1995). These secu-
lar changes in body proportions have implications
for studies on stature estimation. Jantz (1992)
stated that the Trotter-Gleser formula should be
tional relationships. Another consideration is that
body proportions may change due to nutrition and
living conditions. Wales et al. (1992) observed that,
compared to normal children, abused children had
shorter legs in proportion to sitting height.
Regarding estimation of height from long bones,
femur and tibia lengths are more closely related to
stature than the lengths of other long bones. Essen-
tially, it is possible to explain variations in stature
based on differences in the dimensions of these long
bones (Trotter and Gleser, 1952, 1958; Formicola
and Franceschi, 1996); however, it is not clear
whether the femur or tibia has more impact on
changes in body proportions. Feldesman et al. (1990)
found the relationship between femur length and
stature to be virtually constant in adults, regardless
of gender or ethnicity.
Because the tibia and bula are known to be more
strongly positively allometric than the femur (Mead-
26 I. DUYAR AND C. PELIN
TABLE 3. General height estimation formula and group-specific formulae based on tibia length for different
stature categories in study group (mm)
Stature category Regression equation SE R2 F P
Short Stature 951.94 1.890 * tibia length 31.96 0.444 12.78 0.003
Medium Stature 944.82 2.057 * tibia length 36.14 0.546 101.21 0.000
Tall Stature 1,224.15 1.530 * tibia length 35.08 0.645 10.69 0.005
General formula Stature 678.68 2.738 * tibia length 39.35 0.816 528.39 0.000

TABLE 4. Differences between estimated height based on body proportions have also been studied from the
general formula and true height in control group (mm)1 perspective of growth and nutrition. Still, no de-
Stature tailed investigation has examined the effects of such
category n Mean2 SD Minimum Maximum proportional changes on stature estimation. Aldegh-
Short 15 35.14 34.16 10.76 103.03 eri and Agostini (1993) described the relationships
Medium 79 1.60 30.18 51.83 93.38 between lower limb anthropometric measurements
Tall 16 34.12 35.77 114.50 20.85 and stature, and emphasized the importance of
All controls 110 0.98 36.32 114.50 103.03
these with regard to surgical approaches. However,
1
Stature 678.68 2.738 * tibia length. they did not mention proportional changes in rela-
2
Negative values indicate underestimates; positive values indi- tion to stature differences among individuals. For-
cate overestimates.
micola and Franceschi (1996) discussed the poten-
tial for error in stature estimation, particularly in
TABLE 5. Differences between estimated height based on group-
specific formulae and true height in control group (mm)1
tall and short subjects, and stated that the major
axis regression technique is more reliable for indi-
Stature viduals in extreme height categories.
category n Mean2 SD Minimum Maximum
When we used a single regression equation, the
Short 15 10.80 31.65 33.45 82.88 predicted heights were tall for short individuals and
Medium 79 1.13 28.10 42.63 72.82
Tall 16 4.39 20.08 43.54 27.40
short for tall individuals. Formicola and Franceshi
All controls 110 1.64 27.68 43.54 82.88 (1996) reported the same result, and Trotter and
Gleser (1958) noted this as a typical phenomenon of
1
Short: Stature 951.94 1.890 * tibia length. Medium: Stat-
ure 944.82 2.057 * tibia length. Tall: Stature 1,224.15
regression equations. We used group-specic for-
1.530 * tibia length. mulae in addition to the general formula, and the
2
Negative values indicate underestimates; positive values indi- group-specic calculations yielded signicantly
cate overestimates. smaller differences between predicted and true
height. Our results indicate that group-specic for-
ows and Jantz, 1995), we examined the tibia length/ mulae give more accurate height estimates for indi-
stature ratios in short, medium, and tall subjects. viduals in extreme height categories. In summary, it
There were signicant differences among the ratios appears that group-specic formulae may be more
for these categories. Our ndings also conrm that reliable for forensic cases in which height estimates
this ratio increases as height increases. Similar for individuals are important.
changes in the ratios of the lengths of other long On the other hand, in some forensic cases, the
bones to stature have also been reported (Meadows available bones or body parts are isolated. In such
and Jantz, 1995; Konigsberg et al., 1998). These cases, the isolated bones or body parts should be
changes verify our hypothesis that group-specic evaluated according to their own dimensions in or-
formulae are needed for height estimation. der to decide which stature group-specic formula
An extensive study by Konigsberg et al. (1998) would be applied. High correlation between tibia
analyzed different methods for estimating and cali- length and stature is well-known (e.g., Trotter and
brating stature. The authors compared results with Gleser, 1958). Also in the present study, this corre-
inverse calibration, classical calibration, femur-stat- lation was found to be high (r 0.903, P 0.01).
ure ratio, and major and reduced major axis meth- Depending on this fact, tibia length is thought to be
ods. Based on their ndings, Konigsberg et al. (1998, an indicator for stature and stature groupings.
p. 88) argued that when an individual case derives In the present study, tibiae were classied accord-
from the same stature distribution as the reference ing to their lengths into three groups as short, me-
sample, inverse calibration is clearly the preferred dium, and long. As in the classication of body
method. However, when estimating height for indi- height, 15th and 85th percentiles were used as cut-
viduals who were at the extremes, they found that off points; tibiae with a length of 360 mm or less
the square of the mean error was higher with this were determined as short, 361 422 mm as medium,
method. Compared to these results, the procedure and 423 mm and above as long. In addition to stat-
we used in our study yielded a lower error when ure grouping, another grouping was established
estimating height in the extreme categories. based on tibia length. Short, medium, and tall
Many studies of the lower limb/stature ratio high- groups constructed by these two methods were com-
lighted interpopulational differences. Disparities in pared with each other, and no statistically signi-
 STATURE ESTIMATION BASED ON TIBIA LENGTH
cant difference was observed (Fshort 1.571, P
0.05; Fmedium 0.808, P 0.05; Ftall 0.484, P
0.05). In other words, long tibiae mostly belonged to
the tall, and short tibiae to short individuals. These
ndings indicate that grouping depending on tibia
length could be used to identify the stature group to
which a specimen belongs. In order to verify this
suggestion, the previously calculated stature group-
specic regression formulae were applied to the
groups established based on tibia length. The differ-
ences between the estimated and true heights were
8.2 mm, 4.2 mm, and 6.3 mm for short, medium,
and tall groups which were classied based on tibia
length, respectively. When these gures were com-
pared with those in Tables 4 and 5, it was clearly
seen that the results obtained by the application of
the stature group-specic formulae to the groups
based on body height and tibia length were closely
similar with each other. Overall, the procedure we
present in this paper is appropriate for use in pre-
dicting stature grouping.
LITERATURE CITED
Aldegheri R, Agostini S. 1993. A chart of anthropometric values.
J Bone Joint Surg [Br] 75B:86 88.
Breitinger E. 1937. Zur Berechnung der Korperhohe aus den
langen Gliedmassenknochen. Anthropol Anz 14:249 274.
Cameron N, Hiernaux J, Jarman S, Marshall WA, Tanner JM,
Whitehouse RH. 1981. Anthropometry. In: Weiner JS, Lourie
JA, editors. Practical human biology. London: Academic Press.
p 2552.
Cheng JCY, Leungh SSF, Lau J. 1996. Anthropometric measure-
ments and body proportions among Chinese children. Clin Or-
thop 323:2230.
de Mendonca MC. 2000. Estimation of height from the length of
long bones in a Portuguese adult population. Am J Phys An-
thropol 112:39 48.
Duyar I. 1995. Secular trend in height and sitting height of
Turkish schoolchildren in Ankara between 1950 and 1986.
Hacettepe Univ J Faculty Lett 12:113 [in Turkish].
27
Eveleth PB, Tanner JM. 1990. Worldwide variation in human
growth. 2nd ed. Cambridge: Cambridge University Press.
Feldesman MR, Kleckner JG, Lundy JK. 1990. The femur/stature
ratio and estimates of stature in mid- and late-Pleistocene
fossil hominids. Am J Phys Anthropol 83:359 372.
Formicola V, Franceschi M. 1996. Regression equations for esti-
mating stature from long bones of early Holocene European
samples. Am J Phys Anthropol 100:83 88.
Greulich WW. 1957. A comparison of physical growth and devel-
opment of American born and native Japanese children. Am J
Phys Anthropol 15:489 515.
Jantz RL. 1992. Modication of the Trotter and Gleser female
stature estimation formula. J Forensic Sci 37:1230 1235.
Konigsberg LW, Hens SM, Jantz LM, Jungers WL. 1998. Stature
estimation and calibration: Bayesian and maximum likelihood
perspectives in physical anthropology. Yrbk Phys Anthropol
41:6592.
Martin AD, Carter JEL, Hendy KC, Malina RM. 1988. Segment
lengths. In: Lohman TG, Roche AF, Martorell R, editors. An-
thropometric standardization reference manual. Champaign,
IL: Human Kinetics. p 9 26.
Meadows L, Jantz RL. 1995. Allometric secular change in the
long bones from the 1880s to the present. J Forensic Sci 40:
762767.
Meadows Jantz L, Jantz RL. 1999. Secular change in long bone
length and proportion in the United States, 1800 1970. Am J
Phys Anthropol 110:57 67.
Olivier G. 1969. Practical anthropology. Springeld, IL: Charles
C. Thomas.
Sjvold T. 1990. Estimation of stature from long bones utilizing
the line of organic correlation. Hum Evol 5:431 447.
Tanner JM, Hayashi T, Preece MA, Cameron N. 1982. Increase in
length of leg relative to trunk in Japanese children and adults
from 1957 to 1977: comparison with British and with Japanese
Americans. Ann Hum Biol 9:411 423.
Trotter M, Gleser G. 1952. Estimation of stature from long bones
of American whites and Negroes. Am J Phys Anthropol 10:463
514.
Trotter M, Gleser G. 1958. A re-evaluation of estimation of stat-
ure based on measurements of stature taken during life and
long bones after death. Am J Phys Anthropol 16:79 123.
Zhang X, Huang Z. 1988. The second national growth and devel-
opment survey of children in China, 1985: children 0 to 7 years.
Ann Hum Biol 15:289 306.