Professional Documents
Culture Documents
31N, 28
to 25
C by mid-
day; and, summer temperatures average 19
C in the
evening and between 40
and 50
C by midday (Giddy,
1987; Sutton, 1950). Mean annual rainfall (0.3 mm/year)
and humidity are moderately low in the Oasis (Dupras
and Schwarcz, 2001; Sutton, 1947). These climatic condi-
tions are very similar to those during the Romano-
Christian period from which the Kellis 2 cemetery sample
derives, although precipitation in the Oasis was greater
during the Romano-Christian period than today (Giddy,
1987; Stewart et al., 2003).
The arid conditions, low acidity of the soil in the West-
ern Desert, and the consistent mortuary practices
observed within the Kellis 2 cemetery have created ideal
conditions for the exceptional preservation of skeletal
remains, including the preservation of fetal and perinate
individuals. Archaeological evidence from the associated
ancient village of Kellis suggests occupation from AD 50
360 (Hope, 2003); however, numerous radiocarbon dates
from the Kellis 2 cemetery indicate use between AD 100
and 450 (Stewart et al., 2003). At present, 770 individuals
of an estimated 30004000 burials have been excavated,
and of those 725 have been analyzed. Approximately 35%
of the recovered individuals are adults, while the remain-
ing 65% are juveniles (Wheeler, 2009). The mortality pro-
le at Kellis 2 is similar to that expected in a natural
mortality distribution in preindustrial populations
(Tocheri et al., 2005). This study utilizes a cross-sectional
sample of 301 individuals spanning the fetal/perinate
through adult years (Table 1).
Skeletal measurements
Standard skeletal metrics were collected from all indi-
viduals as part of the basic osteological data collection
procedures of the Dakhleh Oasis Project Bioarchaeology
Research Team. A subset of skeletal measurements used
to calculate intralimb proportions was selected for this
study. Data from left-sided elements were examined; how-
ever, right-sided elements were included when the left
side was missing or damaged. Unilateral analyses are
deemed appropriate in this study since ecogeographic pat-
terns in intralimb proportions are likely to affect the body
symmetrically. However, it should be noted that bilateral
asymmetry in limb long bone lengths exists in utero, but
differences between left-sided and right-sided elements
are not statistically signicant (Bagnall et al., 1982).
222 M.M. BLEUZE ET AL.
American Journal of Human Biology
Maximum lengths for the major long bones in the upper
and lower limbs have also been found to show low levels of
bilateral asymmetry in a large, geographically and tempo-
rally diverse sample of modern humans (Auerbach and
Ruff, 2006). Maximum diaphyseal lengths for the
humerus (HL), radius (RL), femur (FL), and tibia (TL)
were measured in individuals with unfused proximal and
distal epiphyses to the nearest tenth of a millimeter using
digital sliding calipers, or to the nearest millimeter using
a standard osteometric board. Maximum bone lengths for
the same four long bones were measured in individuals
with fused proximal and distal epiphyses to the nearest
millimeter using a standard osteometric board following
standard long bone measurement protocols (Buikstra and
Ubelaker, 1994). Maximum bone lengths were converted
to maximum diaphyseal lengths to maintain consistency
in comparing juvenile and adult data (Ruff, 2007). Brach-
ial and crural indices were derived from the maximum
diaphyseal lengths following standard formulae: brachial
index 5maximum radius length/maximum humerus
length 3 100; and, crural index 5maximum tibia length/
maximum femur length 3 100.
Sex and age estimation
Reliable estimations of sex in juvenile skeletal remains
are problematic; therefore, sex estimation was not
attempted for the juvenile sample (Scheuer and Black,
2000a,b). This may limit interpretations of the data since
there are known sex differences in limb long bone growth
velocities and timings, which may ultimately contribute
to sex differences in adult intralimb proportions
(Buschang, 1982; Holliday, 1999; Smith and Buschang,
2004, 2005).
Fig. 1. Map of Dakhleh Oasis and the Kellis 2 site, Egypt.
TABLE 1. Samples
Age cohort Age range N brachial N crural
Fetal/Perinate 941 weeks
a
24 23
Infant 0.00.9 year 62 69
Child 1 (C1) 1.04.9 years 30 34
Child 2 (C2) 5.010.9 years 15 16
Child 3 (C3)
b
11.014.9 years 4 3
Adult female 15.01 years 89 87
Adult male 15.01 years 64 69
Total 288 301
a
While the fetal cohort includes individuals beginning at nine weeks gestation
(Scheuer and Black, 2000a), the youngest individuals in this study are estimated
at 24 weeks gestation.
b
Because of the small sample size, the C3 cohort was combined with the C2
cohort. Mean intralimb proportions are not signicantly different between the
groups, and C3 individuals in this study are all aged near the lower end of the
C3 age range (i.e., 11.011.4 years).
ONTOGENETIC CHANGES IN INTRALIMB PROPORTIONS 223
American Journal of Human Biology
The dentition provides the most reliable estimation of
chronological age in juveniles; however, inter- and intra-
population variation in dental formation and eruption
times exist and different dental aging methods have vari-
ous degrees of accuracy depending on the sample popula-
tion (Foti et al., 2003; Liversidge, 1994; Liversidge and
Molleson, 2004; Saunders, 2000; Smith, 1991). Juveniles
in this study were aged using dental calcication tables
(Moorrees et al., 1963a,b; Smith, 1991). Age estimates
were based on macroscopic examinations of the dentition
since radiographic equipment was not available in the
eld. Multiple teeth from each individual were examined
and suggested age and age ranges were based on crown or
root formation. The assigned dental age was recorded as
the mean of all available ages suggested by each tooth.
When dentition was absent, juvenile age estimates were
based on epiphyseal development and fusion, pars basila-
ris morphology, and fusion of cranial elements (Krogman
and Iscan, 1986; Scheuer and Black, 2000a). Fetal and
perinate age was estimated using a combination of basioc-
ciput metrics, development and fusion of cranial ele-
ments, and long bone diaphyseal lengths (Scheuer et al.,
1980; Scheuer and MacLaughlin-Black, 1994; Sherwood
et al., 2000).
Age terminology used to describe juveniles is inconsis-
tent in the literature and often varies depending on the
objectives of the research (e.g., biological, social) (Hal-
crow and Tayles, 2008). In this study, we use the clinical
pediatric/developmental osteology age terminology to
provide a biological basis for comparisons. Fetus
includes individuals aged nine weeks gestation to birth,
perinate is from 24 weeks gestation to seven postnatal
days (or around the time of birth), infant is from birth
to the end of the rst postnatal year (i.e., 0.00.9 year),
and child is from 1.0 to 14.9 years of age (Halcrow and
Tayles, 2008; Scheuer and Black, 2000a,b). The fetus
and perinate cohorts were pooled since age ranges over-
lap in these groups. Since signicant growth and devel-
opmental changes occur during childhood, we further
divided this cohort to examine shifts in intralimb propor-
tions at a ner scale. Child 1 (C1) encompasses the age
range between 1.0 and 4.9 years of age (early childhood),
child 2 (C2) encompasses the age range between 5.0
and 10.9 years of age (middle childhood), and child 3
(C3) encompasses the age range between 11.0 and 14.9
years of age (late childhood). These subdivisions match
important somatic growth spurts and physiological develop-
ments that occur during childhood (Bogin, 1999; Lewis,
2007), and therefore serve as the rationale for further sub-
dividing the childhood cohort. Because of small sample size,
however, individuals in the C3 cohort were combined with
the C2 cohort. Mean intralimb indices are not signicantly
different between the C2 and C3 groups, and the C3 indi-
viduals in this study were all aged closer to the lower end of
the C3 age range (i.e., 11.011.4 years of age).
Sex estimation for adult skeletal remains was based on
standard osteological techniques using morphological fea-
tures in the skull and pelvis (Buikstra and Ubelaker,
1994). Pubic symphysis morphology based on the Suchey-
Brooks method and sternal rib standards were used to esti-
mate age in adults (Brooks and Suchey, 1990; Iscan and
Loth, 1986). In this study, adults are identied based on
full epiphyseal fusion in the humerus and radius, or full
epiphyseal fusion in the femur and tibia.
Statistical analyses
Brachial and crural indices were compared across age
cohorts with Welchs ANOVA tests since the data is nor-
mally distributed, but the assumption of homogeneity of
the variances is violated (McDonald, 2008). Dunnett-
Tukey-Kramer (DTK) pairwise multiple comparison tests
with family-wise error adjusted for unequal variances and
unequal sample sizes were conducted post-hoc to deter-
mine which groups signicantly differed (Lau, 2009).
Adult sexes were analyzed separately since intralimb pro-
portions have been shown to be a sexually dimorphic trait
(Holliday, 1999). Intralimb indices are graphically repre-
sented in box-and-whisker plots (Figs. 2 and 3). The box
represents the interquartile range and the upper and
lower whiskers represent the maximum and minimum
values, respectively. The median is denoted as the line
within the box. Spearmans rank correlation coefcients
(q) with Fishers z-transformations were calculated
between intralimb skeletal elements (i.e., RL vs. HL and
TL vs. FL) in each age cohort to explore developmental
conservation and evolvability in the brachial index and
crural index. Fishers z-transformations were used to cal-
culate 95% condence intervals for the correlation coef-
cients, and dependent t-tests were used to compare
Fishers z-transformations of correlation coefcients in
the upper and lower limbs within each cohort (Sheskin,
2004). A greater correlation between intralimb elements
Fig. 2. Box-plot of brachial index. Means and standard deviations
as follows: fetus/perinate 582.4462.73, infant 579.05 62.48,
C1 576.1561.17, C2/C3577.2062.25, adult female 578.50 62.55,
and adult male 580.2862.53.
Fig. 3. Box-plot of crural index. Means and standard deviations as
follows: fetus/perinate 587.43 61.32, infant 584.87 62.28,
C1 582.5961.52, C2/C3 583.5661.63, adult female 582.94 62.07,
and adult male 583.4362.18.
224 M.M. BLEUZE ET AL.
American Journal of Human Biology
suggests reduced evolvability, while a lower correlation
suggests greater evolvability (Young et al., 2010). Statisti-
cal signicance is set at P<0.05 and all statistical analy-
ses were carried out using R-Studio.
RESULTS
The mean brachial index is signicantly different among
the age cohorts (F5 25.96, P<0.00). It is greatest during
early ontogeny, declines during early childhood, and
increases again during middle/late childhood (Fig. 2). The
fetus/perinate cohort has a signicantly greater brachial
index than all other groups, and the infant cohort has a sig-
nicantly greater brachial index than the C1 cohort. The
brachial index does not signicantly change during child-
hood. Children in the C1 cohort have a signicantly lower
brachial index than female and male adults, while children
in the C2 cohort have a signicantly lower brachial index
than male adults only (Table 2). The adult pattern is rst evi-
dent during infancy, but is not maintained during childhood.
The mean crural index is signicantly different among
the age cohorts (F537.97, P<0.00). It is greatest in the
fetus/perinate cohort, decreases during infancy and early
childhood, and increases, albeit nonsignicantly, during
middle/late childhood (Fig. 3). The fetus/perinate cohort
has a signicantly greater crural index than all other
groups. Infants have a signicantly greater crural index
than the C1 and adult cohorts. The crural index does not
signicantly change during childhood (Table 2). The adult
pattern is rst evident during early childhood and is
maintained from this point forward.
A signicantly lower correlation is observed between RL
and HL as compared to TL and FL in the fetus/perinate
cohort, and a signicantly greater correlation is observed
between RL and HL as compared to TL and FL in the C1
cohort. The correlations between RL and HL as compared
to TL and FL are not signicantly different in the other age
cohorts (Table 3). We pooled our postnatal juvenile sample
(ages 0.011.4 years) in order to compare the Kellis 2 juve-
niles with published data from Jomon period juveniles
(Temple et al., 2011). The correlations between RL and HL
as compared to TL and FL are not signicantly different in
the Kellis 2 sample (z 50.23, P50.81), which contrasts
with results from Temple et al. (2011).
DISCUSSION
The primary objective of this study was to describe
changes in intralimb proportions during development in
the Kellis 2 sample from the Dakhleh Oasis, Egypt. The
adult pattern in the brachial index is rst evident in
infancy, but is not maintained throughout development.
Conversely, the adult pattern in the crural index appears
during early childhood and is maintained throughout
development. Thus, while adult intralimb proportions are
not present in utero they do appear very early in ontogeny.
This lends further support to the long held view that
intralimb proportions are genetically conserved traits
(Holliday, 1999; Schultz, 1923; Temple et al., 2011).
It has been shown that brachial and crural indices dur-
ing ontogeny are strongly correlated with latitude despite
shifting values over the course of development such that
relative ecogeographic patterns among geographically
diverse populations are maintained during development
(Cowgill et al., 2012; Ruff, 2007). In other words, there are
signicant differences in absolute mean brachial and cru-
ral indices among age-matched cohorts across geographi-
cally diverse populations despite similarities in shifting
patterns throughout development. Shifting patterns in
intralimb proportions during ontogeny are often linked to
differences in growth velocities of intralimb proximal and
distal elements (Cowgill et al., 2012; Temple et al., 2011).
In the Kellis 2 sample, the intralimb proportions are
greatest in the fetus/perinate and infant cohorts, decrease
during early childhood, and increase again during middle/
late childhood. Our results are consistent with previous
work and extend the pattern to the fetus/perinate period.
Since hyperthermia in utero may impede fetal growth and
development, high intralimb proportions may be under
strong selective pressures even before postnatal life (Ger-
icke et al., 1989; Xu et al., 2012). During the rst year of
life, intralimb proportions are signicantly lower than in
utero, but signicantly greater than in early childhood.
These results suggest that accelerated growth velocities of
distal elements relative to proximal elements do not carry
over into infancy.
We hypothesized that adult patterns in intralimb indi-
ces would appear early in ontogeny and be maintained
throughout development. This hypothesis is partially sup-
ported by the results. While mean brachial indices are not
signicantly different between infants and adults, chil-
dren have signicantly lower brachial indices than adults
(barring C2 children and adult females). This may reect
the high growth velocity of the humerus relative to the
radius during childhood and adolescence (Smith and
TABLE 2. P-values from Dunnett-Tukey-Kramer pairwise multiple
comparisons post hoc tests
Brachial index
Infant C1 C2 Adult female Adult male
Fetus/perinate 0.00 0.00 0.00 0.00 0.00
Infant 0.00 0.05 0.76 0.13
C1 0.69 0.00 0.00
C2 0.29 0.00
Adult female 0.00
Crural index
Infant C1 C2 Adult female Adult male
Fetus/perinate 0.00 0.00 0.00 0.00 0.00
Infant 0.00 0.93 0.00 0.00
C1 0.06 0.97 0.44
C2 0.11 0.59
Adult female 0.72
TABLE 3. Spearmans rank correlation coefcients
Cohort Comparison q Fisher-z 95% CI z-value P
Fetal/Perinate RL vs. HL 0.909 1.527 0.8640.941 2.09 0.04
TL vs. FL 0.972 2.092 0.9540.980
Infant RL vs. HL 0.973 2.092 0.9510.982 0.71 0.48
TL vs. FL 0.979 2.298 0.9680.988
C1 RL vs. HL 0.989 2.647 0.9780.995 2.57 0.01
TL vs. FL 0.958 1.946 0.9200.980
C2 RL vs. HL 0.956 1.946 0.8970.985 0.50 0.62
TL vs. FL 0.969 2.993 0.9220.989
Adult ($) RL vs. HL 0.822 1.157 0.7380.878 0.10 0.92
TL vs. FL 0.827 1.188 0.7510.886
Adult (#) RL vs. HL 0.794 1.071 0.6750.867 0.03 0.98
TL vs. FL 0.796 1.099 0.6950.872
ONTOGENETIC CHANGES IN INTRALIMB PROPORTIONS 225
American Journal of Human Biology
Buschang, 2004, 2005). It has been demonstrated that the
brachial index is more strongly correlated with mean
annual temperature than the crural index in adults (Trin-
kaus, 1981), and that the brachial index has a stronger
correlation with latitude in adults than in juveniles (Cow-
gill et al., 2012). Since adult patterns in the brachial index
are not maintained during childhood, the results suggest
that selective pressures from climate acting on the brach-
ial index may be weaker in children than in infants and
adults. The high surface area to body mass ratio in chil-
dren, which enables them to lose dry heat, may ease selec-
tive pressures on the brachial index (Falk, 1998;
Rowland, 2008). Adult patterns in the crural index appear
in early childhood and are maintained throughout middle
and late childhood. It is possible that this reects func-
tional constraints placed on the lower limbs (Temple
et al., 2011; Young et al., 2010).
Correlation analyses indicate a signicantly lower cor-
relation between RL and HL (i.e., greater evolvability)
than between TL and FL in the fetus/perinate cohort,
which suggests that the crural index is more genetically
conserved than the brachial index in utero. These results
are in accordance with previous studies suggesting
greater genetic conservation of the crural index as com-
pared to the brachial index in postnatal life (Frelat and
Mittereocker, 2011; Temple et al., 2011). The greater cor-
relation between lower limb elements as compared to
upper limb elements is not maintained throughout devel-
opment. During early childhood, RL and HL have a signif-
icantly greater correlation than TL and FL. Growth rates
of proximal and distal segments are more variable in the
lower limbs as compared to the upper limbs during child-
hood (ages 310 years), which may explain the signi-
cantly greater correlation between RL and HL as
compared to TL and FL in the C1 cohort (Smith and
Buschang, 2004). The high degree of variation in growth
of the tibia is particularly evident during early childhood,
and may further contribute to the signicantly greater
correlation between RL and HL as compared to TL and
FL (Smith and Buschang, 2004). Animal studies have
shown that exercise increases limb length by increasing
the amount of nutritional solute delivery to growth plates
in weight-bearing limb bones (Jurvelin et al., 1988; Serrat
et al., 2007, 2010). Since children in the early childhood
phase are beginning to walk, the lower correlation of TL
and FL relative to RL and HL may reect growth devia-
tions from the biomechanical demands of bipedalism.
Greater variation in growth rates of lower limb elements
as compared to upper limb elements during early child-
hood coupled with lower limb skeletal element elongation
from increased loading due to the start of walking behav-
iors may explain the greater correlation between RL and
HL as compared to TL and FL in the C1 cohort.
It was previously shown that RL and HL have a signi-
cantly lower correlation than TL and FL in a Jomon
period juvenile sample (Temple et al., 2011). Our pooled
Kellis 2 postnatal juvenile sample does not follow this pat-
tern since the correlations between RL and HL as com-
pared to TL and FL are not signicantly different. These
contrasting results may reect differences in sample con-
struction since the Jomon juvenile sample largely con-
sisted of individuals between 2.1 and 10.9 years of age,
while infants (ages 0.00.9) slightly outnumbered chil-
dren (1.011.4 years of age) in the pooled postnatal Kellis
2 juvenile sample.
Given the climatic extremes experienced in the Dakh-
leh Oasis it is not unexpected to nd elevated intralimb
indices in the Kellis 2 juvenile sample. The Kellis 2 sam-
ple is similar to other warm-adapted populations in this
regard. While cross-sample comparisons are approached
with caution because of differences in age cohort construc-
tion, Kellis 2 juveniles have similar mean intralimb pro-
portions as those reported for similarly age-matched
juveniles from Kulubnarti, Upper Nubia (21