Reviews:Saller's Patriarchy, Property and Death

From William A. Percy
Jump to: navigation, search

Patriarchy, Property and Death

(Cambridge, 1996)
by Richard P. Saller

Beginning in 1984 in a series of articles, Richard P. Saller, joined a bit later by Brent D. Shaw, mistakenly contended that the ancient Romans married late – males on average at age twenty-eight, females at nineteen. They cited hundreds of Latin epitaphs from the late Republic into the Christian empire which proved beyond reasonable doubt that fathers normally commemorated their sons who died before age twenty-eight and their daughters who died before age nineteen; after those ages spouses became the predominant commemorators of each other. From this solid base they erroneously concluded that those were the average ages of marriage. Although otherwise erudite, much of Saller’s Patriarchy, Property and Death is based on these miscalculations. His conclusions were deduced from a faulty axiom. That is, he claimed that the patria potestas was really rather insignificant because most Roman males married for the first time after the deaths of their fathers. In fact very, very few of the specific references collected to date corroborate such a late age for first marriage. Regardless, Saller’s text has been praised to the skies in numerous reviews and gained a widespread acceptance, most notably in Debating Roman Demography (Brill, 2001), authored in part and edited by Walter Scheidel, the dean of Roman demography, and containing a long article by Shaw. No one has yet seriously criticized this gross mistake.

Contrary to the theories of Saller and Shaw, the ancient Romans from Latin speaking provinces, about whom we are informed from manuscripts and literary sources – mostly middle to upper class – married early. As long as he was able, a father always commemorated his son, and until she as a wife produced a living heir, at which point her husband could keep the dowry, a father would also commemorate his married but childless daughter. By age nineteen most wives had a surviving child, so that her husband could keep the dowry and therefore commemorate her.

Earlier demographic studies had contended that ages for first marriage were young. With his magisterial command of Roman customs, Ludwig Friedlander in Darstellungen aus der Sittengeschichte Roms in der Zeit von Augustus bis zum Ausgang der Antonine (in 10 editions between 1864 and 1922) was satisfied to use merely eight female and about twenty-five male examples all drawn from manuscripts. Other German classicists of the 19th and early 20th century had also shown that Roman males in their late teens married girls about fourteen or fifteen. However, as Victorian efforts to protect children from sex and labor increased in the Anglophone world, in particular during the fin-de-siecle and Edwardian ages, a presumably horrified American classicist, A.G. Harkness in 1896, began the attempt to raise the actual ages at which Romans first married by using inaccurate and misleading mathematical averages from the growing corpus of epigraphic evidence then being published.

A sophisticated don from the University of Cambridge, M.K. Hopkins, in “The Age of Roman Girls at Marriage” (Population Studies 18, 1965) repudiated Harkness’ claim, saying that his numbers were too high because of the gross misuse of mathematical averages. Introducing a modal analysis, Hopkins more clearly indicated the significant clusters of lower age ranges and confirmed Friedlander’s work. He added only two more examples of female marriage ages from literary evidence and attempted to settle the issue for all times by amassing more epigraphic evidence for males than Harkness had used. He cited eighty-six pagan and ninety Christian men, believing that he had closed the case. This of course made the Romans guilty of what has recently been dubbed “heterosexual pedophilia.” Strengthening the conclusions of Friedlander and Hopkins, we in The Age of Marriage in Ancient Rome (The Edwin Mellen Press, 2003) have added twenty-one examples, for a total of thirty-one females and seventy examples for males, more than half of which are new, all taken from manuscripts. (See Tables below)

The statistical odds of having but a single one of the definitive first marriage ages of males or females even equaling, much less exceeding, those postulated by Saller and Shaw are so great that my mathematical skills cannot calculate the almost infinitesimal figures. Yet amazingly they have been overwhelmingly accepted and hardly challenged by scholars. How the authorities in classics, demography, and epigraphy, not to mention history and the rest of the social sciences, could have accepted Saller’s and Shaw’s theses blows the mind and raises questions about the credibility and objectivity of the experts.

In contrast to the Greeks, who opted for delayed marriages to keep down birth rates because of the lack of land and other resources to support their citizens adequately, the Romans with ample resources were preoccupied with keeping up the supply of soldiers and therefore encouraged early marriage for both males and females of the citizen class. Saller has gone the farthest astray, but other classicists all to often fail to recognize this fundamental difference in Greek and Roman customs. Among contemporaries very few, most notably John K. Evans (War, women and Children in Ancient Rome. London, 1991) and Tim Parkin (Demography and Roman Society. Baltimore, 1992), have as far as I know, taken serious issue with their errors. Saller’s and Shaw’s findings may indeed be “statistically significant,” but their conclusions demonstrate a complete and shocking lack of understanding of Roman customs as well as a disregard and a disrespect for their own scholarly predecessors. They disputed Hopkins’ data, because much of his is from freedmens’ tombstones. Of course the emancipated freedmen normally married later than free males and often chose older brides, but some did marry young girls.

Commonly and correctly accepted among historians is the notion that in pre-modern societies, mortality rates were very high, exceeding half by age six, and life expectancies low. Following this, one may logically assume a young marriage age – early teens for females and mid-to-late teens for males, for otherwise, populations would not have been able to sustain themselves, let alone expand. There were exceptions, of course, but in any case where early marriage age was not a common practice, as for example with the Spartans or the other ancient Greeks after 630 B.C., there is clearly documented evidence of it and an explanation as to why in my Pederasty and Pedagogy in Archaic Greece (Urbana and Chicago, 1996). There is no definitive evidence pointing to any similar Roman custom. What then motivates someone to play contrarian to giants of scholarship such as Friedlander and Hopkins, when statistical information is scant?

It is perhaps for the same reason that Harkness was willing to distort the data in 1896. By the1980’s, when Saller and Shaw began their analyses that culminated in Patriarchy, Property and Death, the Sexual Revolution of the 60’s and 70’s was succumbing to the holocaust of AIDS, and the conservative fundamentalist Protestants and born-agains along with others in the religious right were beginning to restrict sexual activity, particularly among the young: “Just say no.” Certain classicists began to whitewash their Roman idols from potential charges of child abuse. Saller and Shaw and their supporters seem to have fallen victim to that risk. Regrettably, through their attempts to overturn Friedlander’s and Hopkins’ venerable scholarship, they have contributed to the politically-driven revisionist re-interpretation of Classical history.

I do, however, wish to thank Saller and Shaw for their meticulous collection of the data from the epitaphs, without which I could never have deduced the average age at which Roman husbands who were commemorated lost their fathers (28) and at which wives who died and were commemorated, left living children for their surviving spouses (19). Unlike Shaw, who allowed that he had some evidence in shoe boxes in his mother’s garage that might support us, Saller, understandably busy with his administrative duties and seeking more academic honors, did not respond to our inquiries. Ironically their “statistically significant” data actually helped prove exactly the opposite of what they thought that it had.


> I. First Marriages of Roman Men:
From Appendix I, p. 103

<img src="images/table1.jpg" width="309" height="93">
*bold type indicates ante quem data

II. First Marriages of Roman Women:
From Appendix II, p.121

<img src="images/table1-b.jpg" width="222" height="83">


13, 14, 15,15,15,15, 16,16,16,16, 17,17,17,17,17,17,17, 18,18,18,18,18,18,18, 19,19,19,19,19,19,19, 20,20,20,20,20,20,20,20,20, 21,21,21,21,21,21,21, 22,22,22,22, 23,23,23,23, 24, 24, 25, 26,26, 27, 28, 29, 30, 31,31, 33, 35, 36, 41</font>

Mean: 20.6
Standard Deviation: 6.2
For a random sample of size 83, we may assume that the average is normally distributed with standard deviation
<img src="images/table2-a.jpg" width="79" height="48">

This sample, however, is a sample of convenience; it was obtained from historical records of famous men. We may assume though that it is close to a random sample, at least from the population of upper class men. We also have to assume that marriage customs remained steady during the centuries covered.

We take the null hypothesis to be that the average is 28, and the alternative that it is less. With the above assumptions, we can compute the P-value, that is, the probability that the mean in the sample turns out to be 21 or less if the population mean is 28, as:

<img src="images/table2-b.jpg" width="364" height="33">

(The latter value was computed with Maple.)
Thus, the null hypothesis must be rejected with practical certainty, unless the assumptions can be shown to be invalid.

-courtesy of Gaza Shea
Professor of Math
University of Massachusetts, Boston


11, 12,12,12,12, 13,13,13,13,13, 14,14,14,14,14,14, 15,15,15,15,15,15, 16,16, 17,17

Mean: 14.0
Standard deviation: 1.57

A random sample of size 26, is just barely large enough to assume that the average is normally distributed with standard deviation,
<img src="images/table3-a.jpg" width="79" height="44">

nevertheless, we first assume this, but then obtain another estimate without this assumption as well.

This sample, however, is a sample of convenience; it was obtained from historical records of famous women. We may assume though that it is close to a random sample, at least from the population of upper class women. We also assume that marriage customs remained steady during the centuries covered. (For this reason, we omitted three women for whom records were available from the Christian era.)

We take the null hypothesis to be that the average is 19, and the alternative that it is less. With the above assumptions, we can compute the P-value, that is, the probability that the mean in the sample turns out to be 14 or less if the population mean is 19, as:
<img src="images/table3-b.jpg" width="332" height="33">

(The latter was computed with Maple.)

Thus, the null hypothesis must be rejected with practical certainty, unless the assumptions can be shown to be invalid.

The ridiculously low number we obtained, depends heavily on the validity of the normal approximation, which is questionable. We can avoid it and compute an estimate for the P-value using Chebyshev’s inequality instead, which is valid for any distribution, as:
<img src="images/table3-c.jpg" width="298" height="33">

This estimate, though very crude (in the sense that the true P-value is probably much lower), is much more reliable than the one above, and it is still sufficiently small to enable us to conclude that the null hypothesis, of an average age 19 at first marriage is untenable.</font>

- courtesy of Gaza Shea
Professor of Math
University of Massachusetts, Boston


Statisics, Statistics, and More Lies…

It has been brought to my attention by my former student and co-writer Arnold Lelisthat the analyses I requested from Gaza Shea are, because of my insistence, a slight bit of overkill, and that perhaps my numbers are not totally accurate:

“Table two: the math should be recalculated for 50 – the sample size when the ante-quem dates are removed (there are twenty of these). I have explained how and why the ante-quem data is inherently non-random and slanted specifically to support our position.

“Table three: Here, the sample for the math should throw out the ante-quem data but reinstate Antonia Minor (20), Julia (24), and Galla Placidia (26). This leaves a sample size of 25. The ante-quem data should be removed for the same reason as for Table Two.
As for the three old maids:

“Antonia Minor is extremely well attested in the sources, and there is absolutely no doubt that she married Drusus at age 20 (he was twenty-two), and that this was her first marriage.</font>

“Julia’s marriage also is explicitly attested in the sources. Some have questioned her year of birth on the evidence of her marriage; but no definitive conclusions have been generally accepted in this regard.

“Finally, Galla Placidia’s AAFM of 26 must be retained. This figure represents the highest possible value based on two uncertainties: the marriage to Ataulf occurred NO LATER than January 414, and she was born no earlier than 388. (I suppose this could be considered an ante-quem, therefore. However, the January 414 date appears in some sources, so could be correct.) In any case, at the time she was taken hostage by the Goths, she could have been as young as sixteen – not at all impossible, if the court in Ravenna was undecided how to use her.

“I think it is important to include data that we don’t “like” from time to time; that enhances credibility.”

While it may be true that I have left out some older examples, which might make the data lean more heavily towards older first marriage ages; nevertheless, they would not hinder me from proving that Roman people married at young ages. In addition, the statistics I have presented are merely included as bolsters to understanding what should be a simple exercise in logic, which would not require the burdensome, nit-picking acquisition and analysis of statistics, if current scholarship had not forced us to indulge in it. Just by understanding the nature of pre-modern societies, by acknowledging that mortality rates were high and life expectancies were low, it is not the convoluted math of statistics, but rather the simple laws of arithmetic which make it clear to anyone that unless Romans had a specific and documented custom of delayed marriage – as among pagan Greeks after 630 BC – there can be no question that they married early; for otherwise they would not have flourished into the populous and phenomenal empire that we know of today. I hope that this little essay will provoke competent and genuine accredited classicists, demographers and statisticians to correct the faults of this appendix.
- Posted 10-29-05

Personal tools