Calibrating Paleodemography:The Uniformitarian Challenge...Turned
Problem: convert skeletal age distributions into demographic indicators
One size fits all?Is the problem wrong lasts?
Or, wrong way of measuring feet?
The Uniformitarian Hypothesis (Howell 1976):
Model life tables may be used to calibrate skeletal age distributions of paleo-populations
The Challenge (Paine 1997b:199):
Uniformitarian models (model life tables) may mask important biological and behavioral differences
Age distributions of paleopopulations may mask important biological and behavioral differences
Two kinds of “shoehorning”
Are paleodemographers “shoehorning skeletal series into expected shapes [of model life tables]”?--Paine 1997b
Are skeletal biologists shoehorning skeletal ages into uniform distributions, regardless of time or place?
Calibrate3 collections of dataagainst Coale & Demeny West
Health and Nutrition in the Western Hemisphere: 26 sites, n=5,787
18 published sites, n=5,792 8 Europe, 8 N. America, 1 Mesoamerica, 1 Africa
7 historical burial registers, n= 81,304 (4 Mexico, 3 USA)
Method: Compare
Against each other and a full range of model life tables
Using 4 kinds of statistics
- Proportional hazard models
- Buikstra ratio (d=deaths): d30+/d5+
- Bocquet-Appel & Masset ratio: d5-14/d20+
- Adult ratio: d20+/d35+
1.Uniformitarian Models:effects of fertility and mortality on human populations
Demographers know:fertility has the biggest impact on population age structure (and on the age distribution of deaths).Next figure shows fertility effects:
Fertility varies from GRR = 2 to 6 (average completed family size=4-12 children!)
Mortality is held constant (e0=20 years)
Spread for adults is proportionally large.
Fertility has big effects on age structure of deaths
GRR =2 , 3, 4, 5, 6; e0 = 20
Great variations in mortality have small effect on age distribution of skeletons.The next figure shows mortality effects (stable populations):
Life expectancy varies from 20 to 50 years
Fertility is held constant at GRR = 3.
Spread for adolescents and young adults is small.
Mortality offers a small target
GRR = 3; e0 = 20, 30, 40, 50
Demographers know:fertility has the biggest impact on population age structure (and on the age distribution of deaths).Next figure shows fertility effects:
Fertility varies from GRR = 2 to 6 (TFR=4-12!)
Mortality is held constant (e0=20 years)
Spread for adults is proportionally large.
Calibrate--combining great variations in both fertility and mortalityThe next figure provides stable population backdrop for skeletal data
Accepts the uniformitarian heresy
Fertility ranges from GRR = 2 - 6
Mortality ranges from e0=20, 50
PPT Slide
Model populations:GRR = 6-2; e0=50, 20
2.The Uniformitarian Challenge ...Turned:proportional hazard models of (a) stable populations, (b) Lerna (3600-4000 BP),and (c) Dallas (1915-45)
Whopper assumptions and faux hazard rates
1. Accept whopper assumptions: stable population; no deposition, recovery or aging biases.
2. Compute proportional hazard “rates” for observed and model data (derived solely from age distributions of deaths for both).
3. Examine the shape of the curve of deaths, ignoring youngest (0-4) and oldest.
4. Do fertility-centered paleodemography.
5. Recall difference between real and faux...
Ancient Lerna & modern Dallas TX (African-Americans)--theuniformitarian challenge...turned
“Jaws” effect in skeletal materials (HNWH), but not historical sources.
PPT Slide
PPT Slide
PPT Slide
European prehistoric populations also show “Jaws” effect
Regardless of ancestry, there is a consistent pattern of unprecedentedly high hazard rates from age 25-35 and beyond
Is conventionaluniformitarianism wrong?Next slide: 19th c. Africa...
Consider a near perfectly registered ultra-high mortality population undergoing seasoning.
McDaniels and Preston’s (1994) high mortality model life tables based on Liberia (American Black emigrants in early 19th century)
PPT Slide
Seasoning mortalityof African-Americanssettling in 19thc Liberiais similar to Coale & Demeny models
3. Simple ratios:a. Bocquet-Appel & Masset’s (1982) “Juvenility” index: D5-14/D20+
Minimal requirements for aging
Maximum separation with model life tables
Unfortunately, confidence intervals not defined
“Juvenility” index: Fertility has strong effects; mortality, weak
“Juvenility” index: 4 non-Amerindian populations (HNWH)
“Juvenility” index: HNWH North American data
“Juvenility” index: Meso and South American skeletal populations
Juvenility IndexWhat does it say?
Model populations: great promise
Skeletal datasets: interpretation is problematic
Lack of confidence intervals...
3. Simple ratios:b. Buikstra, Konigsberg & Bullington ratio (1986): D30+/D5+
Minimal requirements for aging
Moderate separation with model life tables
Confidence intervals readily defined (25P5)
Calibrating the Buikstra ratio, including 95% confidence interval
Native American collections show wide variation--most with GRR 6+
Meso- and South American sets--most with GRR 6+, except Marquez
Buikstra ratio:potential unrealized in practice
Point estimates bounded by confidence intervals of GRR +/- 1
Great, puzzling variability
Extremely high fertility (GRR ɰ) suggests too few specimens aged 30+
Yet too few aged 5-9 also likely
3. Simple ratios:c. Adult ratio: D35+/D20+
Minimal requirements for aging
Deposition bias less likely
Moderate separation with model life tables
Confidence intervals readily defined (15P20)
Calibrating the Adult ratio, including 95% confidence interval
Six of 13 HNWH Meso and South American collections look OK
One, perhaps two, of 9 HNWH look OK
Published North Americansets: 4 of 8 look OK
Published European/African sets: 2 of 7 OK
NE Cities--not credible (inmigrants)rural Mexican parishes OK
Adult ratio:potential unrealized in practice
Point estimates bounded by confidence intervals of GRR +/- 1
Great, puzzling variability
Extremely high fertility (GRR ɰ) suggests too few specimens aged 30+
Many range beyond credible levels
4. Conclusions
1. Visual analysis over statistics
2. Pitfalls of paleodemography
3. Of shoes and shoehorning: orthodox uniformitarianism reaffirmed
Conclusions, I: 5 reasons why visuals reveal the uniformitarian turn (and statistics do not)
1. Graphs use hazard rates, not percentages or simple counts.
2. Rates are drawn on a logarithmic scale to show proportional differences.
3. Confidence intervals shown for all data points.
4. Empirical data are displayed over a wide range of stable populations.
5. Graphs facilitate visualization of data and models, whereas statistics may conceal more than they reveal.
II: Paleodemography is methdologically challenged--remember Jaws!!!
1. Skeletal collections are too small to compute rates with much confidence.
2. There is systematic bias in depositing, recovering and/or aging skeletons (too few from age 25/35 and too many for adolescents and young adults) regardless of period or ancestry.
--Historical populations reconstructed from written documents do not show these patterns.
--Nor do African seasoning data (which support orthodox uniformitarianism).
III: Shoes and shoehorning
Lasts are probably OK (Orthodox uniformitarianism reaffirmed)
Whopper shoehorn required for assumptions:
- Real populations are more likely chaotic than stable
- Remains are more likely selective (for sex, age, and status) than complete
END