Bryn Mawr Classical Review 97.8.17

Walter Scheidel, Measuring Sex, Age and Death in the Roman Empire: Explorations in Ancient Demography. Journal of Roman Archaeology, Supplementary Series 21. Ann Arbor: Journal of Roman Archaeology, 1996. Pp. 184. $59.50. ISBN 1-887829-21-0.

Reviewed by Roger S. Bagnall, Classics, Columbia University, New York, NY 10027,

Four studies treat "the biology of brother-sister marriage in Roman Egypt," "digit preference in age records from Roman Egypt," "the demography of the Roman imperial army," and "seasonal mortality in the Roman empire." These sections are largely independent of one another (the absence of an index to the volume speaks loudly) but share a highly quantitative character and strong interest in the natural, particularly the biological, sciences. Indeed, the need to bring scientific work to bear on history is a pervasive theme of the volume. To give away the ending, Scheidel concludes "that, as far as quantitative analysis is concerned, the end of ancient demography is already in sight" (165). This view rests on acceptance of the "emerging consensus" about the main lines of the demographic structure of the ancient world, together with a sense (greatly sharpened by this book) of the limits of the evidence that would allow that consensus to be sharpened in detail. Despite this elegiac tone, however, the book is devoted to precisely this kind of attempt to explore the quantitative evidence for various subjects in much greater detail and to bring out the complexities easily passed over in synthetic treatments.

The longest of the chapters is on close-kin marriage; it is to some extent a preliminary work toward a book on this subject Scheidel is preparing. Three recent articles of Scheidel's published elsewhere also concern various aspects of the subject.1 It has long been known, and much discussed, that the documents of Roman Egypt provide confirmation of Philo's claim that the Egyptian population of his day practiced officially -- and socially -- sanctioned incestuous marriage, in the form of marriage between full or half brothers and sisters. Although the data mainly come from a century or more after Philo, this is probably just the product of the uneven survival of the census declarations, the largest part of the evidence. Mostly these documents come from the Arsinoite nome, and only from there do we have enough evidence to sustain statistical inference; but again, the evidence is thin enough elsewhere (once more, a matter of survival of documents) that it would be unsafe to suppose that the practice was limited to the Arsinoite. The discussion here is based on the data and analysis in Bagnall and Frier 1994,2 but it sets the question in a wider context and poses a different set of questions.

The essence of Scheidel's argument is that, even taking the lowest estimate of the frequency of brother-sister marriage allowed by the evidence of the census declarations and making the most conservative of assumptions about the presence of deleterious recessive genes in the population, sibling marriage at the level indicated should have had devastating effects on the biological fitness of that population. In part these effects would be increased mortality, as lethal recessive genes would become dominant more often as a result of inbreeding; in part they would reduce reproductive success in sublethal but powerful fashion, for example through the births of persons with significant disabilities not visible at birth, who would be less likely to reproduce. Scheidel carefully considers (30 ff.) the possibility that deleterious genes might be eliminated by inbreeding, but he shows that this cannot have happened. He concludes that one must reject two of the possible three conclusions as untenable; these are that the effects of inbreeding were alleviated by adoption and by extramarital conceptions (the first is hardly attested, the second unlikely to have been so common as to have much effect), or that the effects were not as serious as his models predict. As a result, he thinks that the effects of inbreeding must have been considerable "but quietly endured."

Without any expertise in genetics, I cannot claim to evaluate the argument here as it deserves. But two potential weaknesses should be signalled for further discussion. First, many of the potential non-lethal defects would either disable the individual at a time when much of the main reproductive period was already past or would, however unsightly, leave the person still able to reproduce. The very small numbers of individuals involved in the modern studies cited (20-21) do not provide much basis for assessing the number of non-lethals needed to equal a lethal. Second, the calculation of the effects of inbreeding (24-25) is based on pure extrapolation from studies based on first cousins, and no evidence is offered that a straightforward multiplication by four of the excess mortality with first-cousin unions will actually arrive at the excess mortality in sibling unions. Nor is the argument concerning the relationship of such excess mortality to a high existing infant mortality level particularly cogent.

The great problem, however, is that it is impossible to discern any such negative effects, any signs that sibling couples really were at a disadvantage. What would the signs be? Scheidel suggests that having to produce extra children to make up for extra mortality and failure to reproduce is one; but since failure to reproduce might be an effect, are we to look for families with few children (failure to reproduce) or with many (compensation)? The relatively small group we can study (after all, only twenty-four couples) does not give much evidence for either (cf. 15, 26-27).3 Another possibility would be the presence of many disabled adults. For this the census declarations might be helpful (tax exemption as a result of disability is known), but provide no evidence. It must be kept in mind, however, that the sample is extremely small (and not necessarily representative), and few statistically-derived conclusions concerning it deserve much confidence.

The inability to test Scheidel's hypothesis against documentary evidence for the supposed consequences is unfortunate, but it is not itself compelling evidence that he is wrong. Scheidel does not himself offer any explanation for the presence of acceptable incestuous marriage in Roman Egypt, contenting himself with decrying those offered so far and suggesting that those offering such explanations must take more seriously than they have so far the real genetic damage that inbreeding must have caused. The more important that damage, the more powerful the causes of the introduction of sibling marriage must be. At the same time, he insists that the case of Roman Egypt is so limited and so unsuccessful (the practice lasted less than 300 years, it seems)4 that it cannot be used as a counterargument against the view that the incest taboo is a fundamental aspect of human society. To be sure, there is little evidence for close-kin marriage after the Antonine Constitution made everyone subject to Roman law; if the practice ended mainly because Roman law prohibited it for the entire population after 212, it can hardly be argued that it was an evolutionary dead end. What is most striking to me in the conclusion to the article is the sense that, for all the force of the argument that historians must take the realities of genetics seriously, we are left with a situation in which genetics offers us a hypothesis, the documents do not test it, and we have no other way of connecting the two. The same thing is true for Scheidel's lengthy argument (38 ff.) that siblings (at least if brought up together as children) should in fact not have desired one another as mates, followed by his sober assessment that "we may not know anything about the emotional state of brother-sister couples" in the papyri (46).

A bleak assessment might be that we have come a long way to arrive at no conclusions and only at hypotheses that cannot be tested with any available evidence. A more cheerful view might be that the warning that sibling marriage is unlikely to have been good for anyone's health, sex life, or reproductive success is in itself a significant gain, even if we cannot foresee how that warning can be connected with our documents. But the real conclusion seems to me to be that there is something wrong with this picture. It is hard to believe that the metropolis of Arsinoe could have practiced incestuous marriage for 250 years on a large scale without our having some reverberations visible in the papyri. But to repaint the picture is not the task for a review; and it would help to know what it should look like!5

The second chapter looks in great detail at the degree to which the census declarations show either preference or avoidance of particular end-digits in ages. Scheidel provides a context for this examination by looking also at other bodies of data from Roman Egypt: gravestones, mummy labels, nominations to liturgy, tax lists, and contracts. After a detailed analysis of the various statistical tools for measuring digit preference, he finds the Myers' Blended Method the most reliable, with various adjustments in some cases.6 Using it leads to the conclusion that age-rounding (mainly preference for multiples of ten, much less for multiples of five) is rampant in those sources which represent only someone's guess at another person's age (contracts, nominations to liturgy). Census declarations and tax records, carefully gathered and officially controlled, are not surprisingly the least distorted by preferences.7 Gravestones and mummy labels are close enough in behavior to be lumped together for most purposes; they occupy an intermediate position.

All of this leads to a serious nuancing of the notion that age-awareness and precision are a reflection of levels of literacy. Scheidel demonstrates conclusively that one must take account of the nature of the documentation being used and above all who gave the age and for what purpose, before drawing conclusions about the individual whose age is given. It remains true, however, that within the most accurate group of documents there are significant differences between the precision of ages of men and of women, and of city and village dwellers (see esp. 65). (Interestingly, there is no such strong correlation in tombstones: 74.) It is curious that in this connection (66) Scheidel does not explore any possible differences between data contained in declarations submitted by men and those submitted by women, even though it is in most cases easy to distinguish them (the information is given explicitly in complete declarations and is often elicitable in incomplete ones). Such an investigation might help clarify the difficult question of whose knowledge of ages is at stake in the declarations.

There are other interesting conclusions as well, perhaps most intriguing of all that there is a concerted avoidance of the number seven, found almost everywhere except in the male sub-samples of the census data and most widespread among village women. Scheidel argues that this avoidance of 7 was a matter of superstition. He also tries to discern "some general preference for 9 as a powerful figure made up of three times three" (86), but he seems unaware of the literature concerning the avoidance of theta in the papyri as unlucky.8 He recognizes clearly, however, the difficulty in untangling the preference for 9 from the preference for 0, and he sees that preference for 9 is not easy to explain (68). On a more prosaic (but comforting) level, Scheidel argues (70) that if one takes account of digit preference in women's ages, the data of the census declarations support the model life table even more strongly than Frier and I supposed.

Chapter 3 looks at the recruitment requirements of the imperial Roman army and the impact that these may have had on the population of the empire, particularly the Roman citizen population. Much of it is devoted to the question of age at recruitment, a subject approached mainly by subtracting years of service from age at death, as recorded on soldiers' tombstones. Scheidel shows conclusively that the high incidence of age-rounding on these tombstones corrupts the calculation; when the rounded ages are eliminated, the age-range of enlistment shrinks dramatically, with the vast majority joining at 18-20 years. Scheidel then turns to ask how many soldiers made it to retirement; he argues that to natural mortality was added a considerable loss to early discharge (mainly from disability) as well as to desertion and other causes. The concluding part of the chapter considers the potential impact on the imperial budget of changes in the length of service, suggesting that Tiberius' accumulated wealth may have come in part from reducing military expenditures generally (especially not replacing Varus' legions), but in part from reducing the number of veterans entitled to retirement bonuses by extending their term of service.

The final chapter looks at the seasonality of mortality in Rome, Egypt, and North Africa. Bringing out the much higher Roman death rate during the months from July to October, Scheidel is reluctant to offer a single explanation such as malaria, arguing that other fevers, notably those favored by poor sanitation, may have interacted with malaria-produced weakness to provide multiple causes of death. Egypt's curve looks similar to Rome's, but is in advance of it by two months, perhaps the result of the inundation schedule. Carthage, by contrast, shows virtually no seasonal variation in mortality. It is hard to see if this represents lower mortality in general or only a different distribution of it. In any event, the regional variations point to the diversity of some of the contributing elements of the ancient demographic regime.9

This book is not easy reading even for the quantitatively minded, in part because the elements of tables are not always as well labeled as one would wish.10 At times the results of additional stages of analysis do not seem quite commensurate with the effort. But that sense may itself be part of Scheidel's program of showing just how many booby traps there are in the data, how many documentary biases and cultural factors waiting to be uncovered, and in the end the more painful parts of the analysis contribute to our sense of the difficulty and complexity of the enterprise of getting beyond generalities in talking quantitatively about the ancient world.

One other latent but disturbing problem deserves mention in conclusion. Scheidel's concluding "afterword" speaks pessimistically, after the sentence quoted in the opening of this review, about the lack of more quantitative evidence to explore. But quantitative data for the ancient world do not lie around waiting; they are constructed, by many hours of patient work editing texts, assembling references, studying skeletons, and the like. Someone has to recognize the evidence as quantifiable and gather it up, checking it and recording standard information about each item. Scheidel has almost entirely avoided this work by using the compilations of others. Even where creating new data-sets would be possible, as for example with the Terenouthis gravestones (many of which have been published since Boyaval's work of the mid-1970s), he has not tried to do so. Yet larger numbers would presumably have eliminated some sampling flukes and given a greater statistical reliabiity to the results, which (it must be emphasized) often rely on very few documents. And even with the data-sets he uses, he has been reluctant to ask questions that would involve going back and looking at the texts again.11 It is one thing to complain that there are no bricks, another to claim there is no straw. It is not clear which we are dealing with.

All the same, it will be clear that there is in these pages much to stimulate the reader, some salutary cautions, and some solid conclusions. For that, and for the care with which he has tested the reliability of the numbers he has generated, Scheidel deserves our warmest thanks.12


1. "Incest Revisited: Three Notes on the Demography of Sibling Marriage in Roman Egypt," BASP 32 (1995) 143-55 (1: more detailed study of the numbers from the Arsinoite nome, pointing to an extremely high rate of sibling marriage; 2: estimates of consanguinity in the population of Arsinoe; 3: an argument that the published evidence is inadequate to support any hypothesis about the geographical distribution of brother-sister marriage in Roman Egypt); "Brother-Sister and Parent-Child Marriage Outside Royal Families in Ancient Egypt and Iran: A Challenge to the Sociobiological View of Incest Avoidance?" Ethology and Sociobiology 17 (1996) 319-40 (largely a summary of the book chapter, directed at a different audience, but with more discussion of the Iranian evidence); "Brother-Sister Marriage in Roman Egypt," Journal of Biosocial Science 29 (1997) 361-71 (more detail on rates of inbreeding and of inbreeding depression).

2. Roger S. Bagnall and Bruce W. Frier, The Demography of Roman Egypt (Cambridge 1994).

3. It should be borne in mind not only that the census declarations list only surviving children (a point Scheidel is well aware of), but also that they list only those surviving children who still lived in the parental household. Those who had married out are not included in their parents' declaration. Total fertility of the older parents is thus underrepresented by the declarations. Moreover, as siblings married to each other were probably less likely to move out of the parental household, the declarations are likely to overstate the percentage of children of brother-sister marriages who themselves married siblings.

4. Although Scheidel rather oddly says (50) that the phenomenon lasted "over a two-digit number of generations," it is hardly more than a 240-year phenomenon, and much less in the census declarations that Scheidel is discussing: hardly more than five to ten generations.

5. As Scheidel notes in BASP 32 (1995) 154 n. 29, a substantial body of data probably from Ptolemais in Upper Egypt is shortly to appear: R. S. Bagnall, B. W. Frier, and I. C. Rutherford, The Census Register P.Oxy. 984 (Papyrologica Bruxellensia, Brussels 1997). This text at least advances our sense of the geographical and chronological boundaries of the phenomenon by showing only one possible case of sibling marriage among 24 conjugal families recorded in a register probably concerning the city of Ptolemais in the census of 89/90. It thus reinforces the view Frier and I took in 1994 that sibling marriage was more characteristic of lower and middle Egypt than of upper Egypt, and perhaps also that the phenomenon was less common in the first century than the second.

6. One important typographical error in this discussion must be signalled. On 54, line 14, read "Dividing the number of ages ending in 0 (41) by one-tenth of the total number of ages" etc. But even so this is an inaccurate description of Whipple's Index, which uses all ages divisible by 5, not only those divisible by 10. His criticism of Whipple's Index is thus also misinformed; the use of the age range 23-62 neutralizes the supposed distortion he describes.

7. Scheidel has explored numeric bias in the census returns in more detail in another article, "What's in an Age? A Comparative View of Bias in the Census Returns of Roman Egypt," BASP 33 (1996) 25-59. The first section explores digit preference in the census declarations (again concluding that a relationship between accuracy of age statements and literacy cannot be shown); the second discusses the sex ratio in the villages and the metropoleis (but unfortunately not distinguishing slave and free; on this see my remarks in Scripta Classica Israelitica 16 [1997]); the third argues that the purported incidence of twins in the declarations is impossible to reconcile with the facts of biology, concluding that many supposed twins are "false twins" and express a cultural preference for twins, perhaps related to some Egyptian cults.

8. See Armin U. Stylow and J. David Thomas, "Zur Vermeidung von Theta in Datierungen," Chiron 10 (1980) 537-51; their conclusion (551) is that "Welche Motive nun tatsächlich für diese Praxis ausschlaggebend waren letzlich stand hinter allen Formen der Vermeidung von Theta der Glaube an die magische Wirksamkeit seiner Verbindung mit dem Kaisernamen."

9. Many readers will have seen the study of the same subject by Brent Shaw, "Seasons of Death: Aspects of Mortality in Ancient Rome," JRS 86 (1996) 100-138. Shaw has assembled a substantially larger Roman data set (3,943 inscriptions vs. three samples used by Scheidel, the largest 2,125); he also considers Egypt and other areas. A detailed comparison of Scheidel and Shaw on this subject would be instructive but cannot be undertaken here.

10. To give just one example: Table 3.20 (p. 127) deals with expected and implied rates of manpower loss after 20 years of service. The columns are headed "expected mortality," "expected total loss," and "implied loss." But the numbers in these columns are actually the numbers of men surviving in the years in question, not the numbers lost.

11. A striking example is the avoidance of a direct treatment of the relationship between seasonal mortality and age at death, where Scheidel shows (143-44) that child mortality could not account in itself for the visible pattern not by the "laborious operation" of gathering the information but by a more theoretical argument. The argument itself seems sound, but it leaves us missing an interesting exploration that is possible, but time-consuming. Shaw (above, n. 9, 118-19) has compiled and analyzed the data for infants and children in the Roman inscriptions. Even in one case where the data were already assembled, as in the texts listed in the addendum to the Demography of Roman Egypt, Scheidel has not included them in his figures.

12. I am indebted to my daughter, Martha Bagnall, for discussion of the questions of genetics raised in chapter 1.