¹ M.W. Flinn, The European Demographic System 1500–1820 (1981), 19 f.

² Even for these lands, however, possibly the best set of statistics we shall ever have, census figures from Roman Egypt, indicate quite clearly, as against the averages obtained from the epigraphical data from the west, that ‘barely half of all women were married by the age of 20–24 years’ (N, i.e. number analysed = 192), see K. Hopkins, ‘Brother-Sister Marriage in Roman Egypt’, Comparative Studies in Society and History 22 (1980), 333–4 and his Table 3 and fig. 5; the average age at first marriage for girls in this sample seems to lie in the late teens or early twenties. Hopkins called the results ‘striking and unexpected’ when set against ‘the early age at marriage of girls (median 15.5 years) recorded on stone inscriptions in the western half of the Roman empire. By that standard, a considerable proportion of Egyptian families were postponing the marriage of their daughters by up to ten years.’

³ J. Hajnal, ‘European Marriage Patterns in Perspective’, ch. 6 in Glass, D.V. and Eversley, D. E. C. (eds), Population in History (1965), 101–43.Google Scholar

⁴ Flinn, op. cit. (n. 1), 19–20.

⁵ See, for instance, Livi-Bacci, M., A History of Italian Fertility during the last two centuries (1977), 15 (1659–1866) and 100Google Scholar (1896 1900 to 1966–1969); and Bell, R.M., Fate and Honor, Family and Village: Demographic and Cultural Change in Rural Italy since 1800 (1979), 78 ff.Google Scholar

⁶ In the mid-60s, before much of the more recent data collation had taken place, Hajnal, op. cit. (n. 3), 120, made the following guess about the nature of these earlier marriage patterns: ‘It does not seem possible that the populations of medieval Europe had the fully developed European marriage pattern; they must either have had a marriage pattern clearly classifiable as non-European, or else some mixture of the two types with a wider variation of age at first marriage than is found later’ (my italics). He also noted that ‘in the Middle Ages the betrothal of children and the marriage of very young adolescents were apparently widespread throughout the population (not only among the nobility).’

⁷ Herlihy, D. and Klapisch-Zuber, C., Les Toscans et leurs families: une étude du catasto florentin de 1427 (1978), 204 ff.Google Scholar; see the abridged Engl. transl. Tuscans and their Families: A Study of the Florentine Catasto of 1427 (1985) at 202 and 210 ff. See also the review of the first French edition by Smith, R. M., ‘The People of Tuscany and their Families in the Fifteenth Century: Medieval or Mediterranean?’, Journal of Family History 6 (1981), 107–28. I now think that the answer to Smith's question must be that the Florentine families reported by Herlihy and Klapisch–Zuber were species of ‘Mediterranean’ type families, similar to those found in the Roman west. Whether or not one wishes to claim that these types are also in some sense ‘medieval’ is another, perhaps semantic, problem, although there is no reason to believe that, given the probable elements of continuity, they were not that as well.CrossRefGoogle ScholarPubMed

⁸ Bell, op. cit. (n. 5), 89, even though the mean for all first marriages for women in the village was about 20; for the other cases see pp. 78 ff.

⁹ Hopkins, K., ‘The Age of Roman Girls at Marriage’ Population Studies 18 (1964–1965), 309–27.CrossRefGoogle Scholar The article has been very influential in determining the age at marriage of Roman girls accepted by other scholars; see e.g., Weaver, P. R. C., Familia Caesaris: A Social Study of the Emperor's Freedmen and Slaves (1972) at 105–6 and 182–3CrossRefGoogle Scholar— one amongst many such examples. Brunt, P.A., Italian Manpower, 225 B.C. – A.D. 14 (1971), 137 f. depends on Hopkins’ analysis in his statement that the ‘epigraphic evidence shows that the largest number of first marriages were contracted by pagan girls between the ages of 12 and 15, and bears out the deductions that might be made from the more meagre literary testimony’. It is only fair to add that Brunt accepts this analysis with what can only be described as considerable reservations, which he states, and then adds ‘Can we be sure that the poor married so early (when they married at all)?’. He goes on to doubt whether arguments based on the maximum ages permitted for receiving the benefits of alimentary programmes have any relevance for the age-at-marriage problem.Google Scholar

¹⁰ Hopkins, (1964–1965), 313Google Scholar. This is not the place to raise a whole range of other matters relevant to the subject under investigation, but one should, for instance, note how flimsy the evidence is for the consummation of such prepubertal marriages. For the counter arguments see Durry, M., RIDA 2 (1955), 263–73;Google ScholarAnthropos 50 (1955), 432–4; CRAI (1955), 84–91; Gymnasium 63 (1956), 187 ff.; cf. his ‘autocritique’ in RIDA 3 (1956), 227–43, and the comments by Reinach, J., ‘Puberté féminine et mariage romain’, RHD 33 (1956), 268–73. Durry's argument rests in the main on the fact that some girls were married at very young ages and that these ages were probably before menarche. One must also object that, as yet, the problem of residence has not been adequately dealt with; in some cases, at least, one could argue that ‘marriage’ seems not necessarily to have meant an actual move to the husband's house in the case of marriages at very young ages.Google Scholar

¹¹ Hopkins (1964–1965), 319 and his fig. 1.

¹² Ibid., 326.

¹³ Harkness, A. G., ‘Age at Marriage and at Death in the Roman Empire, TAPhA 27 (1896), 35–72,Google Scholar mainly pagan but with some Christian examples; he outlined the basic literary sources (which then became standard fare in the subsequent debate), and then assembled his epigraphical data based on the volumes of CIL then available; Leclercq, H., ‘Mariage’, DACL 10. 2 (1932), 1843–1982, at cols. 1967—1973,Google Scholar whose lists are an updated version to (1932) of those collected by Pelka, O., ‘Aetas Nubilis’, Pt. 111. 1 of his Altchristliche Ehedenkmäler (in Zur Kunstgeschichte des Auslands v, 1901), 47–74Google Scholar: B, ‘Frauen’, 55–69: 165 cases. The latter list was restricted to Christian epigraphical data assembled from CIL, Fabretti, Boldetti, de Rossi (vol. 1), and a number of other collections published before 1900. For Rome, Leclercq is now superseded by the collection in Carletta (1977), see n. 23 below.

¹⁴ For these earlier studies see M. Bang, ‘Das gewöhnliche Alter der Mädchen bei der Verlobung und Verheiratung’, app. 11 in Friedlander, L., Darstellungen aus der Sittengeschichte Roms iv (9th-10th. ed., 1921), 133–41Google Scholar, and Harkness in n. 13 above.

¹⁵ Hopkins (1964–1965), 322; since the social composition of the sample is in fact dominated by slaves and freedmen, however, the suggestion that the dedicators were primarily from the ‘well-to-do middle classes’ needs to be refined.

¹⁶ Ibid., 326; cf. 322 where he denies that the omans tended to record younger marriages disproportionately on stones.

¹⁷ The standard repertoire of references repeated by Pelka, Leclercq, Bang (in Friedländer), Harkness, and Hopkins, amongst others, includes Epictet., Enchirid. 40; Plut., Mor. 138e = Praecepta Coniugalia; Comp. Vit. Lycurg. et Num. 4; Suet., Aug. 62. 1, Calig. 7, Claud. 26. 1; Dio 48. 5. 3; T a c, Ann. 2. 41, 2. 54, 6. 15, 12. 58, 14. 64; Agr. 9; Quintil., Inst. Or. 6, pr. 4; and a few other peripheral instances. Where these are not statements of an ideal, as in the case of Epictetus and Plutarch, the actual cases cited are overwhelmingly related to special marriages within the imperial family or the senatorial ‘aristocracy’.

¹⁸ Hopkins, while citing the instances, often expresses his scepticism of them (e.g. (1964–1965), 315–17); but Brunt, op. cit. (n. 9), 138 seems to be the only scholar consistently to take this bias in the data seriously. He questions the postulation of low ages at first marriage for the mass of the population, and in commenting on the general conditions of ‘the poor’, for instance, he states ‘these conditions make me doubt whether the evidence we possess for the mean age of marriage has any validity for the mass of the population'. Brunt then opts for a much later average age for the general populace of girls, in specific contrast to the younger modes common amongst ‘the wealthy’, noting the specific connection with slavery. At pp. 138–9, he outlines several economic constraints that would militate against the assumption of very young wives by the less well-off urban males (cf. Herlihy and Klapisch-Zuber, op. cit. (n. 7), 221 f.). One might also note that most of these ‘observations’ about Roman girls are made by outsiders to Roman society, principally Greeks, and that they therefore assume the context of ‘anthropological’ curiosities about certain elements of Roman society and their modes of behaviour that were of particular interest to the observers.

¹⁹ Hajnal, op. cit. (n. 3), 113 f. is a sufficient demonstration of this social contrast.

²⁰ Once again, Hajnal, Ibid., 116 f.noted this diffic-ulty some time ago for almost any pre-modern period, but especially for the European Middle Ages and earlier.

²¹ There is a possible example from Britain: RIB 295 = CIL VII, 160 (Viroconium).

²² The Po Valley region and Aemilia in the north do produce some ‘pagan’ examples, but the greater number by far are from the Christian period (Table 2), surely reflecting not only the correlate of ‘urbanism’. but also the shift of late imperial capitals to northern Italy in the period.

²³ Carletta, C., ‘Aspetti biometrici del matrimonio nelle iscrizioni cristiane di Roma’, Augustinianum 17 (1977) 39–51.CrossRefGoogle Scholar Carletta's study is by far the most thorough analysis of the Christian evidence, based on a collation of data from G. B. de Rossi, ICUR, i–m; E. Diehl, ILCV, and H. Zilliacus, S1CV; see also C. Vogel, ‘Facere cum Virginia (-o) sua (-o) annos…; l'âge des époux Chrétiens au moment de contracter mariage, d'après les inscriptions paléochrétiennes’, Revue de Droit Canonique 16 (1966), 355–66, who cites the earlier relevant literature.Google Scholar

²⁴ An analysis of the Leclercq-Harkness figures, for example, will readily show that the average duration of marriage for women dying in their teens to midtwenties, when subtracted from their age at death, will yield an average age at first marriage of about 15, whereas for those women dying in their mid-thirties to mid-forties, the average duration is not commensurately longer, so that for them the subtraction indicates an average age at first marriage in their early twenties.

²⁵ At Rome, 9. 8 (general population), 9. 5 (slaves), and 12. 5 (freedmen) per cent of all funerary stones set up for females were for girls in that age range (the different social distribution is in itself significant, since most of the inscriptions of the type hitherto used in the argument over age at first marriage were those set up by freed elements in the population of Rome and its environs). But any of these death rates would be exceptional, if not unprecedented, in an actual population. For a model population one would expect about 5.2–5.4 per cent female deaths in that age range; see Coale, A. J. and Demeny, P., with Vaughn, B., Regional Model Life Tables and Stable Populations, (2nd ed., 1983): Model South, Level 3.Google Scholar

²⁶ See Flinn, op. cit. (n. i), 22 ff., citing the fundametal paper by Wrigley, E. A., ‘A Simple Model of London's Importance in Changing English Society and Economy, 1650–1750’, P&P 37 (1967), 44–70Google Scholar, and the reply by Sharlin, A., ‘Natural Decrease in Early Modern Cities: A Reconsideration’, P&P 79 (1978), 126–38.Google Scholar

²⁷ For some of these see nn. 24–5 above.

²⁸ Saller, R. P. and Shaw, B. D., ‘Tombstones and Roman Family Relationships in the Principate: Civilians, Soldiers and Slaves’, JRS 74 (1984), 124–56, esp. 1267. In part the argument for expected types of commemorators (in the case of husband as opposed to father/mother) reposes on the structural form of the family, the sentiments attached to it, and the element of residence of the conjugal unit; in part it also depends on legal provisions that lay the obligation on the husband to perform the funeral rites for his deceased wife if he retains the dowry (as I assume to be the case in most instances). For the latter argument see Dig. n. 7. 16; 20; 22; 28 and 29.Google Scholar

²⁹ The samples taken were those of Rome/Italy, North Africa/Lambaesis and the Danube; in none of these regional samples did ‘other persons’ amount to 5 per cent of the total, and often were less than this proportion of the total.

³⁰ Initially, separate samples were taken for each of the Spanish provinces—Baetica, Lusitania, and Tarra-conensis—but no significant differences were found between them, and so the Spanish evidence is pre-sented as a whole.

³¹ See Sailer-Shaw, op. cit. (n. 28), 138–9 for an initial impression of the unusual ‘female dominance’ in the Spanish funerary epigraphy.

³² P. Laslett, ‘Family and Household as Work Group and Kin Group: Areas of Traditional Europe Compared’, ch. 17 in Wall, R. (ed.), Family Forms in Historic Europe (1983), 513–63, at 526 ff.CrossRefGoogle Scholar

³³ Duncan-Jones, R. P., ‘Age-rounding, Illiteracy and Social Differentiation in the Roman Empire’, Chiron 7 (1977), 333–54Google Scholar and ‘Age-Rounding in Greco- Roman Egypt’, ZPE 33 (1979), 169–77.Google Scholar

³⁴ Many of them were from the households of the upper echelons of the administrative élite of the Roman emperor, or of wealthy families in the city; in this respect our distribution reflects that of married servile persons designated contubernales in the Rome sample, see Treggiari, S., ‘Contubernales in CIL 6’, Phoenix 35 (1981), 42–69, at 46 f. (note the concept of ‘extended familia’, which might usefully be applied to private households, as well as the imperial palace), 50 f., 52, and 61: ‘What sort of people commemorate themselves as contubernales? As far as we can tell, predominantly the slaves and ex-slaves of the imperial family, the aristocracy and their dependants, that is, domestics from the city households and administrators in the service of the emperor or senators and equites. Prosperous freedmen of the class of tabernarii and opifices, who must have formed a substantial proportion of the slave and freed population of the capital, but who rarely appear linked with the great houses, hardly show up here.’CrossRefGoogle Scholar

³⁵ Taylor, L. R., ‘Freedmen and Freeborn in the Epitaphs of Rome’, AJPh 82 (1961), 113–32Google Scholar; cf. n. 37 below and the remarks made there by Kleiner.

³⁶ A further conclusion that may then be reached for these data is that although a large sector of the population of Rome has provided us with the bulk of our funerary epigraphy from the city, it is not necessarily the predominant one numerically. If this deduction is valid, it is then highly misleading to make deductions from this corpus of data about the relative proportions of slave/freed and free in the make-up of the whole population of the city. Nevertheless, this has been a common assumption. For example, Brunt, op. cit. (n. 9), 386–7, makes this argument even while wondering whether the statistics taken from epitaphs would tend to underrate the proportion of men of free birth in the whole population. Despite these misgivings, he still opts for the traditional view: ‘It seems to me safe to conclude that slaves and freedmen accounted for well over two-thirds of the urban population in 70, perhaps three-quarters’.

³⁷ See Kleiner, D. E. E., Roman Group Portraiture: The Funerary Reliefs of the Late Republic and Early Empire (1977), at 188–90Google Scholar; Kleiner's materials are restricted to the funerary monuments from the city of Rome and environs that are our subject here.

³⁸ Hopkins (1964–1965), 319—20, ‘It would be interesting to be able to explain the continuation of the old pagan practices and also the general rise in the age of marriage shown by the Christian inscriptions which probably came from the fourth century onwards. It would be possible to see the change as in some way a by-product of Christian asceticism; but that is only partially an explanation, and in part re-labelling. It is still necessary to explain what pressures drove Christians to delay marriage in the cause of asceticism, and to trace their reactions to these changes in pressure and to the changes in social behaviour'. As I hope to demonstrate in what follows, such a privileged explanation, including recourse to ‘Christian asceticism’ is probably unnecessary.

³⁹ Carletta, op. cit. (n. 23), 41—42 and his Table 1, p. 49; the calculations are mine.

⁴⁰ An impression of the difference can be gained quickly by looking at the representative illustrated sample in Zilliacus, SICV.

⁴¹ I have made this point elsewhere with regard to other aspects of family life in the Roman West, see Shaw, B. D., ‘Latin Funerary Epigraphy and Family Life in the Later Roman Empire’, Historia 33 (1984), 457–97, at 483 f.Google Scholar

⁴² CJ 5–4–24 (AD–53°): ‘in quo nuptiarum aetas vel feminis post duodecimum annum accesserit vel maribus post quartum decimum annum completum’; cf. Dig. 23. i. 9 (Ulpian) and Dio 54. 16. 7 (as part of Augustus’ programme); the ‘twelfth year’ and the ‘fourteenth year’ being age 11 and 13 in our terms; the text of Gaius, Inst. 1. 196 is defective at precisely the point where the age is given.

⁴³ Hence it was not difficult for Bang, op. cit. (n. 14), 140—1 to cite many examples of very young ages at first marriage for girls in these centuries; the problem is that they were not an accurate reflection of general practice. For such variance for the centuries before the sixteenth century, and even for Mediterranean contexts, see Herlihy, D., Medieval Households (1985), 103–7Google Scholar, cases where census records reveal a different average pattern from the indications of some of the striking literary examples of marriages at very young ages.

⁴⁴ W. W. Buckland and A. D. McNair, rev. Lawson, F. H., Roman Law and Common Law (1965), 47Google Scholar, with Flinn in n. 1 above; in the United States many states, including Iowa, Kentucky, New York, South Carolina, Texas and Utah set legal minima at 16 for boys and 14 for girls; many of the New England states, including Massachusetts and New Hampshire set them at the Roman/Civil Law limits of 14 for boys and 12 for girls (as did Mississippi); see Howard, G. E., A History of Matrimonial Institutions (Chicago, 1904; repr. New York, 1964), 11. 395–7Google Scholar; Bernard, W., Law for the Family (New York, 1962), 10–13;Google Scholar and Goldstein, J. and Katz, J., The Family and the Law (New York, 1965), 13.Google Scholar

⁴⁵ See Herlihy, n. 43 above; when early modern states began to abandon the Roman/Civil law minimum ages and to institute those normative ages of expected first marriage in legislation, these new ages were very much closer to the actual average age at first marriage; see, e.g., Ozment, S., When Fathers Ruled: Family Life in Reformation Europe (1983), 37–8, 58 ff.Google Scholar

⁴⁶ Ulpian, Tit. 16. I =FIRA 11², 278–9: ‘velut si uterque vel alteruter eorum [sc. vir et uxor] nondum eius aetatis sint, a qua lex liberos exigit, id est si vir minor annorum XXV sit aut uxor annorum XX minor’.

⁴⁷ Sailer, R., ‘Men's Age at Marriage and its Consequences in the Roman Family’, CPh 82 (1987), 21–34Google Scholar, esp. 28–30; cf. Hopkins (1964–1965), 323, noting that sons were commemorated by their parents to much later average ages than were daughters.

⁴⁸ I would like to thank Professor Richard Sailer for his comments and criticisms of this paper, and for allowing me to consult his paper on the age of males at first marriage (see n. 47 above) in advance of its publication. An earlier version of this paper was presented to the annual meeting of the Association of Ancient Historians at the University of Pittsburgh, 4 May 1985. I would like to thank the participants for their perceptive questions; Professor Susan Treggiari, the commentator, for her remarks; and, above all, Professor Keith Hopkins for his suggestions on a subject which he has made his own.