Length of life in the ancient world: a controlled study

[ASCLEPIADES]

J D Montagu

The length of life for a population of ancient Greek

and Roman men with quoted dates of birth and death

has been compared statistically with three sample

populations from different periods. It was found that

the ancients who were born before 100 BC lived as

long as the moderns who died before 1950; they lived

significantly shorter lives than those who have died

in the present half century.

 

FULL ARTICLE HERE

 

Salve, guys! What do think about this study?

[ASCLEPIADES]

Here is the introduction of this study:

 

It is axiomatic that the inhabitants of the ancient world did not live as long as we do today. For example, Deevey(1) reported that the average length of life in ancient Greece was 35 years whereas in classicalRome it was even shorter at 32 years. Dorn (2) revised these figures downwards to less than 30 years and contrasted them with the average of 70 years in North

America and Western Europe in 1959-1960. In a more recent survey Garland3 concludes that life expectancy for both men and women in ancient Greece was 'well below fifty'.

 

The figures quoted by Deevey and Dorn are estimates of the average length of life of every member of the sample population. This includes those who succumbed in infancy and early childhood, thevictims of fatal accidents, and those who met violent deaths. However, did the survivors of these perils live for a shorter time than those of today? A study of

these survivors would supply more information about genetic and constitutional influences on health than the mortality of the total population. A murdered man has nothing to say on this score.

To test the hypothesis that the survivors ofyore did live as long as their counterparts of today, a study has been made of the lengths of life of ancient Greeks and Romans whose dates of birth and death are known to us either firmly or with a close degree of approximation. These have then been compared statistically with similar population samples from recent times.

[Klingan]

The link is broken, could you see if it's possible to fix it? It would be interesting to see what they base their numbers from. I've always used the book "As the Romans Did, second edition, A Sourcebook in Roman Social History" by J Shelton to answer this question, since it's (at least it seems so to me) highly reliable.

 

There it's stated that the average, as well as the median lifespan was 27 years.

[M. Porcius Cato]

Thanks for posting that.

 

I think the most interesting data from the study is that when you go either from "circa" to "firm" dates or from earlier to later dates, you get lower life spans. What this suggests is that there was a tendency to overestimate the ages of eminent men. Moving away from the estimates of our ancient sources to firm records, the most reliable data would be the ages of those born after 100 BC for whom we have definite birth dates. This group had a median age of 58 years compared to the modern median age (for comparable sample) of 78 years.

 

Thus, the life-spans of the Roman elite were shorter than that of the modern elite.

[ASCLEPIADES]

Salve , guys! About THIS LINK; if you can't follow it, you can paste directly the URL:

 

http://www.pubmedcentral.nih.gov/picrender...mp;blobtype=pdf

 

Anyway , here is the methodology section (how they did it):

The ancient population was obtained from the Oxford Classical Dictionary4 and consisted initially of every male entered in the Dictionary who has been accorded either firm dates of birth and death or circa dates. When a date was quoted as a range (usually consecutive years), the mean was calculated. The total number of individuals in the resulting sample was

397. Of these, 99 met violent ends by assassination, forced suicide or death in battle. They were rejected from the series, leaving a final total of 298. Firm dates had been accorded to 70 of these; the remaining 228 were in the circa group. Females were not included on the grounds that the sexes might differ in life expectancy. Modern male populations for comparison with the ancient group were obtained from Chambers Biographical Dictionary5. Thus, both ancient and

modern samples were comparable with respect to males who survived into adulthood and who achieved notability (or notoriety). Three modern groups were examined. One group consisted of males who died in or after 1950 AD. The two other groups died between 1850-1899 and between 1900-1949, respectively. In each group, every third appropriate entry was taken in order to eliminate possible familial influences, i.e.if father and son had both been included. The groups

were compared statistically by means of the nonparametric Mann-Whitney U-test6.

TO BE CONTINUED ...

[M. Porcius Cato]

The groups were compared statistically by means of the nonparametric Mann-Whitney U-test.

 

This made absolutely no sense to me. Ages are a continuous variable, therefore using a nonparametric test is inappropriate.

[ASCLEPIADES]

Salve, guys! If still can't link to the article, you can always try in the site of the British Medical Joural (www.bmj.org) or, even better, read it in paperware at your favoritelibrary. Quoting MPC: "Instead of doing a search on Google, try doing a search on your library's catalogue of BOOKS. They deliver text at amazing speeds without even needing an internet connection, feature high quality images, and--best of all--contain scholarly references and are edited for accuracy. All operating systems work with BOOKS, and thanks to public libraries, they're even open source."

Now, the Results section of this article:

 

The figures for length of life of the ancient population were first analysed in isolation. The 298 observations were grouped in consecutive 5-year periods, when the distribution showed a negative skew with a mode at 75-80 years. The median was 70 years; the range was 19-107. To test the homogeneity ofthe population the 298 members were arranged in chronological order according to year of birth, ranging from 650 BC to 602 AD. When their lengths of life were grouped according to their century of birth, a drop in length of life after the second century BC was strikingly apparent. A lower level was maintained in succeeding centuries. The subgroups with firm dates and with circa dates were therefore examined separately, when both subgroups were seen to show the group trend with drops in the median length of life after 100 BC (Table 1). The decreases were statistically significant in both cases.

 

Table 1. Median length of life in ancient populations with 'circa' and 'firm' dates born before and after 100 Bc. The figures in parentheses denote the numbers in the samples. The P values are for two-tailed significance limits

----------------------------------------------------------------------------------------------------------------------------------------

..............................................'Circa' dates..........................................................'Firm' dates

----------------------------------------------------------------------------------------------------------------------------------------

Born before 100 d.c.................. 72 years (128)........<--z=0.15 NS--->....................71.5 years (30)

.

..............................................z = 2.31 p <0.05.................................................z = 3.03 p <0.005

 

Born after 100 d.c.....................66 years (100).........<--z=2.37 P< 0.02--->..............58 years (40)

-----------------------------------------------------------------------------------------------------------------------------------------

NS=Not significant

TO BE CONTINUED ...

[ASCLEPIADES]

The groups were compared statistically by means of the nonparametric Mann-Whitney U-test.

 

This made absolutely no sense to me. Ages are a continuous variable, therefore using a nonparametric test is inappropriate.

Good question. The independent variable (length of life) is indeed continuous, BUT the dependant variables for each analysis (1.- born before or after 100 BC; 2.- firmness of the dates of birth and death) are not; they are categorical. As far as I now this test is an acceptable alternative under this circumstances.

[M. Porcius Cato]

The groups were compared statistically by means of the nonparametric Mann-Whitney U-test.

This made absolutely no sense to me. Ages are a continuous variable, therefore using a nonparametric test is inappropriate.

Good question. The independent variable (length of life) is indeed continuous, BUT the dependant variables for each analysis (1.- born before or after 100 BC; 2.- firmness of the dates of birth and death) are not; they are categorical. As far as I now this test is an acceptable alternative under this circumstances.

 

You have it backwards. The DV is the thing being measured--in this case, lifespan. The IV is the thing being manipulated (or, acting as a predictor of the DV)--in this case, category of date (before/after 100 BC) and "firmness" of dating. When you have an continuous DV (which lifespan most certainly is) and a categorical IV (which is what they're using), choosing a non-parametric test is typically inappropriate (especially for small samples) because it's likely to make overly conservative errors (i.e., fail to find differences where they really exist). Given that a few assumptions are met (which they don't report on one way or the other), they should have instead used some variant of a general linear model (probably a mixed model would be best).

 

BTW, since their chief claim is that the samples don't differ, their choosing a statistical test that is biased to fail to find real differences is pretty sleazy. In fact, the only reason that this was news was the surprise value of their "finding" that the lifespans of the ancient Romans (at least the elite ones) was no different from the lifespans of the modern ones (at least the elite ones). Tsk tsk tsk.

[ASCLEPIADES]

The groups were compared statistically by means of the nonparametric Mann-Whitney U-test.

This made absolutely no sense to me. Ages are a continuous variable, therefore using a nonparametric test is inappropriate.

Good question. The independent variable (length of life) is indeed continuous, BUT the dependant variables for each analysis (1.- born before or after 100 BC; 2.- firmness of the dates of birth and death) are not; they are categorical. As far as I now this test is an acceptable alternative under this circumstances.

 

You have it backwards. The DV is the thing being measured--in this case, lifespan. The IV is the thing being manipulated (or, acting as a predictor of the DV)--in this case, category of date (before/after 100 BC) and "firmness" of dating. When you have an continuous DV (which lifespan most certainly is) and a categorical IV (which is what they're using), choosing a non-parametric test is typically inappropriate (especially for small samples) because it's likely to make overly conservative errors (i.e., fail to find differences where they really exist). Given that a few assumptions are met (which they don't report on one way or the other), they should have instead used some variant of a general linear model (probably a mixed model would be best).

 

BTW, since their chief claim is that the samples don't differ, their choosing a statistical test that is biased to fail to find real differences is pretty sleazy. In fact, the only reason that this was news was the surprise value of their "finding" that the lifespans of the ancient Romans (at least the elite ones) was no different from the lifespans of the modern ones (at least the elite ones). Tsk tsk tsk.

Salve, MPC. Of course, you're right; I have been corrected and I think you have a valid point Gratiam habeo!

Edited July 19, 2007 by ASCLEPIADES