Hostname: page-component-7c8c6479df-94d59 Total loading time: 0 Render date: 2024-03-28T04:11:24.726Z Has data issue: false hasContentIssue false

Demography, Geography and the Sources of Roman Slaves*

Published online by Cambridge University Press:  14 March 2012

W. V. Harris
Affiliation:
National Humanities Center/Columbia University

Extract

Where did a large-scale Roman slave-owner obtain new slaves? Varro in effect tells us: Ephesus. And the answer would probably have been the same for many generations after his time. But can we work out more systematically and more thoroughly the relative importance of different kinds of sources? The sources which most require consideration are: (1) children born to slave-mothers within the Empire; (2) persons enslaved in provincial or frontier wars; (3) persons imported across the frontiers; (4) the ‘self-enslaved’; and (5) infants abandoned at places within the Empire.

Several years ago, I argued on a number of grounds that the last of these sources, child-exposure, was more important than had previously been recognized. Subsequent reconsideration of the problem has led me to suspect that the source-material under-represents the amount of slave-importation across the frontiers, but not to doubt that child-exposure was very widespread or that it made an important contribution to the slave supply. Of the many subsequent discussions, the most original is that of Ramin and Veyne, who, in an article of 1981 too little attended to in the Anglo-Saxon world, made it appear very likely that those who voluntarily sold themselves into slavery were a larger category than scholars usually imagine. More recently, Scheidel has attempted to revive the case, previously propounded by Shtaerman among others, in favour of the self-reproductivity of the slave population.

Type
Articles
Copyright
Copyright © W. V. Harris 1999. Exclusive Licence to Publish: The Society for the Promotion of Roman Studies

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.)

References

1 De lingua Latina 8.21.

2 Towards a study of the Roman slave trade’, MAAR 36 ( = D'Arms, J. H. and Kopff, E. C. (eds), The Seaborne Commerce of Ancient Rome) (1980), 117–40Google Scholar. Ramin, J. and Veyne, P., ‘Droit romain et société: les hommes libres qui passent pour esclaves et l'esclavage volontaire’, Historia 30 (1981), 475, were less cautious: abandoned children ‘sont sûrement la source principale des esclaves sous l'Empire’.Google Scholar

3 Ramin and Veyne, op. cit. (n. 2), 472–97, repr. in Veyne, P., La Société romaine (1991), 247–80.Google Scholar

4 Scheidel, W., ‘Quantifying the sources of slaves in the Roman Empire’, JRS 87 (1997), 159–69Google Scholar. Shtaerman, E. M., Die Blütezeit der Sklavenwirtschaft in der römischen Republik (1969; the original edition was published in 1964), 70Google Scholar; Shtaerman, E. and Trofimova, M. K., La schiavitù nell'Italia imperiale (1975; original edn 1971), esp. 17 and 24Google Scholar. Her conclusion is more moderate and credible than Scheidel's.

5 Scheidel, op. cit. (n. 4), 165. I wrote (op. cit. (n. 2), 123) that the enslavement of abandoned children was ‘a far more important source’ of slaves than any other Italian or provincial source apart from those who were slaves by birth. The paper of Ramin and Veyne makes me doubt whether I should have written ‘far’. I have never by the way, pace Scheidel (156 n. 2), used the odd expression ‘social life expectancy’.

6 But see Hopkins, K., Conquerors and Slaves (1978), 141Google Scholar. It is also worth consulting the remarks of Henri Wallon on this matter, written in the 1840s without the benefit of CIL or life-tables: Histoire de l'esclavage dans l'antiquite (2nd edn, 1879), I, 158 and II, 101–4.

7 cf. Child-exposure in the Roman Empire’, JRS 84 (1994), 811Google Scholar. Only in classical fields, perhaps, would a degree of tension between what one wrote in 1980 and in 1994 be thought troubling.

8 Riddle, J. M., Contraception and Abortion from the Ancient World to the Renaissance (1992)Google Scholar; Eve's Herbs: a History of Contraception and Abortion in the West (1997). But according to Frier, B. W., ‘Natural fertility and family limitation in Roman marriage’, CPh 89 (1994), 318–33Google ScholarPubMed, there was little family limitation within marriage outside the upper class.

9 Which is not to suggest that all ancient historians must follow this path, or that such study is a vaccine against bad social history. And the question always remains – which economics, which anthropology? As for the present problem, Scheidel is in error in asserting that its difficulties derive solely from ‘lack of demographic conceptualization’ (156) – they also derive from among other things poor evidence, poor interpretation of the evidence, and poor weighing of historical probability.

10 Hopkins, K., ‘Rome, taxes, rent and trade’, Kodai 6/7 (1995/1996), 41Google Scholar. Contrast R. S. Bagnall and B. W. Frier, The Demography of Roman Egypt (1994), xvi–xvii.

11 1989 edition, s.v. model I.2.e. Cf. the definition from Chorley, R. J. and Haggett, P., Socio-Economic Models in Geography (1968 edn), 22Google Scholar, quoted by Finley, M. I., Ancient History: Evidence and Models (1985), 60Google Scholar. This definition stresses simplicity more, and hence covers the second sense mentioned above better than the first.

12 Coale, A. J. and Demeny, P., Regional Model Life Tables and Stable Populations (2nd edn, 1983)Google Scholar.

13 Finley op. cit. (n. n), 61, contrasted model-construction with the meaningless accumulation of facts in books about ancient cities, and that is probably a case in which model-construction can help, partly because the evidence is so unmanageable, as it is not, for example, in the case of the sources of Roman slaves. But Finley's dichotomy was itself seriously misleading.

14 See for instance Cascio, E. Lo, ‘The size of the Roman population: Beloch and the meaning of the Augustan census figures’, JRS 84 (1994), 2340Google Scholar; Scheidel, W., Measuring, Sex, Death and Age in the Roman Empire: Explorations in Ancient Demography (1996), 167–8Google Scholar. F. Coarelli has now argued for a population of the city of Rome as high as 1.2 million in early imperial times (‘La consistenza della città nel periodo imperiale: pomerium, vici, insulae’, in La Rome Impériale. Démographie et logistique (1997), 107).

15 I see the high ratio of slaves as having gradually taken hold in Italy over the course of the middle Republic; the addition of new provinces will have had varying effects on the overall ratio.

16 Thus nothing on this in E. M. Shtaerman et al., Die Sklaverei in den westlichen Provinzen des römischen Reiches im 1.-3. Jahrhundert (1987; original edn 1977), Marinovich, L. P.et al., Die Sklaverei in den östlichen Provinzen u.s.w. (1992; original edn 1977Google Scholar), Parkin, T. G., Demography and Roman Society (1992)Google Scholar, Herrmann-Otto, E., Ex Ancilla Natus (1993)Google Scholar, or Bradley, K. R., Slavery and Society at Rome (1994).CrossRefGoogle Scholar

17 See esp. Bieźuńska-Malowist, I., L'Esclavage dans l'Égypte gréco-romaine II (1977), 156–8Google Scholar; cf. ‘L'Égypte et l'histoire de l'esclavage dans l'antiquité’, in Criscuolo, L. and Geraci, G. (eds), Egitto e storia antica dall'ellenismo all'età araba (1988), 264Google Scholar.

18 In De propriorum animi cuiuslibet affectuum dignotione 9.13 (p. 33 Boer, De (CMG 5,4,1,1) = 5 p. 49Google Scholar Kühn), Galen lets it drop that the number of slaves was the same as the number of male citizens and as the number of women (citizens), namely 40,000. (See Mitchell, S., Anatolia (1993), II, 244Google Scholar, for the view that these numbers referred to the city itself without its chora.) The absolute value of this figure is slight (cf. Scheidel, W., ‘Finances, figures and fiction’, CQ 46 (1996), 222–38CrossRefGoogle Scholar, on the Graeco-Roman passion for the numbers 400, 40,000, 400,000), but the proportions may be roughly right. If the free population was 3.5 times that of the male citizens (Duncan-Jones, R. P., The Economy of the Roman Empire: Quantitative Studies (1974), 264 n. 4Google Scholar), that would mean that slaves made up 22.2 per cent of the total, but it seems unlikely that Galen intended to include very young slave children (since he is discussing the financial assets of the three groups); hence the proportion was probably higher than 22.2 per cent. Parkin, op. cit. (n. 16), 175 n. 187, prefers a multiplier of 4, not 3.5, but Caesar, , BG 1.29Google Scholar does not support this: the men who could bear arms (among the Helvetii) must have been a smaller set than the adult males. Those who believe in low estimates for the slave population regularly belittle Galen's information.

19 Rathbone, D., Economic Rationalism and Rural Society in Third-century A.D. Egypt (1991), 8991Google Scholar, 106–7.

20 See further ‘Between Archaic and Modern: some current problems in the history of the Roman economy’, in Harris, W. V. (ed.), The Inscribed Economy. Production and Distribution in the Roman Empire in the Light of Instrumentum Domesticum (JRA Supplementary Series 6) (1993), 25–7Google Scholar.

21 See for instance Higman, B. W., ‘Household structure and fertility on Jamaican slave plantations: a nineteenth-century example’, Population Studies 27 (1973)CrossRefGoogle ScholarPubMed, at 527 (repr. in Beckles, H. and Shepherd, V. (eds), Caribbean Slave Society and Economy (1991), 250Google Scholar).

22 Ramin and Veyne, op. cit. (n. 2), 481. See for instance Dio Chrys. 15.3–5.

23 cf. ‘Child-exposure’, op. cit. (n. 7), 14.

24 ‘Towards a study’, op. cit. (n. 2), 119–20; and see below, p. 69.

25 BC 1.7.29, an assertion rejected by E. Gabba ad loc. Herrmann-Otto, who discusses this passage without coming to any firm conclusion, errs in implying (op. cit. (n. 16), 234 n. 6) that the scholars she lists, including me, have gone so far as to suppose that there was no natural reproduction of slaves ‘worth mentioning’ under the Republic, which would be a bizarre position.

26 It could be argued that Appian's comment is better evidence if it is not authentically Republican; but it is in any case a slender reed.

27 Varro, , RR 2.10.6Google Scholar, and Colum. 1.8.19 are invoked by Scheidel, 169.

28 ‘Nos quidem … feminis quoque fecundioribus, quarum in subole certus numerus honorari debet, otium nonnumquam et libertatem dedimus, cum complures natos educassent. Nam cui tres erant filii [sons or children?] vacatio, cui plures libertas quoque contingebat’. Cf. Parkin, op. cit. (n. 16), 122. Such a rule must have put female infants at risk. On the desirability and affordability of vernae see ‘Towards a study’, op. cit. (n. 2), 118–19, 120.

29 The implications of Fragmenta de iure fisci 13 (FIRA, ed. Riccobono II, 629) are obscure.

30 Cicero translated this book in his youth (De off. 2.87), and the translation was widely read; see S. B. Pomeroy's commentary, p. 70. She is mistaken, however, in saying (p. 299) that for Xenophon slaves ‘born at home’ are the only acceptable ones: as far as I can see, he nowhere implies any such opinion, nor does Oec. 7.34 provide any evidence that, even in Xenophon's imagination, ‘Ischomachus’ slaves evidently do more than reproduce their numbers'.

31 With respect to pre-Severan times, this rests on an argument from silence and probability. Then there is the comment of Ulpian, in Dig. 5.3.27Google Scholar.pr.: ‘quia non temere ancillae eius rei causa comparantur ut pariant …’, ‘since slave-women are not commonly acquired so that they may produce children’; for the sense of non temere cf. Suet., , De gramm. 4.5Google Scholar, Gell. 20.5.4, and OLD sense 3. It used to be debated, not unreasonably, whether non temere was interpolated: De Martino, F., Storia economica di Roma antica (1980), 265–6Google Scholar. (T. Kinsey in A. Watson (ed.), The Digest of Justinian (1985), erroneously translates ‘slave girls are not acquired solely as breeding stock’.) There is no contradiction between Ulpian's words and Dig. 21.1.14.1 (Ulpian again) or 19.1.21.pr. (Paulus), texts which confirm the obvious fact that when women slaves were purchased the purchasers were sometimes (as they must normally have been) interested in their ability to bear children.

32 Scheidel, op. cit. (n. 4), 157 n. 14.

33 Thomas, H., The Story of the Atlantic Slave Trade 1440–1870 (1997), 57aGoogle Scholar. But other historians such as H. S. Klein dismiss this claim (personal communication).

34 See Tac., Germ. 25.

35 See, for instance, Eltis, D. and Engerman, S. L., ‘Fluctuations in sex and age ratios in the Transatlantic slave trade, 1663–1864’, Economic History Review 46 (1993), 308–23CrossRefGoogle Scholar, who say that the sex imbalance is normal for a migrating population. For an interesting attempt to combine factors in the source regions with differential demand see R. Olwell, Masters, Slaves and Subjects: the Culture of Power in the South Carolina Low Country 1740–1790 (1998), 28 n. 44. See now Klein, H. S., The Atlantic Slave Trade (1999)Google Scholar.

36 cf. Eltis, D. and Richardson, D., ‘West Africa and the Transatlantic slave trade: new evidence of long run trends’, in Eltis, and Richardson, (eds), Routes to Slavery: Direction, Ethnicity and Mortality in the Transatlantic Slave Trade (1997), 32–3.Google Scholar

37 cf. Harris, ‘Towards a study’, op. cit. (n. 2), 119–20.

38 That is, ‘the number of daughters that a cohort of newborn girl babies will bear during their lifetime assuming a fixed schedule of age-specific fertility rates and a fixed set of mortality rates’, Shryock, H. S., Siegel, J. S. et al. , The Methods and Materials of Demography (1976 edn), 315Google Scholar.

39 Scheidel, op. cit. (n. 4), 169 (end) is wrong to imply that we know such families to have been common among Roman slaves, and also mistaken in implying that Bagnall and Frier, op. cit. (n. 10), lend support to this view (see 156–9 for their most pertinent comments). See below for the argument that the sex-ratio detectable in the census-returns from the Egyptian chora (more female slaves than male) reverses the pattern prevailing in the Roman Empire as a whole.

40 For details see John, A. M., The Plantation Slaves of Trinidad, 1783–1816: a Mathematical and Demographic Enquiry (1988)Google Scholar.

41 See for instance Craton, M., ‘Death, disease and medicine on the Jamaican slave plantations: the example of Worthy Park, 1767–1838’, Histoire Sociale – Social History 9 (1976), 237–55Google ScholarPubMed, repr. in H. Beckles and V. Shepherd (eds), Caribbean Slave Society and Economy (1991); R. B. Sheridan, Doctors and Slaves: A Medical and Demographic History of Slavery in the British West Indies, 1680–1834 (1985).

42 Lamur, H., ‘Demographic performance of two slave populations of the Dutch speaking Caribbean’, Boletin de Estudios Latino Americanos y del Caribe 30 (1981)Google Scholar, cited from Beckles and Shepherd, op. cit. (n. 41), 216.

43 For some of the fatal remedies favoured by doctors during the Atlantic journey see Kiple, K. F. and Higgins, B. T., ‘Mortality caused by dehydration during the Middle Passage’, in J. E. Inikori and S. L. Engerman (eds), The Atlantic Slave Trade (1992), 321–37, esp. 327CrossRefGoogle Scholar.

44 See most recently Bankole, K. O., Slavery and Medicine: Enslavement and Medical Practices in Antebellum Louisiana (1998)Google Scholar.

45 There were epheremal official attempts in the second century A.D. to protect slaves from certain extremes of punishment and overwork; cf. Garnsey, P., Ideas of Slavery from Aristotle to Augustine (1996), 93–7Google Scholar.

46 Shaw, B. D., ‘The cultural meaning of death: age and gender in the Roman family’, in D. I. Kertzer and R. P. Sailer (eds), The Family in Italy from Antiquity to the Present (1991)Google Scholar, at 87, making use of the work of P. Zanker and D. E. E. Kleiner. Some slaves did of course live within stable family structures.

47 Fogel, R. W. and Engerman, S. L., Time on the Cross (1974), 126–44Google Scholar. The classic treatment, also much discussed, is Gutman, H. G., The Black Family in Slavery and Freedom, 1750–1925 (1976)Google Scholar, chs 2–4; see too P. Kolchin, American Slavery, 1619–1877 (1993), 138–43. Scheidel, op. cit. (n. 4), 163 n. 29 gives quite the wrong impression on this matter.

48 Herrmann-Otto, op. cit. (n. 16), esp. 264.

49 This matter is ignored in Scheidel's account.

50 This case has been reinforced by Bagnall, R. S., ‘Missing females in Roman Egypt’, Scripta Classica Israelica 16 (1997), 121–38Google Scholar.

51 Bagnall and Frier, op. cit. (n. 10), 94.

52 op. cit. (n. 10), 157.

53 ‘Towards a study’, op. cit. (n. 2), 119.

54 Higman, B. W., Slave Populations of the British Caribbean 1807–1834 (1984), 118Google Scholar.

55 e.g. Treggiari, S., ‘Family life among the staff of the Volusii’, TAPhA 105 (1975), 395Google Scholar.

56 See esp. Shaw, op. cit. (n. 46), at 81–2.

57 77 per cent of the commemorated household staff of Livia appears to have been male (Treggiari, S., ‘Jobs in the household of Livia’, PBSR 43 (1975), 58Google Scholar), as were 66 per cent of the commemorated town slaves of the Statilii and Volusii (these numbers include freedmen) (Treggiari, op. cit. (n. 55), esp. 395, who hesitantly argued, n. 10, that since burial clubs open to women were in question ‘they should have had an equal chance of being commemorated’). At Carthage, 76 per cent of the recorded members of the imperial household were male: P. R. C. Weaver, Familia Caesaris (1972), 172.

58 P.Oxy. XLIV.3197. No other Roman inventory of comparable size seems to have been published.

59 de Ste. Croix, G. E. M., The Class Struggle in the Ancient Greek World (1981), 588Google Scholar, with proper reservations about the potential significance of such evidence.

60 12.4.3: some authorities said that food should be served by persons without sexual contacts, i.e. male or female children; 8.2.7: a boy or an old woman should be put in charge of stray chickens. This is not a rich harvest from many hundreds of pages. In 12.3.5–9 he describes the duty of the vilica, and she seems strangely isolated from other women.

61 cf. Patterson, O., Slavery and Social Death (1982), 134Google Scholar. Scheidel, op. cit. (n. 4), 169, attributes to Eltis and Engerman, op. cit. (n. 35), 321, the view that the demographic effect of an unnatural sex-ratio on a slave population ‘should not be overrated’, implying apparently that it was never important, which is not at all what they say.

62 If we descended to the era of Justinian, or even of Diocletian, the story might be different (see the end of this article); the point is not to deny that the natural sex-ratio ever reasserted itself, but that it did so quickly. The Aezani text of Diocletian's Price Edict shows that female slaves received the same valuation as males in only one age-group, from eight to sixteen; prospective fertility is likely to be one of the causes (cf. Scheidel, W., ‘Reflections on the differential valuation of slaves in Diocletian's Price Edict and in the United States’, Münstersche Beiträge zur antiken Handelsgeschichte 15, 1 (1996), 6779Google Scholar).

63 Alföldy, G. used this expression in ‘Die Freilassung von Sklaven und die Struktur der Sklaverei in der römischen Kaiserzeit’, RSA 2 (1972), 97129Google Scholar, but stepped back from it in the Nachträge accompanying the reprint in Die römische Gesellschaft (1986), 286–331, at 330; these Nachträge do not address the problem of the slave-supply. Against Alföldy's theory of easy manumission: ‘Towards a study’, op. cit. (n. 2), 118; Wiedemann, T. E. J., ‘The regularity of manumission at Rome’, CQ 35 (1985), 162–75CrossRefGoogle Scholar.

64 Phil. 8.32: ‘cum in spem libertatis sexennio post sumus ingressi diutiusque servitutem perpessi quam captivi frugi et diligentes solent’ (49 B.C.–43 B.C. = 6). This cannot be nonsense. Rather, certain kinds of slaves could hope for freedom after six years. Perhaps wealthy Romans already made a mental division of slaves into quasi-classes, as they certainly did later on (Ulpian in Dig. 47.10.15.44). Incidentally Scheidel's assertion (op. cit. (n. 4), 158) that I hypothesize a ‘staggering amount of social mobility’ is spun out of nothing.

65 op. cit. (n. 6), 139.

66 Weaver, P. R. C., ‘Children of freedmen (and freedwomen)’, in Rawson, B. (ed.), Marriage, Divorce and Children in Ancient Rome (1991), at 179–82Google Scholar.

67 For the concept of cohort fertility see, e.g., Newell, C., Methods and Models in Demography (1988), 5262Google Scholar, Halli, S. S. and Rao, K. V., Advanced Techniques of Population Analysis (1992), 42–5CrossRefGoogle Scholar. The cohort we are considering will not, of course, all have been born into slavery.

68 Coale and Demeny, op. cit. (n. 12), 57, 82, 399, 449. The applicability of Model West (or South) to the Ancient World needs to be re-examined in the light of the history of causes of mortality (none of the 130 tables underlying Model West goes back earlier than 1870 (Coale and Demeny 12), a date later than, among other things, Lister's discovery of antisepsis; five of the twenty-two tables underlying Model South are from Italy, 1876–1910, all the others are from 1900 or later (ibid.)), but the problem cannot be pursued here. It is unlikely, for example, that many if any of the countries whose statistics went to make up Model West had nearly as high a level of infant mortality from child-abandonment. Level 3, incidentally, means that the table concerns a population in which eo = 25.

69 Pleket, H. W., ‘Wirtschaft’, in Fischer, W. et al. , Handbuch der europäischen Wirtschafts- und Sozial- geschichte (1990), I, 57Google Scholar, Shaw, B. D., review of Parkin, T. G., Demography and Roman Society, CPh 89 (1994). 190–1Google Scholar.

70 On the assumption that there were about 105 live male births for every 100 female ones.

71 cf. Duncan-Jones, R., Structure and Scale in the Roman Economy (1990), 100–1.CrossRefGoogle Scholar

72 If the sex-ratio of the slave population was really as high as 300, self-reproduction would have required even more absurd levels of fertility; but I do not dismiss the possibility that after some decline such a population might reach the sex-ratios used in the text.

73 Applying Model South would require an even higher GRR.

74 Michael Haines's extrapolation (personal communication) from the estimate of A. J. Coale and N. W. Rives that the Total Fertility Rate of the whole black population, slave and free, of the US in 1850–1859 was 7.90 (Population Index 39 (1973), 26Google Scholar).

75 op. cit. (n. 61), 133.

76 Scheidel, op. cit. (n. 4), 159.

77 Somalia : see esp. Periplous Marts Erythraei 13. Sahara: CIL VIII.4508, with Mattingly, D.J., Tripolitania (1994), 156Google Scholar; Brett, M. and Fentress, E., The Berbers (1996), 68–9Google Scholar. For Mauretania see ‘Towards a study’, op. cit. (n. 2), 126.

78 Scheidel, op. cit. (n. 4), 159–60.

79 The notion that in Roman times there were major population movements in NE Europe, having gone through a period of unpopularity, seems to be taking hold again; see e.g. Heather, P., The Goths (1996), 4850Google Scholar.

80 For a survey of the evidence see ‘Towards a study’, op. cit. (n. 2), 124. If as Tacitus says (Germ. 19) the Germans did not expose infants, that may have been because in case of necessity they, in essence, exported some of them as slaves; they supposedly exported persons who were enslaved for gambling debts (Germ. 24).

81 There is no demographic reason why there should not have been 40,000 slaves imported every year over a long period, though if compelled to guess I would opt for a lower figure. Scheidel, op. cit. (n. 4), 164 n. 34, seems to imply that the figure of 70,000 a year, said to be the maximum reached in the Atlantic slave trade, means that 40,000 is too large a figure, but the one figure has no bearing on the other.

82 But see, e.g., Crook, J., Law and Life of Rome, 90 B.C.–A.D. 212 (1967), 5960Google Scholar; Alföldy, op. cit. (n. 63), 125 (315 in the 1986 repr.).

83 De ben. 4.13.3. They are among those who ‘summam utilitatem aliis adferunt’.

84 Clem. Rom. 1.55.2; Papinian, in Dig. 41.3.44Google Scholar pr. (‘frequenter ignorantia liberos emimus’); Petr., , Sat. 57.4Google Scholar; Dio Chrys. 15.23 (?); Ulpian, in Dig. 21.1.17.12Google Scholar, 28.3.6.5.

85 For a soldier, self-sale was not surprisingly a capital offence, Dig. 48.19.14 (Macer).

86 cf. Aubert, J.-J., Business Managers in Ancient Rome: a Social and Economic Study of Institores, 200 B.C.–A.D. 250 (1994), 194Google Scholar.

87 Ramin and Veyne, op. cit. (n. 2), 496, consider it to be the third great source together with foundlings and self-reproduction.

88 See Garnsey, P., Famine and Food Supply in the Graeco-Roman World (1988)CrossRefGoogle Scholar.

89 Braudel, F., Civilization and Capitalism, 15th–18th Century. I. The Structures of Everyday Life: The Limits of the Possible (1981; original edn 1979), 77Google Scholar.

90 cf. ‘Child-exposure’, op. cit. (n. 7), 1, 20–1.

91 Ramin and Veyne, op. cit. (n. 2), 477.

92 Scheidel, op. cit. (n. 4), 164.

93 In Tilly, L. A.et al., ‘Child abandonment in European history: a symposium’, J. of Family History 17 (1992), 15CrossRefGoogle Scholar. See also The theoretical possibility of extensive infanticide in the Graeco-Roman world’, CQ 32 (1982), 114–16CrossRefGoogle Scholar. P. Brulé has hypothesized that more than 50 per cent of female infants were exposed in some Hellenistic cities, Enquête démographique sur la famille grecque antique’, RÉA 92 (1990), 233–58Google Scholar.

94 Scheidel, op. cit. (n. 4), 165 n. 37, attempts to dispose of the small proportion of the comparative evidence which he takes notice of, but it is not clear what reason or reasons he advances for doing so. Not a demographic one certainly. His point seems to be that high levels of abandonment were brought about by the existence of foundling hospitals, which were of course unknown in antiquity. But very high levels of abandonment are known from, indeed commonplace in, other worlds without foundling hospitals (Dickeman, M., ‘Demographic consequences of infanticide in man’, Annual Review of Ecology and Systematics 6 (1975), 130CrossRefGoogle Scholar); and the basic premise is faulty, for we must assume that abandoning parents had at least a rough idea that foundling hospitals were dangerous — and being cauldrons of disease, it is quite possible that traditional foundling hospitals led to a higher mortality rate than Graeco-Romana exposure did.

95 Engelmann, H. and Knibbe, D., ‘Das Zollgesetz derProvinz Asia’, Epigraphica Anatolica 14 (1989), 11. 11–12, 98–9, 117–22Google Scholar.

96 cf. ‘Toward s a study’, op. cit. (n. 2), 126–8. I should also have mentioned the allusion to the slave trade in the first-century customs law of Caunos (JHS 74 (1954), 97105Google Scholar = SEG XIV (1957), no. 639)Google Scholar.

97 See the probably Flavian inscription published by Herrmann, P., ‘Neues vom Sklavenmarkt in Sardeis’, Arkeoloji Dergisi 4 (1996), 175–87Google Scholar (the text has been quoted elsewhere, e.g. SEG XLIII (1994), p.3111)Google Scholar.

98 The question of the characteristic architecture of Graeco-Roman slave-markets will be re-examined at a conference organized by Elizabeth Fentress which is due to take place at the American Academy in Rome in June 2000.

99 About the slave-trade there is more to say elsewhere in view of such studies as Coarelli, F., ‘“Magistri capitolini” e mercanti di schiavi nella Roma repubblicana’, Index 15 (1987), 175–90Google Scholar.

100 ‘Towards a study’, op. cit. (n. 2), 121. For detailed but inconclusive discussion of the pay of the vigiles see Sablayrolles, R., Libertinus miles. Les cohortes de vigiles (1996), 333–42.Google Scholar

101 Scheidel, op. cit. (n. 4), 156.