close

Вход

Забыли?

вход по аккаунту

?

An elusive paleodemography A comparison of two methods for estimating the adult age distribution of deaths at late Classic Copan Honduras.

код для вставкиСкачать
AMERICAN JOURNAL OF PHYSICAL ANTHROPOLOGY 132:40–47 (2007)
An Elusive Paleodemography? A Comparison of Two
Methods for Estimating the Adult Age Distribution
of Deaths at Late Classic Copan, Honduras
Rebecca Storey
Department of Anthropology, University of Houston, Houston, TX 77204-5020
KEY WORDS
skeletal adult age estimation; bayesian estimation; auricular surface seriation
ABSTRACT
Comparison of different adult age estimation methods on the same skeletal sample with
unknown ages could forward paleodemographic inference, while researchers sort out various controversies.
The original aging method for the auricular surface
(Lovejoy et al., 1985a) assigned an age estimation based
on several separate characteristics. Researchers have
found this original method hard to apply. It is usually
forgotten that before assigning an age, there was a seriation, an ordering of all available individuals from youngest to oldest. Thus, age estimation reflected the place
of an individual within its sample. A recent article
(Buckberry and Chamberlain, 2002) proposed a revised
method that scores theses various characteristics into
age stages, which can then be used with a Bayesian
method to estimate an adult age distribution for the
sample. Both methods were applied to the adult auricular surfaces of a Pre-Columbian Maya skeletal population from Copan, Honduras and resulted in age distributions with significant numbers of older adults. However,
contrary to the usual paleodemographic distribution, one
Bayesian estimation based on uniform prior probabilities
yielded a population with 57% of the ages at death over
65, while another based on a high mortality life table
still had 12% of the individuals aged over 75 years. The
seriation method yielded an age distribution more similar to that known from preindustrial historical situations, without excessive longevity of adults. Paleodemography must still wrestle with its elusive goal of accurate
adult age estimation from skeletons, a necessary base
for demographic study of past populations. Am J Phys
Anthropol 132:40–47, 2007. V 2006 Wiley-Liss, Inc.
Paleodemography based on skeletons has been a contentious and ever-evolving undertaking for the last 40
years (Petersen, 1975; Hoppa, 2002), but recent controversy has questioned how to determine, with some accuracy, an age distribution of deaths. This remains one of
the useful pieces of demographic information available
about past populations. However, skeletal age distributions differ from the age distribution of deaths of recent
and historical populations. Skeletons can be aged fairly
accurately, with well-defined estimates of error, as juveniles and young adults younger than 30 years (Cox,
2001). However, current methods for estimating the age
of older adults are considered to lack both precision and
accuracy (Cox, 2001). The ultimate question for paleodemography is whether traditional skeletal aging techniques have revealed a real pattern for past populations,
or are they just responsible for what are actually deviant
patterns of ages at death? Questions do arise, because
skeletal samples often have estimated age distributions
of deaths that are different from those of model life
tables often used by researchers as emblematic of the
human pattern. Skeletal samples seem to have higher
proportions of prime age adults and low proportion of
individuals over 50 years of age at death (Storey, 1992;
Paine, 1997).
Model life table patterns are based on recent historical
populations with unprecedented longevity, while skeletally derived patterns have very low overall survivorship
with high adult mortality levels compared to those during childhood (Meindl and Russell, 1998). Historical
demographers have indicated that the relationship of
infant/child mortality to adult mortality can differ from
that in model tables (Woods, 1993; Wrigley et al., 1997).
From the reconstitution studies of early modern England, it was not until the second half of the 18th century
that child and adult mortality rates have the relationship found in the Coale et al. (1983) Princeton model
tables (Wrigley et al., 1997). ‘‘Viewed in terms of the
Princeton North or West model life tables adult mortality was far too high relative to rates in infancy and
childhood in the seventeenth century’’ (Wrigley et al.,
1997). There is evidence of the plasticity of the age structure of deaths, and thus, the differences of paleodemographic skeletal samples from models should not be dismissed as erroneous (Meindl and Russell, 1998).
The similarity of the paleodemographic pattern even
in very different cultural and temporal circumstances
(Paine, 1997) does seem to suggest that traditional age
estimation techniques are more likely to be the problem.
Present methods seem to slightly overage young adults
and then dramatically underage older individuals (Molleson and Cox, 1993). Certainly, recent efforts have been
devoted to improving age estimation techniques as a first
C 2006
V
WILEY-LISS, INC.
C
Grant sponsors: Government of Honduras; Fulbright Foundation;
University of Houston.
Correspondence to: Rebecca Storey, University of Houston,
Department of Anthropology, 4800 Calhoun Boulevard, Houston, TX
77204–5020. E-mail: rstorey@uh.edu
Received 19 December 2005; accepted 8 August 2006.
DOI 10.1002/ajpa.20502
Published online 31 October 2006 in Wiley InterScience
(www.interscience.wiley.com).
ELUSIVE PALEODEMOGRAPHY
step to answering the ultimate question in paleodemography of whether aging methods are the cause of the
skeletal age distributions.
Skeletal samples have a distribution of age indicators,
and the paleodemographer attempts to recover the distribution of ages responsible for the pattern of indicators. This is the crux of the recent Rostock Manifesto
(Hoppa and Vaupel, 2002), a concerted attempt to develop better ways to estimate the age structure of a skeletal population. Reference skeletons of known age are
used to validate age indicators, but the criticism of much
past practice in paleodemography is that age distributions for skeletons tend to mimic the underlying age distribution of reference populations (Bouquet-Appel and
Masset, 1982). While certainly not all agree (Van Gerven
and Armelagos, 1983), most researchers seem to accept
this as a problem and have forwarded solutions (Chamberlain, 2001; Hoppa and Vaupel, 2002; Konigsberg and
Frankenberg, 2002).
Of course, one also has to wonder about the underlying uniformitarian assumption (Howell, 1976; Hoppa,
2002), whether recent, documented populations are good
models for aging individuals from very different environments and lives. In fact, there is much to cast doubt on
it, as aging is recognized to be quite plastic and to be
influenced by both genetics and environment. However,
there is really no alternative to the use of the mostly
modern reference populations available, but the uniformitarian assumption needs extensive testing with welldocumented historical samples. While awaiting further
research, what can researchers do with skeletal populations lacking historical documentation but which they
need to analyze? One answer is to compare various adult
age estimation methods available for the same skeletal
indicator and hope that the results might be commensurable so as to strengthen interpretations.
An acceptable method for estimating adult age must
not dramatically underage older individuals, and it should
not yield an age distribution of deaths that mimics the
reference population on which the method was developed. Recent methods have advocated employing Bayesian estimation to relate age-related morphological stages
of an indicator from a reference population to those of a
skeletal sample (Ackroyd et al., 1999; Hoppa and Vaupel,
2002). This method allows the calculation of the probability of age conditional on the stage of the age indicator,
and the probability of a skeleton being a particular age
at death, given the stage of the indicator. It is supposed
to remove the mimicry of the reference sample, because
the information from the reference population is mediated by the use of an appropriate prior probability not
dependent on the age distribution of deaths in the reference population.
The selection of a good age indicator is of more importance to the whole problem. The pubic symphysis of the
pelvis has long been a favorite age indicator for adults,
and a set of Bayesian posterior probabilities for estimating age has been presented (Chamberlain, 2001). However, the pubic symphysis is likely a poor indicator of
advanced adult age, and its use only exacerbates the
underlying problem of underaging, and thus underestimation, of old adults. The pubic symphysis completes its developmental changes during the fourth decade and that is
the limit of its use for age estimation, as later changes are
all degenerative and highly variable (Lovejoy et al., 1997).
The auricular surface of the pelvis promises a better
method for estimating adult age, because it appears to
41
track age changes into older ages and tends to be better
preserved in most archaeological collections (Meindl and
Russell, 1998). The original age estimation method of
Lovejoy et al. (1985b) presented eight modal 5-year
stages, from 20 to 60þ, based on features of surface texture, transverse organization, porosity, and arthritic
changes. The modal stages were developed using skeletons from the early 20th century Haman-Todd collection
as the reference sample. This method has been criticized
because many auricular surfaces do not fit easily into
modal stages, and it tends to overage the young and
underage older individuals (Saunders et al., 1992). The
features overlap various modal stages that make it difficult to assign an individual surface to a stage (Buckberry and Chamberlain, 2002). What is usually forgotten
and not applied by critics is that the method requires
that age stages are assigned only after auricular surfaces are seriated from youngest to oldest. The advantages of the seriation method include comparing all
individuals and checking for consistency of application
during the process of assigning estimated ages (Lovejoy
et al., 1985a). From personal experience, the method also
makes clear the age range present in the sample and
discourages underestimating how many older individuals
are present. The seriation method has been recently
criticized because it really does not solve the problem of
mimicry of the reference sample and also does not provide for the measurement of statistical uncertainty in
age estimates that many researchers now believe must
be part of the process (Konigsberg and Frankenberg,
2002).
Several recent systems to make features of the auricular surface easier to score and to improve age estimation
for adults have been published, one using transition
analysis (Boldsen et al., 2002) and one, multiple regression (Igarashi et al., 2005). Buckberry and Chamberlain
(2002) also devised a quantitative scoring system based
on five aging characteristics from the original Lovejoy
and coworkers descriptions to make age estimation easier and more replicable to use. Two characteristics have
five states: transverse organization, scored from one
(90% of the surface transversely organized) to five (no
transverse organization) and surface texture, from one
(90% finely granular) to five (50%þ dense bone). Three
characteristics have only three stages: microporosity and
macroporosity (from one, no micro- or macroporosity,
two, on one demiface, three on both demifaces) and apical changes (from one, sharp and distinct, to three, apex
no longer a smooth arc). These are totaled into composite
scores for each surface (maximum of 19), which are then
grouped into seven age stages. The composite scores and
age stages were based on the Spitalfields known age
sample from the 18th and 19th centuries (Molleson and
Cox, 1993). To remove the effect of the reference population age structure, uniform prior probabilities were used
to estimate posterior probabilities of age for use with
other skeletal samples (Table 1). This is the simplest
model for prior probabilities in Bayesian estimation.
Ideally, all auricular surface aging methods should be
compared by tests on a different reference population
with known ages, but the opportunity to do so has not
yet materialized or been attempted. Although the transition analysis method (Boldsen et al., 2002) was published the same year as Buckberry and Chamberlain,
lack of familiarity with the former method at the time of
the field season in 2004 limited comparison to only the
latter and seriation methods. However, the focus here is
American Journal of Physical Anthropology—DOI 10.1002/ajpa
42
R. STOREY
TABLE 1. Quantitative scoring posterior probabilities by stage1
Age stage
Age
15–24
25–34
35–44
45–54
55–64
65–74
75þ
I
0.86 (1.00)
0.14 (0.00)
0.00
0.00
0.00
0.00
0.00
II
0.33
0.42
0.21
0.04
0.00
0.00
0.00
(0.28)
(0.48)
(0.24)
(0.00)
III
0.27
0.18
0.20
0.12
0.07
0.02
0.13
(0.38)
(0.16)
(0.32)
(0.09)
(0.02)
(0.00)
IV
0.00
0.17
0.07
0.16
0.20
0.14
0.27
(0.27)
(0.24)
(0.15)
(0.06)
(0.12)
V
0.00
0.03
0.19
0.20
0.18
0.26
0.13
VI
(0.10)
(0.05)
(0.24)
(0.23)
0.00
0.00
0.02 (0.03)
0.04 (0.09)
0.11 (0.19)
0.10 (0.19)
0.73 (0.49)
VII
0.00
0.00
0.00
0.27
0.00
0.24
0.49
(0.09)
(0.07)
(0.08)
(0.77)
1
Adapted from Buckberry and Chamberlain (2002): Table 13. The last two age groups (75–94) have been combined here as 75þ.
Numbers in parentheses represent corrections to the published posterior probabilities (personal communication, A. Chamberlain).
on the comparison of aging methods to a skeletal sample
and what inferences might be made about the age distribution of deaths, pending further tests of aging methods.
More specifically, the interest is in how auricular surface
methods compare in estimating the number of older individuals.
How different are the age distributions of deaths from
the Lovejoy and coworkers and Buckberry and Chamberlain methods? The best outcome would be similar
results, as this might indicate what the pattern of adult
longevity in the sample might be. One difficulty in comparing the seriation and quantitative scoring methods is
that the raw data for the method is fully reported in
Buckberry and Chamberlain, with scores and ages for all
individuals used in the Spitalfields reference sample.
This will allow a test for mimicry. Lovejoy et al. (1985a)
only report on the bias test by decade for their method
and only that the age distribution of deaths for the Todd
skeletons was chosen for similarity to a survivorship
curve chosen from the literature. The particular survivorship curve with numbers of individuals by age is not
specified. Only the total numbers of reference individuals were given. The bias tests seems to show slight overaging of the young and underaging of the older adults,
although all of the mean errors are less than 10 years,
even for the 60þ individuals. There is no way to tell if
the resulting age distribution mimics the Todd skeletons.
Although the seriation method certainly falls into the
more traditional age estimation methods that depend on
experience and the ‘‘intuition’’ of the osteologist, it must
be demonstrated that the quantitative scoring method
will prove superior.
METHODS AND MATERIALS
Both auricular surface methods were applied to Late
Classic skeletons from the Maya site of Copan, Honduras. The Maya were an important Pre-Columbian civilization, and Copan was one of the major centers during
the Late Classic period (A.D. 650–1000), the apogee at
Copan (Coe, 1999). After the Late Classic, the Copan
area, in common with many nearby Maya sites, was
gradually abandoned. After a period of expansion, the
Copan population began to decline after the nineth century, declining by almost half in the 10th century and
more gradually thereafter (Webster, 2002). All individuals were recovered in and around residences, representing the attritional mortality of the period without any
large accumulations of skeletons indicating catastrophic
mortality. The skeletal sample is large, comprising more
than 400 individuals spanning over a thousand years,
but the bulk of the sample is from the Late Classic period. Only individuals from this period were included in
the sample. These individuals come mostly from elite
residences, but almost one-third were recovered from
lower-status residences from both urban and rural contexts. Thus, the full range of Late Classic Copan society
is represented.
Adults over age 20 were the focus, although one individual probably closer to age 17 was included (youngest
individual in sample with a complete auricular surface).
Only 107 individuals had well-preserved auricular surfaces. No differences by sex were observed in the aging criteria of either method (Lovejoy et al., 1985b; Buckberry
and Chamberlain, 2002); therefore, sexes will be pooled
in the Copan sample. There were 49% females and 51%
males in the sample.
Seriation is intended to allow consistency in age determination, as one continually judges the aging characteristics and does not assign an age until the end, to avoid
inadvertent shifts in observer criteria through time
(Lovejoy et al., 1985a), as was stressed during personal
training in the method. Multiple seriations, each at least
a year apart, were conducted in the 20 years the sample
has been studied. Each seriation was carried out without
any notes on age estimations previously determined, but
each time new individuals, as they were studied that
field season, were included. The maximum number of
seriations for one individual surface is four, and 36 surfaces have been seriated more than once. The seriated
surfaces were placed into the eight modal age stages.
The Buckberry and Chamberlain method was applied
in 2004 to 79 surfaces with the help of one graduate student, who has only limited experience with adult skeletal aging. The distribution of ages was calculated as suggested by Chamberlain (2000) and Buckberry and Chamberlain (2002), where the individuals in each of the
seven age stages were distributed proportionally according to the posterior probabilities of each 10-year age
group. So, all individuals who scored in age stage II, for
example, had 0.33 placed 15–24, 0.42 in 35–34, and so
on (Table 1). Thus, the technique is not intended for
individual auricular surface estimation, but to provide
an estimation of ages based on the distribution of age
indicators within a skeletal sample.
RESULTS
In applying the method of Buckberry and Chamberlain, the scoring of the characteristics for percent of
transverse organization and surface texture changes
caused disagreement on about 20% of the surfaces
between the two observers. The other three characteristics were more easily scored, although it seemed that
more than three categories for these would have been
helpful. This would, however, have probably increased
American Journal of Physical Anthropology—DOI 10.1002/ajpa
ELUSIVE PALEODEMOGRAPHY
43
Fig. 1. Distribution of scores on the aging characters of Buckberry and Chamberlain for Copan auricular surfaces. All scores
are from youngest to oldest character. Transverse organization is scored from 1 ¼ 90% surface is transversely organized to 3 ¼
25%–49% is transversely organized to 5 ¼ no transverse organization; Surface texture from 1 ¼ 90% fine grained to 3 ¼ 50% coarse
grain to 5 ¼ 50% dense bone; Microporosity and macroporosity from 1 ¼ none to 3 ¼ present on both demifaces; Apical changes
from 1 ¼ none to 3 ¼ irregular, no longer a smooth arc (Buckberry and Chamberlain, 2002).
Fig. 2. Age distribution of
deaths for Copan by Bayesian
estimation compared with Spitalfields reference sample. The
Copan distribution is based on
the corrected uniform prior probabilities, while the C&D model
is based on priors derived from
West Model 1 (Coale et al. 1983).
the amount of disagreement. Out of a total of 395 observations (5 times 79), there was agreement on 352 (89%).
The distribution of scores for each characteristic reveals that the older states predominate in Copan (Fig. 1).
Scores of one and two indicate the younger states for
each characteristic. Only macroporosity had a majority
of surfaces with less than score three, while the other
characteristics are dominated by scores indicating the
older character states. Apical change is the only character impacted by preservation, but still scores of three,
the oldest, are the most common. Application of the posterior probabilities to the composite scores calculated by
Buckberry and Chamberlain and applied as they suggest
(Table 1) yields a very old age distribution of deaths
(Fig. 2). Contrary to the more usual age distribution of
deaths in paleodemography, Copan here has 57% of the
individuals over 65 years old at death, and only 19%
younger than age 45. It is even significantly older than
the Spitalfields known-age individuals used as the reference sample underlying the probabilities (K-S test, P <
0.01). Thus, the Bayesian estimation prevents simple
replication of the Spitalfields age distribution, but the KS test does not prove that there is no mimicry, only that
the two distributions are significantly different in some
way. Copan also has more individuals aged over 65 than
Spitalfields, even though Spitalfields does have a relatively old age structure with underrepresentation of
young adults according to Buckberry and Chamberlain
(2002: 236).
The posterior probabilities (Table 1) have several unrealistic characteristics, such as zero probability of being
55–64 in Stage 7. There is also a higher probability of
being 75þ if in auricular Stage 6 than Stage 7. Since
Copan has 38% of the sample in Stage 6 and only 18%
in Stage 7, these sample measures reveal potential problems with the application of this method. Correction of
these and other errors in the original posterior probabilities (Table 1) increases the proportion aged less than 45
years to 23% versus 19% and decreases those over age
65 from 57% to 52%. However, these corrections do not
American Journal of Physical Anthropology—DOI 10.1002/ajpa
44
R. STOREY
Fig. 3. Age distribution of
deaths by seriation method. Seriation 1 is based on the youngest
age estimations, while Seriation 2
is based on the oldest.
otherwise change the significant difference from the Spitalfields reference sample.
The Rostock Manifesto (Hoppa and Vaupel, 2002) indicates that a likely model of the distribution of life spans
in the skeletal population should provide prior probabilities for the age estimation. Chamberlain (2001) suggested that model life tables could provide such priors. A
set of posterior probabilities based on Coale et al. (1983)
West Level 1 model life table (e0 ¼ 20 years) may provide a reasonable alternative to the uniform priors (probabilities from personal communication, A. Chamberlain)
(Fig. 2). The resulting age distribution of deaths for
Copan is younger, significantly different from the original Bayesian estimation based on uniform priors (K-S
test, P < 0.01), but not from the Spitalfields age distribution (K-S test, P > 0.05).
With the seriation method, there were a total of 63
multiple agings (on 36 surfaces). Of these, 22 (35%)
resulted in different age classes in different seriations,
almost always with less than 5 years difference in age
estimation. Nine of these cases (14%) had a two to three
stage difference (changing the age estimate by 10–15
years). However, two surfaces also fell into the exact
same modal stage twice. The rest (39 instances) placed
in the same modal stage two, three, or four times. Thus,
there were shifts through time in the observer’s criteria,
and the intraobserver error could be considered high (at
best, 14%). The results of the youngest age distribution
were compared with the oldest one, to see what effect
these 22 discrepancies had on the age distribution of
deaths.
Seriation also yields a somewhat older age distribution, with the majority of individuals aged as over 40
years at death. Seriation 1 uses all the youngest estimates, and Seriation 2 the oldest (Fig. 3), but there are
no statistical differences (K-S test, P > 0.05) between
the two age distributions. There are any number of reasons why there could be such divergent estimations on
the seriations by one researcher, but probably the single
most important is that each seriation took place on the
last day in the field season, so there was no opportunity
to leave them and return to recheck after a few days.
The differences could also have been influenced by the
particular combination of surfaces to be seriated; more
younger or older ones might have made a surface appear
to be in a different modal age than the original estimation. One of the criticisms of a method that depends on
classification of an entire surface into one age category
is that it is unlikely that all aging features will senesce
at the same time, so that it is difficult to assign a surface
clearly to one age stage (Boldsen et al., 2002). This criticism is not totally unwarranted. The seriation method is
based on assessing all surfaces simultaneously, and conducting multiple seriations clearly does not make up for
that lack. On the other hand, there is no change in the
ultimate age distribution of deaths. Mean age at death
for Seriation 1 is 44.4 years versus 46.6 years for Seriation 2. It is not possible to compare the seriation results
here with the original Todd distribution to judge mimicry, as the distribution used by Lovejoy and et al.
(1985b) has not been published.
Figure 4 compares the age distributions from both
Bayesian estimations and the seriation method for the
Copan sample. The seriation has more individuals aged
25–49, a common finding of paleodemographic samples
and different from many model life tables. In Seriation
2, the bulk of deaths (71%) occurred between ages 35–
60, while 18% were estimated to be older than age 60.
Both Bayesian estimations have more deaths over age
60. Mean age of the seriation distribution is 46.6 years,
while that of the corrected quantitative scoring is 63
years and that based on the Coale and Demeny model
table is 56. Both are significantly different from the
seriation distribution (K-S tests, P < 0.01). Unfortunately, the results of the two methods were not similar.
The Bayesian estimations yield an age distribution with
significant numbers of older adults, but their adult longevity may be older than reasonable for a pre-industrial,
Pre-Columbian skeletal sample like Copan.
Given our understanding of historic pre-modern mortality patterns (Jeune and Vaupel, 1995), it seems unrealistic that the majority of adult deaths would be over
age 65 and 40% would be older than age 75, as in the
estimation based on uniform prior probabilities. The evi-
American Journal of Physical Anthropology—DOI 10.1002/ajpa
ELUSIVE PALEODEMOGRAPHY
Fig. 4. Comparison of age distribution of deaths for Copan
adults by seriation and Bayesian estimation. Seriation 2 is
based on the oldest age estimations. The other two distributions
were derived by Bayesian estimation. The C&D model is based
on the high mortality model life table West Level 1 (Coale et al.
1983), while the uniform model is based on the corrected uniform prior probabilities.
dence available seems to indicate that relatively few
individuals who entered adulthood lived to age 50 and
beyond. Some of this evidence about adult longevity is
based on skeletons. For example, Boldsen and Paine
(1995) look at what might be learned in post-glacial
Europe. Given the problems with skeletal samples, they
found that they cannot determine the probabilities of
reaching ages older than 90. Instead, the probability of
surviving to age 40, having reached age 20, was only
about 40% throughout most of the history. Increases in
adult survivorship only occur much later. Boldsen
(1995), looking at the site of Tirup, Denmark, for medieval mortality, found that less than 25% of males and
15% of females reached age 50. With 700 births, not
even one male or female would have reached 80, and
mortality during adulthood is about five times as high as
in contemporary Danes. Of course, these are based on
skeletons and thus might suffer from the myriad age
estimation problems.
There is also evidence based on historical documentation from pre-modern populations. Based on censuses
from second to third centuries AD in Roman Egypt,
Bagnall and Frier (1994) determined that only about
1/5 of females lived from their teens into their 60s.
Zhao (1995) demonstrates from fairly large Chinese
genealogies that a few individuals were able to reach
over 95 years at death. Also, Zhao suggests that adult
mortality for the past thousand years has been high
and fairly stable, and that information from such Chinese genealogies indicates that life expectancies at
ages 30 and 50 were lower than those in Coale and
Demeny model table West Level 5, which are 28.3 and
16.9 more years, respectively, for females. Fluctuations
in mortality levels among adults were also found by
Wrigley et al. (1997) in late 16th and 17th century
England. Age structure of mortality is plastic. For their
earliest period, life expectancy at age 25 was 30 years
with survivorship to age 55 then *50%. Some few data
45
on males from earlier times suggest very high levels of
mortality, where life expectancy at age 20 was perhaps
only 27 more years on average (Wrigley et al., 1997).
Using historical life tables, Thatcher (1995) finds that
in the 18th century England the chance of living to 100
was 1 in 100,000. He has also constructed some life
tables from age 50 upwards, which indicate that about
40–50% survive to age 70, but then only 10–20% to age
80. Wilmoth (1995), whose main purpose was to estimate the proportion of centenarians, does conclude that
they would be present with what he calls the birth
of civilization, not because mortality improved, but
because world population would be sufficiently large.
More interestingly, like Zhao, Wilmoth finds that high
levels of adult mortality have been present until the
last couple of centuries and that life expectancy at age
50 was about 14 more years. Wilmoth points out that
the earliest dependable life tables underlying Coale
and Demeny have life expectancies at birth of 33.4 for
males and 35.5 for females, and these were already in
the industrial era. At least from available data, it does
seem that pre-modern adult mortality would not have
produced many individuals over 75 years at death.
The historical patterns also indicate that few adults
survived to age 60, and perhaps age 40, in some situations.
Coming from a very different environment and culture
than the historical situations discussed above, is it likely
that Copan would have elevated survivorship to old age?
Copan represents a tropical environment with a stone
technology in a Pre-Columbian period. Diet was overwhelmingly maize-based with little protein input during
the Late Classic, based on stable isotope analysis of some
skeletons (Reed, 1994). Evidence of morbidity is ubiquitous on the skeletons (Storey et al., 2002), and there is
nothing to suggest that Copan would have been particularly advantaged over other preindustrial populations.
On the other hand, the historical data indicate that the
age distribution of deaths produced by seriation (higher
mortality in ages 30–50 than industrial-era populations,
with only about 1/5 attaining ages older than 60) are
reasonable by current understanding. The Bayesian estimation based on the Coale et al. (1983) high mortality
life table is closer to the historical data, but still has
56% of the deaths over age 55. The auricular surface
indicators do indicate a generally older adult population,
whether in the original Lovejoy et al. characteristics or
in the Buckberry and Chamberlain system.
The Buckberry and Chamberlain posterior probabilities are an attempt to overcome some of the problems
with adult age estimation and provide more opportunity
to extend the older ages in a skeletal age distribution of
deaths. But these may provide an over-correction. A criticism of uniform priors to obtain posterior probabilities
is that they place disproportionate weight on very old
and unlikely ages at death (Boldsen et al., 2002). The
older Spitalfields reference sample simply exacerbated
this tendency. The Coale and Demeny models have
weaknesses already mentioned. The very high mortality
model still yields an age distribution with a majority
older than age 55, though more plausible than the uniform priors estimation. The e50 for this estimation is 16
more years, which is only slightly higher than Wilmoth’s
estimate, but the proportion aged as 75þ years is 12%,
which may still be unrealistic. Unfortunately, the
Bayesian estimations still need tweaking. The seriation
method results in a more even loss of adults from age 45
American Journal of Physical Anthropology—DOI 10.1002/ajpa
46
R. STOREY
on. While this method may tend to overestimate proportions in their 30s and 40s at death, it seems a more realistic age distribution of deaths for this population. The
e50 is 11.6 more years for Seriation 2, which is not much
less than the e50 ¼ 14 estimate for preindustrial populations (Wilmoth, 1995).
CONCLUSION
The auricular surface of the pelvis has the potential to
produce more realistic age estimations for adults and aid
in the recovery of the age distribution of deaths of a
skeletal sample for which there are no historical data.
Two methods of age estimation have been contrasted on
the Copan Classic Maya skeletons, suggested as one possible way to improve current paleodemographic inference, given all the uncertainties surrounding adult age
estimation. All resulting age distributions reveal an
older adult sample, which is what the archaeological
context would indicate. This was a population in decline
for at least half the Late Classic Period (for example,
Freter, 1997). In declining populations, older adults
should predominate, as they do here. Unfortunately, the
two methods yielded significantly different age distributions of deaths and do not make interpretation of Copan
easy. The more traditional age estimation method, seriation, although lacking in statistical rigor and clarity, provides results more in line with information from historical demography and provides perhaps a more reasonable
starting point for further demographic inference. The
result here shows that Bayesian estimation does increase the proportion of individuals aged over age 55 in
a skeletal sample, but the examples employed reveal the
importance of choosing reasonable priors. In particular,
the use of uniform prior probabilities should be avoided
in Bayesian estimation. More comparisons need to be
done of age estimation methods, and reasonable models
of the distribution of life spans in the past need to be
identified. Paleodemography is improving the estimation
of older adults when they are present, but the pattern of
adult longevity of the last two centuries is probably not
applicable to earlier populations.
ACKNOWLEDGMENTS
This article has benefited greatly from the comments
of Richard Meindl and the anonymous reviewers, for
which the author is very grateful. The Copan study is
conducted with the permission of the Instituto Hondureño de Antropologı́a e Historı́a. Many students and colleagues from the United States and Honduras have all
been involved in this study which would not have been
possible without their help and input.
LITERATURE CITED
Ackroyd RG, Lucy D, Pollard AM, Roberts CA. 1999. Nasty, brutish, but not necessarily short: A reconsideration of the statistical methods used to calculate age at death from adult
human skeletal and dental age indicators. Am Antiq 64:55–
70.
Bagnall RS, Frier BW. 1994. The demography of Roman Egypt.
Cambridge: Cambridge University Press.
Boldsen JL. 1995. Patterns of advanced age mortality in the medieval village Tirup. In: Jeune B, Vaupel JW, editors. Exceptional longevity: From prehistory to present. Odense: Odense
University Press. p 37–46.
Boldsen JL, Milner GR, Konigsberg LW, Wood JW. 2002. Transition analysis: A new method for estimating age from skeletons. In: Hoppa RD, Vaupel JW, editors. Paleodemography:
Age distributions from skeletal samples. Cambridge: Cambridge University Press. p 73–106.
Boldsen JL, Paine RR. 1995. Defining extreme longevity from
the mesolithic to the middle ages: Estimates based on skeletal
data. In: Jeune B, Vaupel JW, editors. Exceptional longevity:
From prehistory to present. Odense: Odense University Press.
p 25–36.
Bouquet-Appel J-P, Masset C. 1982. Farewell to paleodemography. J Hum Evol 11:321–333.
Buckberry JL, Chamberlain AT. 2002. Age estimation from the
auricular surface of the ilium: A revised method. Am J Phys
Anthropol 119:231–239.
Chamberlain AT. 2000. Problems and prospects in palaeodemography. In: Cox M, Mays S, editors. Human osteology in
archaeology and forensic medicine. London: Greenwich Medical Media. p 101–116.
Chamberlain AT. 2001. Palaeodemography. In: Brothwell DR,
Pollard AM, editors. Handbook of archaeological sciences. Chichester: Wiley. p 259–268.
Coale AJ, Demeny P, Vaughn B. 1983. Regional model life tables
and stable populations. New York: Academic Press.
Coe M. 1999. The Maya, 6th ed. London: Thames and Hudson.
Cox M. 2001. Assessment of age at death and sex. In: Brothwell
DR, Pollard AM, editors. Handbook of archaeological sciences.
Chichester: Wiley. p 237–247.
Freter A. 1997. The question of time: The impact of chronology
on Copan prehistoric settlement demography. In: Paine RR,
editor. Integrating archaeological demography: Multidisciplinary approaches to prehistoric population, (Occasional papers
24). Carbondale, IL: Center for Archaeological Investigations.
p 21–42.
Hoppa RD. 2002. Paleodemography: Looking back and thinking
ahead. In: Hoppa RD, Vaupel JW, editors. Paleodemography:
Age distributions from skeletal samples. Cambridge: Cambridge University Press. p 9–28.
Hoppa RD, Vaupel JW. 2002. The Rostock manifesto for paleodemography: The way from stage to age. In: Hoppa RD, Vaupel
JW, editors. Paleodemography: Age distributions from skeletal
samples. Cambridge: Cambridge University Press. p 1–8.
Howell N. 1976. Toward a uniformitarian theory of human paleodemography. In: Ward RM, Weiss KM, editors. The demographic evolution of human populations. New York: Academic
Press. p 25–40.
Igarashi Y, Uesu K, Wakaebe T, Kanazawa E. 2005. New
method for estimation of adult skeletal age at death from the
morphology of the auricular surface of the ilium. Am J Phys
Anthropol 128:324–339.
Jeune B, Vaupel JW, editors. 1995. Exceptional longevity: From
prehistory to the present. Monographs on population aging 2.
Odense: Odense University Press.
Konigsberg LW, Frankenberg SR. 2002. Deconstructing death in
paleodemography. Am J Phys Anthropol 117:297–309.
Lovejoy CO, Meindl RS, Mensforth RP, Barton TJ. 1985a. Multifactorial determination of skeletal age at death: A method
and blind tests of its accuracy. Am J Phys Anthropol 68:1–14.
Lovejoy CO, Meindl RS, Pryzbeck TR, Mensforth RP. 1985b.
Chronological metamorphosis of the auricular surface of the
ilium: A new method for the determination of adult skeletal
age at death. Am J Phys Anthropol 68:15–29.
Lovejoy CO, Meindl RS, Tague RG, Latimer B. 1997. The comparative senescent biology of the hominoid pelvis and its
implications for the use of age-at-death indicators in the
human skeleton. In: Paine RR, editor. Integrating archaeological demography: Multidisciplinary approaches to prehistoric
population, (Occasional papers 24). Carbondale, IL: Center for
Archaeological Investigations. p 43–63.
Meindl RS, Russell KF. 1998. Recent advances in method and
theory in paleodemography. Ann Rev Anthropol 27:375–399.
Molleson T, Cox M. 1993. The Spitalfields project, Vol. 2. The
anthropology: The middling sort. New York: Council for British Archaeology.
American Journal of Physical Anthropology—DOI 10.1002/ajpa
ELUSIVE PALEODEMOGRAPHY
Paine RR. 1997. The role of uniformitarian models in osteological paleodemography. In: Paine RR, editor. Integrating
archaeological demography: Multidisciplinary approaches to
prehistoric population, (Occasional papers 24). Carbondale,
IL: Center for Archaeological Investigations. p 191–204.
Petersen W. 1975. A demographer’s view of prehistoric demography. Curr Anthropol 16:227–237.
Reed DM. 1994. Ancient Maya diet at Copan, Honduras, as determined through the analysis of stable carbon and nitrogen isotopes. In: Sobolik KD, editor. Paleonutrition: The diet and
health of prehistoric Americans, (Occasional paper 22). Carbondale, IL: Center for Archaeological Investigations. p 210–221.
Saunders SR, Fitzgerald C, Rogers T, Dudar JC, McKillop H. 1992.
A test of several methods of skeletal age estimation using a documented archaeological sample. Can Soc Forensic Sci J 25:97–118.
Storey R. 1992. Life and death in the ancient city of Teotihuacan. Tuscaloosa: University of Alabama Press.
Storey R, Marquez Morfin L, Smith V. 2002. Social disruption
and the Maya civilization of Mesoamerica: A study of health
and economy of the last thousand years. In: Steckel RH, Rose
JC, editors. The Backbone of History: Health and nutrition in
the Western Hemisphere. Cambridge: Cambridge University
Press. p 308–338.
47
Thatcher AR. 1995. A note on some historical data on old age
mortality. In: Jeune B, Vaupel JW, editors. Exceptional longevity: From prehistory to present. Odense: Odense University Press. p 117–124.
Van Gerven DP, Armelagos GJ. 1983. ‘‘Farewell to paleodemography?’’ Rumors of its death have been greatly exaggerated. J
Hum Evol 12:353–360.
Webster D. 2002. The fall of the ancient Maya. London: Thames
and Hudson.
Wilmoth JR. 1995. The earliest centenarians: A statistical analysis. In: Jeune B, Vaupel JW, editors. Exceptional Longevity:
From prehistory to present. Odense: Odense University Press.
p 125–169.
Woods R. 1993. On the historical relationship between infant
and adult mortality. Popul Stud 47:195–219.
Wrigley EA, Davies RS, Oeppen JE, Schofield RS. 1997. English
population history from family reconstitution 1580–1837.
Cambridge: Cambridge University Press.
Zhao Z. 1995. Record longevity in Chinese history—evidence
from the Wang genealogy. In: Jeune B, Vaupel JW, editors.
Exceptional longevity: from prehistory to present. Odense:
Odense University Press. p 93–108.
American Journal of Physical Anthropology—DOI 10.1002/ajpa
Документ
Категория
Без категории
Просмотров
5
Размер файла
267 Кб
Теги
two, death, distributions, honduras, method, latex, comparison, paleodemographic, adults, copan, classic, age, estimating, elusive
1/--страниц
Пожаловаться на содержимое документа