Hostname: page-component-78c5997874-lj6df Total loading time: 0 Render date: 2024-11-08T09:27:53.851Z Has data issue: false hasContentIssue false
  • English
  • Français

Emperors, aristocrats, and the grim reaper: towards a demographic profile of the Roman élite*

Published online by Cambridge University Press:  06 November 2012

Walter Scheidel
Walter Scheidel
Affiliation:
Darwin College, Cambridge
Get access Rights & Permissions [Opens in a new window]

Extract

The opening pages of the annals of the Roman monarchy tell of long-lived rulers and thriving families. Augustus lived to the ripe age of seventy-six, survived by his wife of fifty-one years, Livia, who died at eighty-six, while her son Tiberius bettered his predecessor's record by two more years. Augustus’ sister Octavia gave birth to five children, all of whom lived long enough to get married; Agrippa left at least half a dozen children, and perhaps more; Germanicus, despite his tender age at death, was survived by no fewer than three sons and three daughters. At the same time, longevity and abundant offspring went hand in hand with early death and misery. More than a few luminaries of Augustus’ court were less fortunate than their elders: Marcellus died at twenty-three, the elder Drusus at twenty-nine, Augustus’ heirs Gaius and Lucius Caesar at twenty-three and eighteen. Drusus had lost a number of children save three that survived; Germanicus buried three of his six sons as infants.

Type
Research Article
Information
The Classical Quarterly , Volume 49 , Issue 1 , May 1999 , pp. 254 - 281
Copyright
Copyright © The Classical Association 1999

Access options

Get access to the full version of this content by using one of the access options below. (Log in options will check for institutional or personal access. Content may require purchase if you do not have access.)

Footnotes

*

All biographical data on emperors have been drawn from two recent standard works of reference, D. Kienast, Römische Kaisertabelle: Grundzüge einer römischen Kaiserchronologie (2nd edn, Darmstadt, 1996) (for the period up to A.D. 395), and the three volumes of J. R. Martindale, The Prosopography of the Later Roman Empire, II–III B (Cambridge, 1980–92). Graham Burton, Richard Duncan-Jones, Bruce Frier, Peter Garnsey, Keith Hopkins and a referee kindly offered comments on earlier versions of this paper.

References

1 Thus Suet. Claud. 1.6: ‘ex Antonia minore complures quidem liberos tulit, verum tres omnium reliquit’.

2 Cf. e.g. Syme, R., Roman Papers IV, ed. Birley, A. R. (Oxford, 1988), pp. 418–29Google Scholar (‘Neglected children on the Ara Pacis’), who draws attention to two children of L. Domitius Ahenobarbus (cos. 16 B.c.) and Antonia who must have died as minors.

3 Coale, A. J. and Demeny, P. (with Vaughan, B.), Regional Model Life Tables and Stable Populations (2nd edn, New York and London, 1983), pp. 43–4Google ScholarPubMed. On the uses of model life tables in ancient history, see e.g. Parkin, T. G., Demography and Roman Society (Baltimore and London, 1992), pp. 67–90Google Scholar. Newell, C., Methods and Models in Demography (New York, 1988)Google Scholar offers a lucid general introduction. Model life tables describe the age structure and various demographic properties of ‘ideal’ populations that are characterized by different but constant age-specific rates of mortality. Such models are based on large amounts of empirical data; in the case of the Coale-Demeny life tables, information derived from 652 data sets from actual populations was used to construct four sets (labelled ‘North’, ‘South’, ‘West’, and ‘East’; ‘West’ is regarded as an average pattern) of twenty-five tables for each sex. The twenty-five tables correspond to twenty-five different mortality levels that reflect a wide range of values for mean life expectancy at birth from around twenty (Level 1) to eighty years (Level 25). A term such as ‘Model West Level 4 Males’ is made up of three coordinates (regional set, mortality levels/life expectancy, and sex) that locate a given life table within the grand total of 200 ‘stationary’ populations (where ‘stationary' refers to a population with a constant size and a constant age structure).

4 While Level 3 (with a mean life expectancy at birth [= e 0] of 25 years for women and 22.9 for men) has been considered representative of the Roman population at large, Level 6 (with rates of 32.5 and 30.1 years) is sometimes applied to the upper classes. Frier, B. W., ‘Roman life expectancy: Ulpian's evidence’, HSCPh 86 (1982), 213–51Google ScholarPubMed, discusses Roman legal evidence suggestive of Mortality Level 2 (but cf. Parkin [n. 3], pp. 27–41); according to Bagnall, R. S. and Frier, B. W., The Demography of Roman Egypt (Cambridge, 1994), pp. 75–110CrossRefGoogle Scholar, the census returns of Roman Egypt are consistent with Level 2 for women and Level 4 (?) for men (but cf. R. Sallares and W. Scheidel, Disease and the Demography of the Roman World, forthcoming); Frier, B. W., ‘The demography of the early Roman Empire’, Cambridge Ancient History, 2nd edn, vol. XIGoogle Scholar (forthcoming), posits Level 3 as a rough average for the empire. A life expectancy at birth of about 30–32.5 years for Roman senators was first canvassed by Hopkins, K., Death and Renewal (Cambridge, 1983), pp. 147–8CrossRefGoogle Scholar. Saller, R. P., Patriarchy Property and Death in the Roman Family (Cambridge, 1994), p. 45CrossRefGoogle Scholar, considers Level 6 ‘the probable upper limit of life expectancy at birth’. Duncan-Jones, R., Structure and Scale in the Roman Economy (Cambridge, 1990), pp. 93–6CrossRefGoogle Scholar, argues that the album CIL IX 338 is consistent with Model South Level 6 (or 4) for the decurions of Canusium (but see below, third section). In general, there is no obvious reason why the wealthy should have outlived the poor: see the final paragraph of this paper.

5 Moving averages reduce chance irregularity in small sets of data; they are centred on a given target digit or bracket and assign the mean of the values of this digit or bracket and of a number of adjacent ones (two, four, six, …) to the target digit or bracket.

6 Age 20–29: Constantia, daughter of Constantius II and wife of Galerius, aged c. 21; Julia Drusilla, sister of Caligula, aged 21–23; Flavia Julia, daughter of Titus, aged c. 28. Age 30–39: Eudocia, daughter of Valentinian III, aged 33; Constantina, daughter of Constantine I, aged c. 34; Faustina the Elder, wife of Antoninus Pius, aged c. 35. Age 40–49: Licinia Eudoxia, wife of Valentinian III, aged c. 40; Faustina the Younger, wife of Marcus Aurelius, aged c. 46. Age 50–69: Vibia Sabina, wife of Hadrian, aged c. 51; Pulcheria, daughter of Arcadius, aged 54; Galla Placidia, daughter of Theodosius I, aged 57 or 58; Ulpia Marciana, sister of Trajan, aged 50–68; Pompeia Plotina, wife of Trajan, aged over 53; Matidia, niece of Trajan, aged over 51. Age 70+: Helena, mother of Constantine I, aged a 80; Livia, wife of Augustus, aged 86; Domitia Longina, wife of Domitian, aged over 70.

7 Hopkins (n. 4), pp. 146–9; accepted by Talbert, R. J. A., The Senate of Imperial Rome (Princeton, 1984), p. 133Google Scholar.

8 Talbert (n. 7), pp. 131–4, esp. p. 132. It deserves notice that in reorganizing the senate, Sulla seems to have believed that an annual intake of twenty thirty-year old quaestors would create a body of 600 senators.

9 Talbert (n. 7), p. 134, cf. pp. 15–16.

10 Ibid., p. 18.

11 Saller (n. 4), pp. 45–6,48–65.

12 Duties and age of patricians: Alföldy, G., Konsulat und Senatorenstand unter den Antoninen: Prosopographische Untersuchungen zur senatorischen Führungsschicht (Bonn, 1977), p. 36Google Scholar; Talbert (n. 7), p. 18. The average number of consuls rose from four since 5 B.C. to six in the first half of the first century, six to ten under the Flavians, six to eight under Trajan, eight under Hadrian, eight to ten under Antoninus Pius, ten under Marcus Aurelius and twelve under the Severans: Talbert, p. 21, and see Alföldy, p. 20 for Pius; this yields a mean of about 9.6 for the years from A.D. 138 to 180. Proportion of patricians: Alföldy, p. 56, calculates that between A.D. 138 and 161, thirty to forty out of 220 to 225 consuls (or 13–18 per cent) were patricians. Age of consulate for plebeians: Alföldy, p. 36. Emperors and ex-consuls: from A.D. 138 to 161, 4.5 per cent of consuls were rulers or ex-consuls (Alföldy, p. 20); of sixty-nine reasonably well attested consuls from A.D. 138 to 192, six held the consulate twice (three of them were patricians), or 8.7 per cent (Alföldy, pp. 327–45), but their share of all consulates must have been much lower (cf. also Eck, W., Senatoren von Vespasian bis Hadrian: Prospographische Untersuchungen mit Einschluß der Jahres- und Provinzialfasten der Statthalter [Munich, 1970], pp. 60–3Google Scholar, for iteration under Domitian). Five per cent seems a reasonable mean for Pius and Marcus.

13 Talbert (n. 7), pp. 18, 20.

14 Cf. e.g. Hopkins (n. 4), p. 149; Talbert (n. 7), p. 21.

15 There is no compelling reason to think so: Hopkins (n. 4), p. 147, n. 39.

16 Etienne, R., ‘La démographie de la famille d'Ausone’, Annales de démographie historique (1964), 15–25Google Scholar, and ‘La démographie des families impériales et sénatoriales au IVe siècle après J.C.‘, in Transformation et conflits au IVe siècle ap. J.-C. (Bonn, 1978), pp. 133–68Google Scholar.

17 For the same and other reasons, the data gathered by Etienne, R., ‘Ces morts que l'on compte dans la dynastie flavienne’, in Hinard, F. (ed.), La mort, les morts et l'au-delà dans le monde romain (Caen, 1987), pp. 65–90Google Scholar, are unsuitable for demographic analysis.

18 Thus Etienne, ‘Families’ (n. 16), p. 136, n. 21. From A.D. 315, he reckons with a minimum age of forty for the consulate and with forty-two to fifty for the praefect of Rome. See below.

19 Based on Etienne, ‘Ausone’ (n. 16), 21, and ‘Families’ (n. 16), Tables II–VII.

20 The same procedure has been adopted by Bagnall and Frier (n. 4), pp. 82, 101.

21 Duncan-Jones (n. 4), pp. 93–6. I shall discuss this inscription in more detail in another paper, ‘Death and renewal in a Roman local élite: the album of Canusium reconsidered’, CQ (forthcoming).

22 Ibid., p. 94. For a similar calculation, see the preceding section of this paper.

23 Cf. also Frier, B. W., ‘Statistics and Roman society’, JRA 5 (1992), 289Google Scholar.

24 Parkin (n. 3), pp. 137–8.

25 Saller (n. 4), p. 17, n. 22, mainly on the skipping of the quaestorship.

26 Jacques, F., Le privilège de liberté: politique impériale et autonomie municipale dans les cités de l'Occident romain (161–244) (Rome, 1984), pp. 489–93Google Scholar.

27 Saller (n. 4), p. 18. For another, less critical, discussion of Duncan-Jones' model, see Patriarca, F. Dal Cason, ‘Considerazioni demografiche sulla lista decurionale della tabula di Canusium’, Athenaeum 83 (1995), 246–8Google Scholar.

28 Duncan-Jones briefly refers to both mechanisms but does not fully appreciate their possible impact on the structure of office-holding. He also seems to take it for granted that each aedile surviving to the age of forty became a duovir, which is far from self-evident.

29 I derive this mortality level from my analysis of the Roman senate (above, second section). Adjacent levels would not significantly affect the results.

30 A. Caesellius Proculus (the twelfth most senior councillor listed in the album), L. Faenius Merops (no. 13), and Q. Iunius Alexander (no. 15).

31 See Jacques (n. 26), pp. 474 and 475, n. 148 (and cf. p. 476, n. 152).

32 Jacques (n. 26), p. 474.

33 T. Ligerius Postuminus (the most senior councillor), T. Aelius Rufus (no. 4), T. Aelius Flavianus (no. 5), and C. Galbius Soterianus (no. 8). On the patrons of Canusium, see Silvestrini, M., ‘Aspetti della municipalità di Canusium: l'albo dei decurioni’, MEFRA 102 (1990), 597–602CrossRefGoogle Scholar.

34 Compare Frier, ‘Demography’ (n. 4).

35 See Coale and Demeny (n. 3), pp. 42–4: at Level 2 (Model West), 53 per cent of males are dead by age 5; at Level 6, 36 per cent of all females die during the same period.

36 Although a few graveyards document plausibly high levels of infant burials (e.g. Pomeroy, S. B., Families in Classical and Hellenistic Greece: Representations and Realities [Oxford, 1997], p. 121Google Scholar and n. 66), most skeletal samples clearly conceal the true extent of early mortality; social factors often seem to have distorted the age-composition of cemetery populations. See e.g. Parkin (n. 3), pp. 41–58.

37 Thus Frier, ‘Life expectancy’ (n. 4); Bagnall and Frier (n. 4).

38 This problem is discussed by Woods, R., ‘On the historical relationship between infant and adult mortality’, Population Studies 47 (1993), 195–219CrossRefGoogle ScholarPubMed. Cf. also Preston, S. H., McDaniel, A. and Grushka, C., ‘New model life tables for high-mortality populations’, Historical Methods 26 (1993), 149–59CrossRefGoogle Scholar.

39 More often than not, even these sources fail us, as Syme, R., Roman Papers VI, ed. Birley, A. R. (Oxford, 1991), p. 241Google Scholar, notes: ‘Forgotten wives and evanescent children, such is the constant rubric of mortality at Rome.’ In The Augustan Aristocracy (Oxford, 1986), pp. 21–4Google Scholar, Syme draw sattention to the impact of epidemics on the Roman élite.

40 Ktesias FGrHist 688 F 15 § 49 (König, F. W., Die Persika des Ktesias von Knidos [Graz, 1972], pp. 19–20)Google Scholar; Hornblower, S., Mausolus (Oxford, 1982), p. XXVIGoogle Scholar. Both cases involved brother-sister marriage, which may well have increased the incidence of child mortality: see Scheidel, W., Measuring Sex, Age and Death in the Roman Empire: Explorations in Ancient Demography (Ann Arbor, 1996), pp. 9–51, esp. pp. 20–9Google Scholar.

41 Plut. Tib. Gracc. 1.3–5 (but cf. Moir, K. M., ‘Pliny N.H. 7.57 and the Marriage of Tiberius Gracchus’, CQ 33 [1983], 144)CrossRefGoogle Scholar.

42 Quint. 6 pr. 4–10.

43 Plut. Mor. 608C, 609D, 610E, 611D.

44 Fronto, , De Nepote Amisso, 2 and 4 (pp. 220 and 224, ed. Hout, van den) (Loeb, trans. Haines, C. R.)Google Scholar.

45 As usual, the following list is based on Kienast, pp. 139–40. For a somewhat different reconstruction with a total of fourteen children, see Birley, A., Marcus Aurelius: A Biography (2nd edn, London, 1987), pp. 239, 247–8Google Scholar. Since eight of them are listed as having died as small children, the tally (unlike some of the names) is the same. Cf. Ameling, W., ‘Die Kinder des Marc Aurel und die Bildnistypen der Faustina Minor’, ZPE 90 (1992), 147–66Google Scholar.

46 HA Max. Balb. 5.2: ‘cui fratres quattuor pueri fuerunt, quattuor puellas, qui omnes intra pubertatem interierunt’.

47 Like Vespasian, his brother Sabinus also produced two sons and one daughter who lived long enough to be married. One of them, T. Flavius Clemens, fathered at least seven children, only three of whom are known by name. All we know is that some of them survived long enough to be tutored by Quintilian but in each case even their approximate age at death remains obscure. For all we know, most or all of them could have died in childhood. Seven successive generations of Flavians are discussed by Etienne (n. 17), who at one point speculates (69) that at least three of Clemens' children were already dead by A.D. 96: the truth is that we cannot tell, and that the lack of proper evidence usually prevents us from investigating the families even of close relatives of Roman emperors.

48 HA Tac. 16.4: ‘et Floriani liberi et Taciti multi exstiterunt’.

49 Saller (n. 4), p. 57 Table 3.2.d, p. 63 Table 3.3.d. On thes e simulations (which are only available for Mortality Levels 3 and 6), see pp. 43–69; they ‘[generate] a model population by simulating the basic event s of birth, death and marriage, month by month, in accordance with the age-specific probabilities of those events as established by the demographic parameters' (p. 44). The final column in Table 4 is based on the following set of assumptions: life expectancy at birth was 25 or 32. 5 years (for women, here applied to either sex); mean age at first marriage was twenty-five for men (and fifteen for women); first marriages were distributed between the ages twenty and forty for men (and from twelve to thirty-three for women); after the death of a spouse, men would remarry up to the age of sixty (pp. 45–6). (Different parameters apply for the simulation of non-senatorial kinship patterns.)

50 If we speculatively assign three sons to Tacitus, we arrive at a mean of 0.787 sons, against 0.745 in the model, and 0.793 daughters. A general point deserves attention: the figures shown in Saller's tables, whilst calculated to higher accuracy, are rounded off to a single digit behind the comma. the cumulation of rounded-off figures to obtain a grand total and an overall mean therefore introduces a possible margin of error which, however, cannot exceed a range of plus/minus 3.7 sons or 6.6 per cent for the grand total and is unlikely to be larger than plus/minus 3.3 per cent; it is thus fairly negligible.

51 It is true that two of the last three emperors leftover one-sixth of all surviving sons. This peak is, however, offset by the long line of emperors without sons who preceded them. the match between evidence and model does not depend on the cut-off date: for the period from AD. 14 to 395, a mean of 0.722 surviving sons compares well with 0.744 sons in the simulation; if we stop counting in A.D. 476, the figures are very similar, 0.708 and 0.758, respectively.

52 Based on Saller (n. 4), p. 58 Table 3.2.e, p. 64 Table 3.3.e.

53 Ibid., p. 47.

54 Syme (n. 39), p. 235, stresses early marriage for Roman senators. See also Treggiari, S., Roman Marriage: Iusti Coniuges from the Time of Cicero to the Time of Ulpian (Oxford, 1991), pp. 401–2Google Scholar; Saller (n. 4), pp. 26, 45–6.

55 Betzig, L. and Weber, S., ‘Presidents preferred sons’, Politics and the Life Sciences 14 (1995), 61–4CrossRefGoogle Scholar, discuss male-biased sex ratios among the children of high-status parents.

56 Widespread family limitation: Riddle, J. M., Contraception and Abortion from the Ancient World to the Renaissance (Cambridge MA/London, 1992), pp. 1–107Google Scholar, and Eve's Herbs: A History of Contraception and Abortion in the West (Cambridge MA/London, 1997), pp. 10–90Google Scholar. For related ideas, see Wierschowski, L., ‘Der historisch-demographische Kontext der severischen Abtreibungsverbote’, Laverna 7 (1996), 42–66Google Scholar. Élite behaviour: Brunt, P. A., Italian Manpower 225 B.C-A.D. 14 (Oxford, 1971, repr. 1987), p. 142Google Scholar, claims that ‘it was surely the limitation of [aristocratic] families that explains their disappearance’. For a more cautious discussion of Roman élite fertility, see Hopkins (n. 4), pp. 78–107. Of course, emperors might be considered a special case within the élite: see below. Comparative evidence on élite fertility: e.g. Johansson, S. R., ‘Status anxiety and demographic contraction of privileged populations’, Population and Development Review 13 (1987), 439–70CrossRefGoogle Scholar.

57 Lead poisoning as a cause of sterility has finally gone outof fashion: see e.g. Scarborough, J., ‘The myth of lead poisoning among the Romans: an essay review’, Journal of the History of Medicine 39 (1984), 469–75Google Scholar; L., and Needleman, D., ‘Lead poisoning and the decline of the Roman aristocracy’, EMC n.s. 4 (1985), 63–86Google Scholar. (This is not to say that habitual ingestion of wine mixed with leaded sapa could not have resulted in serious health problems: cf. Eisinger, J., ‘Lead and wine: Eberhard Gockel and the colica pictonum‘, Medical History 26 [1982], 279–302CrossRefGoogle Scholar; Nutton, V., ‘Bleivergiftung’, Der Neue Pauly 2 [1997], 709.Google Scholar) There is also little point in replacing the lead-poisoning theory with speculation that gonorrhoea (for which there does not seem to be any specific evidence) may have been a common cause of sterility in Roman societv (Needleman, pp. 86–94).

58 Frier, B. W., ‘Natural fertility and family limitation in Roman marriage’, CPh 90 (1994), 318–21Google Scholar.

59 Widows include the elder and the younger Antonia and Vipsania Agrippina; others died young: see above, p. 258.

60 Later miscarriages, which become more frequent towards the end of the reproductive period, cannot be ruled out. Cf. Birley (n. 45), p. 191, for the possibility that she was pregnant at the time of her death.

61 Although the Historia Augusta (8.1) ascribes two daughters to Septimius Severus, their existence is doubtful; moreover, they could also have been borne by his first wife, Paccia Marciana.

62 Constantine I was married tofausta from A.D. 307 to 326 but four of their five children appear to have been born between A.D. 316 and c. 320. The curious lack of children during the first decade of their marriage was probably not the resultof family planning; rather, this gap points to the existence of earlier children who died soon after birth and remain unknown to us.

63 Thus already Syme, R., Roman Papers III, ed. Birley, A. R. (Oxford, 1984), p. 1237Google Scholar.

64 Mart. 10.63. Since her age at death and the ages of her children remain unknown, we can only very roughly estimate the odds against ten siblings surviving to the death of their mother; they must have been of the order of 1,000 to 1. This scenario is rendered even more remarkable by the fact that, as Martial notes, all these children shared the same father.

65 Treggiari (n. 54), p. 404. However, Pliny might also have referred to the six surviving children of Germanicus and Agrippina.

66 See above, n. 45.

67 She is assumed to have married in A.D. 4 or 5. If she was sixteen or seventeen then, she would have given birth to her first child at about seventeen or eighteen; this guess cannot be wide of the mark. Her reproductive cycle was cut short by Gennanicus' death in A.D. 19, one or two years after the birth of their ninth child. For a discussion of her reproductive history, see Lindsay, H., ‘A fertile marriage: Agrippina and the chronology of her children by Germanicus’, Latomus 54 (1995), 3–17Google Scholar (who considers the existence of a tenth child possible: 7–8).

68 For a possible example, see above, n. 62. It is not clear how long it took the elder Tiberius Gracchus and his wife Cornelia to produce twelve children. The view that these births occurred within a period of twelve years strains credulity (even if we allow for the possibility of multiple births: thus Parkin [n. 3], p. 181, n. 16). Moir (n. 41), pp. 136–45, plausibly argues that depending on the date of Cornelia's marriage, her twelve children were spaced outover fifteen or even twenty-six years. In the light of the evidence presented above, the implied mean birth intervals of 1.4 and 2.4 years both seem acceptable.

69 The families studied by Etienne readily prove this point: see above, nn. 16 and 17. As a consequence, scholars usually evoke ‘soft’, literary evidence associating the Roman élite with various forms of family limitation: e.g. Hopkins, K., ‘Contraception in the Roman Empire’, CSSH 8 (1965), 124–51Google Scholar; Frier (n. 58), p. 332.

70 For concepts such as ‘Family status maintenance’ and ‘Social reproduction in polities’, see Hopkins (n. 4), pp. 61–2. Even social reproduction is difficult to measure: Burton, G. P., ‘The inheritance of the consulate in the Antonine period: a problem revisited’, Phoenix 49 (1995), 218–31CrossRefGoogle Scholar.

71 Suder, W., ‘Démographie des femmes de l'ordre sénatorial (Ier–IIe siècle ap. J.-C.): fecondité’, Antiquitas 18 (1993), 199–201Google Scholar, based on Raepsaet-Charlier, M.-T., Prosopographie des femmes de l'ordre sénatorial (Ier–IIe s.) (Leuven, 1987)Google Scholar.

72 Suder (n. 71), p. 200 (where, strangely, he opts for two children per woman).

73 Suder somewhat disingenuously claims that ‘ce qui pour nous est important, c'est non pas d'avoir des chiffres absolus, mais des proportions et un'ordre de grandeur' (ibid., p. 201), but if he means to imply that these figures are to be taken as ‘la limite inférieure’ (p. 199) for the Total Fertility Rate, we must not forget that this limit may well be very far removed from the true rate.

74 Etienne, ‘Families’ (n. 16), Table X (97 children of 54 couples).

75 See Betzig, L., ‘Roman polygyny’, Ethology and Sociobiology 13 (1992), 309–49CrossRefGoogle Scholar, for a substantial survey (and Darwinian interpretation) of pertinent Roman evidence, and her Despotism and Differential Reproduction: a Darwinian View of History (Hawthorne, 1986), esp. pp. 70–86Google Scholar, and ‘Medieval monogamy’, Journal of Family History 20 (1995), 181–216, at 183–94Google Scholar, for comparative material.

76 see eeg Finley, M. I., Ancient Slavery and Modern Ideology (London, 1980), pp. 95–6Google Scholar; Kolendo, J., ‘L'esclavage et la vie sexuelle des hommes libres à Rome’, Index 10 (1983), 288–97Google Scholar; Herrmann-Otto, E., Ex ancilla natus: Untersuchungen zu den “hausgeborenen” Sklaven und Sklavinnen im Westen des römischen Kaiserreiches (Stuttgart, 1994), pp. 256Google Scholar, n. 59, 310–12 (cf. my review in Tyche 11 [1996], 274–8Google Scholar, with references to comparative evidence).

77 Syme (n. 63), p. 1238.

78 This calculation is based on Model West Mortality Level 4; see above in the section on the life expectancy of Roman emperors.

79 We would have tofall back on the extreme case of the Sasanid king Sapur II, who was crowned at birth and lived to be seventy, to better this record.

80 Hopkins, K., ‘Rules of evidence’, JRS 68 (1978), 186Google Scholar.

81 Livi-Bacci, M., Population and Nutrition: An Essay on Europeandemographic History (Cambridge, 1991), pp. 63–7CrossRefGoogle Scholar, and Johansson, S. R., ‘Food for thought: rhetoric and reality in modern mortality history’, Historical Methods 27 (1994), 101–25, at 113–14CrossRefGoogle Scholar, summarize pertinent research.

82 Cf. Johansson (n. 81), 114, for comparative evidence. There is little to support the notion that Roman ‘upper-class mortality may have been substantially lower than general levels’ (Frier [n. 58], p. 332, n. 57). Residence in the city of Rome was in itself a serious health hazard: see Scheidel (n. 40), pp. 139–53 (on seasonal mortality), and also pp. 124–9 (on mortality in military units stationed in the capital); Sallares and Scheidel (n. 4), ch. 1 (on malaria in Rome).

83 Soranus, Gyn. 2.18, asks mothers to abstain from breast-feeding their babies during the first twenty days after delivery, when in fact maternal milk at that stage is about three times as rich in protein as mature human milk and provides the child with vital antibodies. The same injunction applied to wet-nurses who had very recently given birth (Gyn. 2.20). Garnsey, P., ‘Child rearing in ancient Italy’, in his Cities, Peasants and Food in Classical Antiquity: Essays in Social and Economic History (ed. Scheidel, W.) (Cambridge, 1998), p. 264CrossRefGoogle Scholar, argues that while Soranus' medical advice (which did not pass uncontested) ‘would have significantly reduced the life chances of the children concerned’, it ‘would have had limited impact outside an upper-class clientele’. We may suspect that if anyone at all was harmed by fanciful medical theories, it must have been the Roman élite, including imperial families. In most cases, however, infections would take their toll regardless of medical interference.