Dennis P. Van Gerven George J. Armelagos ... - Anthropology

Dennis P. Van Gerven
Department of Anthropology,
University of Colorado, Boulder,
Colorado 80309, U.S.A.
George J. Armelagos
Department of Anthropolo~, University
of Massachusetts, Amherst,
Massachusetts, 01002, U.S.A.
Keywords: paleo&mography, lijb,
tables, skeletal age.
Received 12 October 1982 and
accepted 12 October 1982
"Farewell to Paleodemography?" Rumors of
Its Death Have Been Greatly Exaggerated
Bocquet-Appel & Masset (1982) have recently proposed that
paleodemography be abandoned in the light of two seemingly
insurmountable deficiencies. First, they contend that the age structures of
skeletal samples reflects nothing more than the age structures of their
reference populations, that is to say, the samples by whose criteria they were
aged. Second, they contend that age estimates made on adult skeletons lack
sufficient accuracy to permit legitimate demographic analysis. Such
estimates are, in the authors' view, nothing more than "random
fluctuations and errors of method".
The purpose of the present research is lo considcr Bocquet-Appel &
Masset's assertions in light of the empirical evidence produced by
paleodemography. Our results fail to support either of the authors'
allegations. In fact, skeletal samples do not invariably reflect the structure
of their reference populations and age estimates do not produce the random
fluctuations predicted by the authors' a priori criteria for age estimation.
Most importantly, our data indicates that the authors fail to consider the
methods and accomplishments of paleodemography in their entirety and
thereby present a biased and inadequate discussion of the salient issues.
1. Introduction
In a recent article entitled "Farewell to Paleodemography" Bocquet-Appel & Masset
(1982) review and critically assess what they t~el to be the principal methods and
accomplishments of paleodemography. They are particularly concerned with the accuracy
by which individual skeletons can be aged, and with the degree to which skeletal ages can
be combined into meaningful demographic profiles. Their assessment is far from
encouraging. Beginning first with the creation of demographic profiles, the authors
compare several skeletal populations with their reference populations, that is, the
populations by which the skeletons were aged. For example, the age distribution of
skeletons from ancient Nubia (Swedlund & Armelagos, 1969) is compared to the age
distribution of the McKern-Stewart sample of Korean War dead by whose pubic age
criteria the Nubians were presumably aged. On the basis of such comparisons the authors
conclude that skeletal populations arc." doomed to emerge as demographic copies of their
reference populations.
The authors then proceed to the question of aging the individual skeleton. After
considering various methods separately and combined, they estimate the highest
correlation between chronological and skeletal age to be approximately 0"8. Based on a
number of a priori criteria and assumptions, this level of accuracy isj udged to be completely
inadequate. The authors suggest, for example, that in order to insure a 95% level of
accuracy in age classification for a population whose actual ages range over a 72-year span,
only 1"4 age categories can be meaningfully created.
When then is the fate ofpaleodemography and those who continue to estimate ages and
create age categories? The authors leave little doubt.
The scholars who persist in this course will only obtain artifacts; the information conveyed by
the age indicators is so poor that the age distributions thus available can hardly reflect anything
but random fluctuations and errors o/method [emphasis ours] (Bocquet-Appel & Masset, 1982;
p. 329).
Journal ojHuman Evolulion (1983) 12, 353 360
0047-2484/83/040353 + 08 $03.00/0 9 1983 Academic Press Inc. (London) Limited

354 D. P. VAN GERVEN AND G. j. ARMELAGOS
The authors' farewell to paleodemography is clearly not a personal adieu but an obituary
for the inquiry as a whole. Before the death becomes official, however, som e reappraisal of
the author's diagnosis seems warranted. We intend to consider carefully whether the
authors, themselves, have adequately and accurately portrayed the methods, objectives
and accomplishments of paleodemography. It is our position that the methods employed
by paleodemographers are far from inadequate. It is also our position that the results of
paleodemography lend absolutely no support to the contention that they represent nothing
more than "random fluctuations and errors of method". The authors' harsh appraisal must
stand not only in light of their personal criteria but the criteria reasonably, conscientiously
and successfully applied by others. We will then consider whether the authors have
adequately considered and assessed the wider-ranging objectives and accomplishments of
paleodemography.
2. The Relationship Between Skeletal Samples and
Their Reference Populations
In order to determine whether skeletal samples are, as suggested by Bocquet-Appel &
Masset, passive reflections of their reference populations, two skeletal samples from
Sudanese Nubia were selected for analysis. The first represcnts 201 individuals excavated
near Wadi Halfa and reported by Swedlund & Armelagos (1969). Although the reference
population associated with this sample by Bocquet-Appel & Masset was the
McKern-Stewart Korean War dead, the system most extensively used was that developed
by Todd (1920) and later modified by Brooks (1955). The second sample consists of 162
individuals excavated at the site of Kulubnarti (Van Gerven et al., 1981). This sample was
also aged using changes in the os pubis as descibed by Todd and later modified by Brooks.
Age categories for the two samples were arranged in five-year intervals from 15 through
55+. This was also done for Todd's original sample (Todd, 1920). In order to facilitate
statistical comparison, the data was then cast into cumulative mortality curves (Figure 1).
Statistical comparisons between each skeletal sample and the Todd reference sample were
made using the Kolmogorov-Smirnov two-sample test for cumulative frequencies as
discussed by Siegel (1958) and Lovejoy (1971).
As illustrated in Figure 1, differences between the Nubian samples and the Todd
Figure 1. A comparison of cumula-
tive mortality between two ancient
Nubian populations and their refer-
ence population based on Todd's
(1920) analysis of 306 American
males. [Kulubnarti (---), Wadi
HalIR ( ), Todd ( .... ).]
g
d
I00
9O
8O
7O
6O
/ .-- o z
20 /-~ //"
,oL..~176
15 20 25 50 55
Age
o/~ 5~
o~ //~ /
/o /
// !
/ .o
o /
.I-"
/
I I
40 45 50 55*

"FAREWELL TO PALEODEMOGRAPHY~: A REPLY 355
reference sample are extensive. With the exception of the first age category, all differences
are significant at the 99% confidence level. Furthermore, the two skeletal samples are
strikingly different from one another. These results in no way support the contention that
skeletal samples correspond primarily to the age structure of their reference sample. The
results do, on the other hand, correspond to previously observed differences among skeletal
populations aged according to a common reference group. Green and co-workers (1974)
for example, found important differences between skeletal samples from the site of
Meinarti in Sudanese Nubia corresponding to periods of village growth and decline, they
also found significant differences in life expectancy related to social status within the
Meinarti population. It must be emphasized that all the Meinarti skeketal remains were
aged in the same manner. Such differences, while making excellent biocultural sense, are
difficuIt to explain if skeletal remains reflect only the demographic structure of their
reference sample or random fluctuations and errors of method.
3. The Determination of Age at Death From Skeletal
Remains
The second critical issue in Bocquet-Appel & Masset's "Farewell to Paleodemograpy"
concerns the ability of skeletal biologists to assign accurately ages at death to skeletons.
While they accept the utility of age-dependent criteria applied to subadults, the authors are
far more skeptical of age-related changes in the adult skeleton. It is suggested that unless
the correlation between chronological age and skeletal age reaches 0'9 or better, the
prospects for paleodemography are dim. It is our contention that this a priori assessment is
of little practical value in determining the actual utility of skeletal age data.
In order to assess the impact of errors in age estimation, the sample of 306 males
analysed by Todd (1920) for age changes in the os pubis were cast into a
cumulative mortality curve based on their chronological ages. The sample was then cast
into the same age categories based on pubic ages determined according to Todd without
knowledge of chronological age. Pubic age values were derived from Todd's ten stage
system by averaging the oldest and youngest age associated with each individual's stage
assignment. All individuals whose oldest stage value was 10 (50+ years) were assigned to
the 50-55+ age category. While this procedure may have resulted in our overaging some
individuals classified by Todd as stage 9-10, it allowed us to avoid averaging his 50+ value
as a fixed integer.
As indicated in Figure 2, the cumulative frequency distributions created by
chronological and pubic age data are strikingly similar and are in fact tied through the first
two age groups. A maximum difference of 9% occurs at age 40 and is statistically
significant. Difl~rences between all other age groups are not significant at the 95%
confidence level. It is important to emphasize that the Todd system is seldom used in
isolation. As Bocquet-Appel & Masset themselves demonstrate, the ability to assess age at
death improves when multiple criteria are applied. Consequently, it is doubtful that
skeletal age indicators produce nothing of interest for demographic purposes. Put another
way, if the pubic ages derived from the Todd sample were our only basis for reconstructing
their mortality profile, would it be reasonable to conclude that we had reconstructed
nothing but "random fluctuations and errors of method"?
As a final point, it must be recognized that the age distribution cannot be assessed
beyond the 50-55+ category. While only a small percentage (usually between 1 and 10%)

35G D.P. VAN GERVEN AND G.J. ARMELAGOS
of skeletal remains fall in this interval, Bocquet-Appel & Masset contend that this
constitutes a major source of demographic error. They, however, offer little in the way of
concrete support for their contention that a large portion of adults in prehistoric societies
lived well into their 60s, 70s and even 80s. They site mortality among the Dobe !Kung as
exemplifying what they believe to be a typical mortality pattern with mode adult mortality
Figure 2. A comparison of cumula-
tive survivorship based on chrono-
logical age (---) and pubic age
( .... ) estimates for a sample o["
306 American males studied by
Todd (1920).
tO0
90
80
70
60
50
40
50
- 20-
..~./~
'~ F
15
o /
20 25 30 35
Age
.,y
o/
...4,. o
o /
/,
I k ~ I
40 45 50 55+
after 60 and even 70 years, however, this data constitutes little in the way of evidence either
for prehistoric mortality or the Dobe !Kung themselves. Howell's (1979) analysis of Dobe
!Kung mortality is based on developmental estimates of subadult ages which are then
extrapolated to adult mortality estimates using model lifetables for modern populations!
How this data can be cited as superior to paleodemographic reconstructions is not made
clear.
4. The Wider-Ranging Objectives and
Accomplishments of Paleodemography
If, as Bocquet-Appel & Masset suggest, paleodemographic reconstructions provide
nothing more than random fluctuations and errors of method, the consequences of such
errors and fluctuations should be apparent in all aspects of paleodemographic
analysis--not simply in the construction of mortality profiles. It is important to emphasize
that physical anthropologists have not limited their use of demographic data to questions of
mortality and survivorship. Researches into paleopathology as well as growth and
development have made abundant use of mortality data. The accomplishments of this
large body of research must also be given consideration before paleodemography is laid to
rest.
Recent research into the age-related loss of skeletal tissue known as osteoporosis (Dewey
el al., 1969; Van Gerven et aL, 1969) clearly illustrates the value of age data derived from
skeletal remains. In order to determine whether ancient human populations experienced
osteoporotic bone changes, Dewey and co-workers measured cortical bone thickness at the
proximal one-third of femurs from 203 Nubians excavated near Wadi Halfa. The
individuals are from the sample a portion of which was discussed earlier and illustrated in
Figure 1. As previously indicated, all ages at death were determined in the field using

"FAREWELL TO PALEODEMOGRAPHY': A REPLY 357
changes in the os pubis. Observations of cortical thickness were not a factor in age
determination. Measurements of cortical thickness were made on isolated femur sections
prepared in the physical anthropology laboratories at the University of Utah without
knowledge of the age or sex of any specimen. After cortical thickness values had been
obtained for the sample, the age data was added and mean cortical thickness was
determined for four age groups as illustrated in Figure 3.
o
o
-I0
-us
0
-5
g -20
-25
Figure 3. Percentage reduction in average cortical thickness (osteoporosis)
for males and females in (a) a modern American and (b) Ancient Nubian
sample.
- o (a)
>o
O" \
\
- ~9
\
\
I I I o
21-30 51-40 41-50 51,
o
0
5
g~
g o
5
- o (b)
\o o-.d ~
\
-- \
\
_ x
\O,
x x
"O
I I
16-2122-31 32-41 42-50+
Age Age
The results were then compared to the pattern of bone loss previously documented for a
modern American sample (Bartley & Arnold, 1965) of known age using the same
technique of cortical bone measurement. While the age groups used by Dewey do not
correspond precisely to the modern series, the similarity between the ancient Nubian and
modern American patterns of cortical bone loss is striking.* Both indicate a rapid rate of
cortical thinning with advancing age among women with a much lower rate among men.
Before we can conclude, as Bocquet-Appel & Masset do, that skeletal age data is
spurious, it must be explained how observations of bone loss, made independently of age,
could produce such a coherent pattern. Indeed, if Bocquet-Appel & Masset are correct,
such an analysis of osteoporosis should have produced no discernible change in cortical
thickness by age--that is, the relationship should have emerged as a random one.
The successful application of a demographic approach to phenomena other than
mortality has not been limited to osteoporosis. Lovejoy & Heiple (1981), for example,
found a correlation of 0'97 between long-bone fractures and years at risk (determined from
skeletal age) in a Late Woodland population from the Libben site in Northern Ohio.
Beyond a simple measure of correlation, however, their demographic approach provided
additional insight into the fracture process as it impacted different age groups. According
to Lovejoy & Heiple (1981):
Rates of fracture show marked elevation in two periods of the life cycle: adolescence/young
adulthood and old age. These are the two periods of the life cycle in which one would most likely
expect such high rates, if accidental trauma were the primary cause (p. 538).
*In order to tacilitatc comparison, values beyond age 50 were combined by
averaging in the modern sample.

358 D.P. VAN GERVEN AND G. J. ARMELAGOS
These results also correspond closely to the pattern determined from a demographic survey
of fractures in modern England and Wales (Buhr & Cooke, 1959) in which fractures
associated with high levels of activity accumulated during the young years while other
kinds of fractures accumulated during older age.
Further examples could be cited but the point would be the same. When age-related
events and processes affecting skeletal remains are examined independently of age
estimation, i.e. demographically, their patterns of occurrence make sense in light of
modern skeletal biology. In no case do the results reflect the random fluctuations and
errors of method predicted by Bocquet-Appcl & Masset. The problem, therefore, seems to
lie more with the authors' "farewell" and its underlying assumptions than with age
estimation and paleodemography.
Before concluding, one final issue must be addressed~ At the outset of their discussion,
Bocquet-Appel & Masset acknowledged the accuracy with which age can be determined
for subadult remains and accordingly exempt this segment of skeletal samples from their
assessment. However, inasmuch as subadult remains regularly constitute over 50% of
skeletal samples, and the authors conclude by dismissing all of paleodemography, some
consideration of subadult remains and their value to paleodemography seems warranted.
The Kulubnarti sample discussed earlier provides an excellent example of how valuable
subaduh materials can be to paleodemographic analysis. A comparison of probabilities of
dying and mean life expectancies between early and later Christian cemeteries at the site
has revealed a higher mortality rate among subadults (birth through 14 years) during early
Christian times (Van Gerven et al., 1981). In order to determine whether this difference is
due to differences in childhood stress and morbidity, probabilities of dying within and
between the two cemeteries were compared to frequencies of cribra orbitalia. Cribra
orbitalia is a lesion of the superior surface of the eye orbit observed in many Nubian
populations that has been related to dietary iron deficiency as well as parasitic and
bacterial infections (Carlson, et al., 1974; Van Gerven et al., 1981). In both cemeteries the
frequency of cribra orbitalia shows a high correspondence to probabilities of dying from
infancy through the early adult years. Of even greater significance, the higher probabilities
of dying among subaduhs in the early Christian cemetery correspond to higher frequencies
of the lesion.
These results support the contention that differences in childhood mortality from early
to late Christian times are primarily due to differences in biological stress acting on these
populations. It appears that the greater regional autonomy experienced by these people
during the later Christian period was a positive influence on their biological well being. In
terms of the issue at hand, it is particularly significant that subadult mortality is the most
sensitive barometer of that biocuhural change.
However, as with adult paleodemography, the analysis ofsubadults has not been limited
to questions of mortality. For example, the analysis of bone growth and development,
assessed independently of age determination, has been a major area of demographic
inquiry.
Studies of ancient Amerindian (johnston, 1962) as well as Nubian (Mahler, 1968;
Hummert & Van Gerven, 1982) children have demonstrated important correspondences
between prehistoric and modern growth patterns even though "... some degree of error of
error is introduced by the very fact that the sample is skeletal. It does not represent the
normal, healthy, population from which it was drawn" (Johnston, 1962, p. 249).
Mahler, for example, found that the velocity and symmetry of long-bone growth among

"FAREWELL TO PALEODEMOGRAPHY": A REPLY 359
ancient Nubians was broadly similar to that of modern Americans studied by Maresh
(1955) with the exception that Nubians had a later and stronger adolescent growth spurt.
Mahler hypothesized that the stronger growth spurt observed for the Nubians reflected
catch-up growth resulting from an inadequate childhood diet and reduced pre-adolescent
growth.
Recently, Huss-Ashmore (1978) and Martin & Armelagos (1979) have investigated the
nutritional basis fbr premature osteoporosis in juvenile and young adult Nubians and their
results lend strong support to Mahler's earlier hypothesis. When long-bone length was
plotted against cortical thickness and midshaft width for the Nubian sample, it became
apparent that Nubian juveniles maintained long-bone length at the expense of normal
cortical bone development. Following research carried out by Garnet al. (1966) on modern
children, Huss-Ashmore suggested that this decrease in cortical bone may be evidence for
protein-energy malnutrition in ancient Nubia. Here, then, we have a major new hypothesis
created from demographic research on subadult remains.
Studies such as this are not of simply passing importance, they represent a major arena
of paleodemographic inquiry. Data produced by such research has provided important
new insights into the dynamics of population adaptation as well as the biology of growth
and development. And yet, if Bocquet-Appel & Masset are to be taken seriously, such
research must either be dismissed or defined as non-demographic. In our view, neither
solution is appropriate. We would prefer to acknowledge the real and important limitations
within which all paleodemographers work, improve our methods when possible, and
acknowledge the past, present and future value of paleodemographic inquiry.
5. Conclusions
In conclusion, for all of their grim projections and dire forecasts, Bocquet-Appel & Masset
fail to account for one overriding fact; the results of paleodemographic research make
sense. If demographic reconstructions are nothing more than passive reflections of their
reference populations, important and bioculturally meaningful differences documented
within and between skeletal populations should not exist, but they do exist. If age estimates
combine to produce nothing more than random fluctuations and errors of method, such
random fluctuations should be apparent in the study of age-dependent processes, but no
such randomness is apparent. Most importantly, ifpaleodemography is to be dismissed, it
should be found wanting in its entirety, but it is not. The authors ignore more than they
consider. The study of subadult remains is entirely disregarded in their farewell as is the
modern fluorescence in paleopathology and growth and development. These areas owe
their current resurgence to a populational perspective provided by paleodemography.
What, then, is the status ofpaleodemography? The answer is clear. Rumors of its death
have been greatly exaggerated.
References
Bartley, M. H. & Arnold, J. S. (1965). Sex Differences in Human Skeletal Involution. Section III of Progress Report
AM 09379 to National Institute of Arthritis and Metabolic Disease, National Institute of Health. United States
Public Health Service.
Bocquet-Appel, J. & Masset, C. (1982). Farewell to paleodemography. Journal of Human EvOlution 11, 321-333.
Brooks, S. T. (1955). Skeletal age at death: the reliability of cranial and pubic age indicators. American Journal of
Physical Anthropology I3, 567 597.
Buhr, A.J. & Cooke, A. M. (1959). Fracture patterns. Lancet 1,531-536.

360 D. P. VAN GERVEN AND G. J. ARMELAGOS
Carlson, D. S., Armelagos, G. J. & Van Gerven, D. P. (1974). Factors contributing to the etiology of cribra
orbitalia in prehistoric Nubia. Journal of Human Evolution 3, 405-410.
Dewey, J. R., Armelagos, G. J. & Bartley, M. H. (1969). Femoral cortical involution in three Nubian
archeological populations. Human Biology 41, 13-28.
Garn, S. M., Rohmann, C. G. & Guzman, M. A. (1966). Malnutrition and skeletal development in the preschool
child. In Preschool Child Malnutrition, pp. 43-62. National Academy of Sciences-National Research Council,
Washington, D.C.
Green, S., Green, S. & Armelagos, G.J. (1974). Settlement and mortality of the Christian site (1050 A.D.-1300
A.D.) of Meinarti (Sudan). Journal of Human Evolution 3, 297-311.
Howell, N. (1979). Demography of the Dobe! Kung. New York, San Francisco & London: Academic Press.
Hummert, J. R. & Van Gerven, D. P. (1982). Long bone growth in a population from Kulubnarti, Sudanese
Nubia (55(>1450 A.D.). American Journal of Physical Anthropology 57, 198.
Huss-Ashmore, R. (1978). Nutritional determination in a Nubian skeletal population. American Journal of Physical
Anthroplogy 48, 407.
Johnston, F. E. (1962). Growth of the long bones of infants and young children at Indian Knoll. American Journal of
Physical Anthroplogy 20, 249-254.
Lovejoy, C. O. (1971). Methods tbr the detection of census error. American Anthropologist 73, 101-109.
Lovejoy, C. O. & Heiple, K. G. (1981). The analysis offi'actures in skeletal populations with an example from the
Libben site, Ottowa County Ohio. American Journal of Physical Anthroploogy 55, 529-541.
Mahler, P. E. (1968). Growth of the Long Bones in a Prehistoric Population from Sudanese Nubia. Thesis. University of
Utah.
Maresch, M. M. (1955). Linear growth of long bones of extremities ti'oru infancy through adolescence. American
Journal of Diseases of Children 89, 725 742.
Martin, D., & Armelagos, G. J. (1979). Morphometrics of compact bone: an example from Sudanese Nubia.
American Journal of Physical Anthroplogy 51, 571-578.
Siegel, S. (1956). Nonparametric Statistics for the Behavioral Sciences. New York: McGraw-Hill.
Swedlund, A. C. & Armelagos, G.J. (1969). Un recherche en paleodemographie: la Nubia Soudanaise. Annales:
Economic, Societes, Civilization 24, 1287-1298.
Todd, T. W. (1920). Age changes in the pubic bone, I: the male White pubis. American Journal of Physical
Anthropology 3, 285-344.
Van Gerven, D. P., Armelagos, G. J. & Bartley, M. H. (1969). Roentgenographic and direct measurement of
femoral cortical involution in a prehistoric Mississippian population. American Journal of Physical Anthropology 31,
23-38.
Van Gerven, D. P., Sandford, M. K. & Hummert, J. R. (1981). Mortality and culture change in Nubia's Batn el
Hajar. Journal of Human Evolution 10, 39~408.