¹ Crook, J. A., Law and Life of Rome (1967), 98Google Scholar, quoted in Bradley, K. R., Discovering the Roman Family: Studies in Roman Social History (1991), 125Google Scholar; see also Treggiari, S., Roman Marriage: Iusti Coniuges from the Time of Cicero to the Time of Ulpian (1991) 410Google Scholar; Champlin, E., Final Judgments: Duty and Emotion in Roman Wills, 200 B.C.–A.D. 250 (1991), 103–4Google Scholar; Kertzer, D. I. and Saller, R. P. (eds), The Family in Italy from Antiquity to the Present (1991), 10, 72–3Google Scholar; Wallace-Hadrill, A., Houses and Society in Pompeii and Herculaneum (1994), 92Google Scholar. Bradley's several articles on the Roman family, it should be noted, have attempted to question and destabilize, though not entirely reject, the consensus.

² Saller, R. P., ‘Familia, Domus, and the Roman conception of the family’, Phoenix 38 (1984), 336–55CrossRefGoogle Scholar, at 342; for fuller discussion of the definitional issues, see Saller, R. P., Patriarchy, Property and Death in the Roman Family (1994), 74–88.CrossRefGoogle Scholar

³ ibid., 343; see also Herlihy, D. H., Medieval Households (1985), 2–4Google Scholar; Corbier, M., ‘Constructing Kinship in Rome: Marriage and Divorce, Filiation and Adoption’, in Kertzer, and Saller, , op. cit. (n. 1), 127–44, at 129.Google Scholar

⁴ ibid., 344.

⁵ ibid., 355.

⁶ Saller, R. P. and Shaw, B. D., ‘Tombstones and Roman family relations in the Principate: civilians, soldiers and slaves’, JRS 74 (1984), 124–56, at 145Google Scholar. See also Shaw, B. D., ‘Latin funerary epigraphy and family life in the later Roman Empire’, Historia 33 (1984), 457–97Google Scholar.

⁷ Saller and Shaw, op. cit. (n. 6), 131. I concentrate on the civilian population of Saller and Shaw's study. Their entire study is more complex and thorough than can be portrayed in this paper; they give much attention, for example, to a comparison of civilian, servile, and military commemorations and geographical differences that are not relevant for my purposes here.

⁸ ibid., 134.

⁹ ibid., 137, 145–6. It should be noted that in his recent book, Saller tends to use more images of continuum from inner to outer family than a strict dichotomy of nuclear/extended (Patriarchy, Property and Death in the Roman Family (1994), 100). I will argue in this article that a continuum rather than dichotomy should be used even more so.

¹⁰ As to the rationale for choosing these inscriptions from Tituli Asiae Minoris (hereafter TAM) rather than studying the more numerous and more recently edited inscriptions from somewhere like Ephesus, there are a few reasons. By using TAM I was able to compare inscriptions that came from several locations but were collected, organized, and edited similarly. Moreover, after reading around in different corpora of funerary inscriptions, I became convinced that a certain style of inscription, which I here call the ‘familial’ inscription, offered greater possibilities for familial analysis. The Ephesus corpus, for example contains fewer inscriptions conforming to this style. Finally, by selecting inscriptions from different places and in smaller numbers, I was able to divide the analysis into smaller, more manageable ‘chunks’, thereby making it easier for others to check my method and conclusions without having to work through the entire body of evidence.

¹¹ Although none of these inscriptions is dated, most are likely to be from the second or third century C.E., judging from archaeological evidence, names of the persons mentioned in the inscriptions, and the forms of the inscriptions. For the second and third centuries as a time of growth in the ‘epigraphic habit’ of the Roman Empire, see MacMullen, R., ‘The epigraphic habit in the Roman Empire’, American Journal of Philology 103 (1982), 233–46CrossRefGoogle Scholar. Of course, the inscriptions probably come from a wide range of dates, but we will probably not be far off by placing them in this general time.

¹² TAM II.949 (vol. 2, fasc. 3, inscription number 949). The various words I have taken to refer to ‘spouse’, ‘husband’, or ‘wife’ (σύμβιος, or often simply ἀνήρ, or γυνή) do not indicate whether or not the couple was legally married or not. Since slaves, who could not legally marry, often use the same terms as free persons to refer to their relationships, we should never take the terms to imply necessarily a legal relationship. The question is not one addressed by my organization of the data.

¹³ TAM II.1008.

¹⁴ TAM II.1000, see also 1159.

¹⁵ Even in Latin inscriptions it is often difficult to discern relations, and this even though Latin has many more precise terms for specific familial relations than Greek; for a discussion of such difficulties and possibilities in Latin inscriptions, see Corbier, op. cit. (n. 3), 130–2.

¹⁶ It must be remembered that when I give percentages in this study I am referring only to percentages of inscriptions, which is not to be taken as making any claim about percentage of the population. Numbers of inscriptions must not be simply extrapolated to represent numbers within the population. Thus my claim that 75 per cent of these inscriptions reflect extended family funerary commemorations is not intended to suggest that 75 per cent of the households of the population were extended. I do believe that a substantial number of inscriptions reflecting a certain social formation can be taken to imply a substantial number of such formations in the society, but the proportions and percentages remain unknown.

¹⁷ Although believing that geographical differences in family structures cannot account for the extent of the discrepancy in this case, I do not want to imply that regional differences are unimportant; such variables should be taken into account more often than they usually are. See, for example, the warnings by Bellemore, J. and Rawson, B., ‘ALVMNI: the Italian evidence’, ZPE 83 (1990), 1–17Google Scholar, at 2–3. Note also the findings of Bagnall, R. S. and Frier, B. W., The Demography of Roman Egypt (1994)CrossRefGoogle Scholar, indicating that families in villages tended to be larger and more complex than those in the metropoleis (49). It should be noted, however, that their findings overall agree much more closely to mine (that extended and multiple-family households were, if not in the majority, significant and numerous) than to Saller and Shaw's (see 59–60).

¹⁸ Saller and Shaw, op. cit. (n. 6), 131–2.

¹⁹ See, for example, instances of the frérèche form in Roman Egypt: Bagnall and Frier, op. cit. (n. 17), 64–6.

²⁰ It should be remembered that two occurrences of the same kind of relationship in one inscription (two sons for their father, for example) is counted as one relationship. Thus, I suppose that when the one woman provides for four persons of uncertain relationship, but who seem not to be kin, that should be counted as one non-nuclear relationship (amicus?) rather than as four (11.952). Counting all four persons, in any case, would raise the number of non-nuclear relationships in comparison to nuclear, but not so as to alter the overall picture substantially, in which the nuclear family relationships predominate. For the meaning of θρεπτός see n. 25 below.

²¹ I have noted that often a relationship is mentioned without naming the person directly, thus leaving open the possibility that the persons (usually children or descendants) do not yet exist. If we count only named persons from these ten inscriptions, or persons we can be certain exist, we come up with percentages emphasizing the nuclear family a bit less: 64 per cent (fourteen out of twenty two relationships) to 36 per cent. While this percentage is lower than the usual results given by Saller and Shaw for their inscriptions, it nevertheless shows how their method tends to exaggerate the presence of the nuclear family structure.

²² I should emphasize that I am not disagreeing with the use of the inscriptions to establish the existence of relationships (which Saller does effectively, I believe, in his Patriarchy, Property and Death in the Roman Family, 25–41Google Scholar), but only to the use of one-to-one inscriptions to say anything about the boundaries of the family.

²³ Even the traditional form of father, then mother, then child(ren), in that order, occurs in 119 of the Olympian inscriptions (well over 50 per cent of the family inscriptions), even when the inscription does not contain simply a ‘nuclear family’, thus demonstrating the ideological ubiquity of the traditional triad even in more complex familial structures.

²⁴ TAM II.947. For another case of a multi-generational extended inscription, see TAM II.1129, in which a man provides for himself, his wife, his two sons, their wives, their grandchildren (unnamed), and great-grandchildren (unnamed). In this case also, the grandchildren and great-grandchildren probably do not yet exist, but their inclusion on the tombstone demonstrates what kind of family was expected to be appropriate — and possible.

²⁵ TAM II.951 (see also 11.1105, in which three brothers provide for themselves, their respective wives, and their mother). The translation of τρόφιμος is uncertain (it could refer to their ‘nurse’), but it probably refers to something like a ‘nursling’, that is, a person taken up by persons not biologically related. If so, τρόφιμος would be basically equivalent to θρεπτός, a social role reflected many times in these inscriptions from Asia Minor. The precise social role indicated by the Greek θρεπτός, is debated by scholars. In the end, it does not seem to be equivalent to the Latin verna, which would mean a home-bred slave. Rather, it probably refers to some (unofficial or quasilegal?) practice of ‘rearing’ children not one's own, perhaps by taking in an exposed infant, buying a child with a view to rearing it in one's own family, or unofficially ‘adopting’ a child. See A. Cameron, ‘θρεπτοί and Related Terms in the Inscriptions of Asia Minor’, in W. M. Calder and J. Keil (eds), Anatolian Studies Presented to W. H. Buckler (1939), 27–62; Nani, T. G., ‘ΘΠΕΠΤΟΙ,’ Epigraphica 5–6 (1943–1944), 45–84Google Scholar. I believe John Boswell's statement that θρεπτός meant ‘slave born in the household’ is contradicted, at least for Asia Minor, by the large number of such occurrences in familial inscriptions and the way they occur in the inscriptions. I take θρεπτός to be more like the Latin alumnus than verna, and that it did not, at least in these inscriptions, reflect a legally servile role. Of course, it may have been something like a servile role, and its lower status may be reflected in the fact that such persons' names are usually given in the latter part of the familial inscription. See Boswell, J. E., ‘Expositio and oblatio: the abandonment of children and the ancient and medieval family’, AHR 89 (1984), 10–33Google ScholarPubMed, at 15 n. 8. For a survey of literary evidence on ‘exposure’ among Greeks, see Germain, L. R. F., ‘Aspects du droit d'exposition en Grèce’, Revue historique de Droit français et étranger 47 (1969), 112–20Google Scholar; see also P. Garnsey, ‘Child Rearing in Ancient Italy’, in Kertzer and Saller, op. cit. (n. 1), 48–65, at 51–6. For recent discussion of the exact meaning and status of alumni, see Rawson, B., ‘Children in the Roman Familia,’ in Rawson, (ed.), The Family in Ancient Rome: New Perspectives (1986), 170–200Google Scholar, at 173–86; Nielsen, H. S., ‘ALUMNUS: a term of relation denoting quasi-adoption,’ Classica et Mediaevalia 38 (1987), 141–88Google Scholar. It is unusual, but not unknown, for two people to be ‘co-rearers’ of one person, as may be the case in this inscription; two apparently unrelated men call one girl their alumna in a Latin inscription, see Weaver, P. R. C. and Wilkins, P. I., ‘A lost alumna’, ZPE 99 (1993), 241–4Google Scholar.

²⁶ Note terminology such as ‘larger extended family units’ (Saller and Shaw, op. cit. (n. 6), 124).

²⁷ This is also in line with other studies that emphasize the regular smallness of families, usually including no more than two sons and one daughter. See Eyben, E., ‘Family planning in Antiquity’, Ancient Society 11/12 (1981/1982), 75Google Scholar; see also Pomeroy, S. B., Goddesses, Whores, Wives, and Slaves: Women in Classical Antiquity (1975), 197–8Google Scholar; Bagnall and Frier, op. cit. (n. 17), 67–8.

²⁸ TAM II.979. For the explanation of the term θρεπτός see n. 25 above.

²⁹ TAM II.1009 and 1014.

³⁰ The numbers given for Apollonis, Magnesia/Sipylum, Hierocaesarea, and Attalia are too small to be very helpful, so I do not discuss them in the text; they are offered in the table, however, for purposes of comparison. Although they are small samples, they all offer some evidence for the healthy presence of extended family structures, well exceeding the numbers one would expect given Saller and Shaw's study.

³¹ In Saller and Shaw's reckoning, servile and freed relations are not included in extended family counts. I have included them in my extended category, but excluding those inscriptions that are categorized as extended only because of inclusion of servile or freed dependents would not substantially alter the results. The affected inscriptions would number only one or two for Termessus, one or two for Bithynia (some of the inscriptions are open to different interpretations), and two for Olympus. If θρεπτοί/αί are included as ‘servile’ relations (which they would not be if we consider them equivalent to Saller and Shaw's alumni/ae) the number of affected inscriptions grows to ten for Bithynia and six for Olympus. In other cases, servile persons are included in extended family inscriptions, but their exclusion would not mean categorizing the inscription differently.

³² I have no way of discerning if any of these people would correspond to the ‘lodgers’ identified as nonkin members of households in the Egyptian census papyri studied by Bagnall and Frier, op. cit. (n. 17) 65–6; but clearly that is one possibility.

³³ TAM III.400. In III.495 two slave women (of different owners) are buried together. In some cases, two men or two women who do not seem to be siblings share an inscription (III.303, 320, 394, 635).

³⁴ These figures are in line with studies from other times and cultures. Even in those cultures which are categorized as ‘extended family’ cultures, the actual percentage of extended family structures compared to nuclear structures is seldom a majority. For example, Tsuneo Yamane, in a study on twentieth-century Japan, notes that ‘in 1920, when the traditional family system in Japan was being enforced legally and when the lineally-extended household was preferred culturally. 55 Per cent of the families were nuclear’. In 1970 the percentage of nuclear families had risen to 63.4 per cent. (‘The Nuclear Family Within the Three Generational Household in Modern Japan’, in Beyond the Nuclear Family (1977), 79–95Google Scholar, at 80–1.) Note also the range of percentages of either extended or multiple households for various parts of Europe from the seventeenth to early nineteenth centuries. Only in two cases listed by one study do extended-multiple-family households rise over 50 per cent, and those are only 52.5 per cent and 53.2 per cent (Wachter, K. W., with Hammel, E. A. and Laslett, P. et al. , Statistical Studies of Historical Social Structure (1978), 92–3Google Scholar; note also the figures from different studies compared in Segaline, M., Historical Anthropology of the Family (1986), 23–4CrossRefGoogle Scholar, tables 1.3 and 1.4). Thus even in those societies where non-nuclear structures are quite common and even culturally preferred, they are seldom completely dominant or even the majority of house holds. For the ancient Mediterranean, therefore, we may be demanding too much to expect extended structures actually to outnumber nuclear structures. Even at only 30 per cent, they would represent an important social and cultural presence, and one quite uncommon in modern, Western experience.

³⁵ TAM IV.220; 231; 234; 238; 247; 249; 255; 260.

³⁶ In many of these inscriptions, as with those from Olympus, the provider mentions ‘children’ or ‘grand-children’, and sometimes even a σύμβιος (wife or husband) without naming them. In some cases, at least, I suspect that the person or persons may not actually (yet?) have those relations, but mentions them in the inscription due either to convention or to the expectation that such relations will materialize in the future, or both. For example, in TAM II.1128, two brothers provide an inscription for themselves and their wives and children; the wife of only one of the brothers, however, is named, which would be confusing unless the other brother does not yet have a wife. In another case, it appears that the text originally read, ‘Eutyches Daphnionos, Olympene, built the tomb for myself and my wife and children’; the name of the wife, however, seems to have been added after the words ‘my wife’ at some later time (TAM II.1036, and see editor's note). In other inscriptions, however, names of relations are not given even though they seem to have existed at the time of the erection. In TAM II.974, for example, Agathopous Hoplonos does not give the name of his wife and children, but he obviously did have a wife since he gives the name of the woman who ‘reared’ her (see also 11.1068, 1096, 1119).

³⁷ TAM IV.215; 262; 314.

³⁸ This would be the implication of Saller and Shaw's logic when they construe fewer funerary commemorations for non-nuclear family members to provide evidence of the relative unimportance of the extended family.

³⁹ See the discussion by Casey, J., The History of the Family (1989), esp. 14, 166Google Scholar, who suggests that definitions of ‘family’ should perhaps come at the end of such studies rather than at the beginning: ‘It could be argued, indeed, that “definition” is what all family history is really about, the last chapter of the book rather than the first’. See also Bagnall and Frier, op. cit. (n. 17), 57 n. 19. For an excellent recent essay on theoretical issues in historiography of the family, see Peskowitz, M., ‘ “Family/ies” in Antiquity: Evidence from Tannaitic Literature and Roman Galilean Architecture’, in Cohen, S. J. D. (ed.), The Jewish Family in Antiquity (1993), 9–36Google Scholar.

⁴⁰ Ulpian, Digest L. 16.195; see Herlihy, D., ‘House holds in the Early Middle Ages: Symmetry and Sainthood,’ in R. McC. Netting et al., Households: Comparative and Historical Studies of the Domestic Group (1984), 383–406, at 385.Google Scholar

⁴¹ Daube, D., ‘Dodges and rackets in Roman Law’, Proceedings of the Classical Association 61 (1964), 28–30Google Scholar; for wives married without being in manu: Gardner, J., Women in Roman Law and Society (1986), 76CrossRefGoogle Scholar; Clark, G., ‘Roman women’, Greece and Rome² 28 (1981), 193–212CrossRefGoogle Scholar, at 203–4; Treggiari, op. cit. (n. 1), 16–36.

⁴² For the assumption of co-residence as the defining factor in delimiting family, see, for example, Morgan, D. H. J., Social Theory and the Family (1975), 207Google Scholar; Masnick, G. and Bane, M. J. et al. , The Nation's Families: 1960–1990 (1980), 25Google Scholar; Segaline, op. cit. (n. 34), 13. These are just a few examples selected almost at random.

⁴³ See D'Arms, J. H., Romans on the Bay of Naples: A Social and Cultural Study of the Villas and Their Owners from 150 B.C. to A.D. 400 (1970), 123–33Google Scholar; Champlin, E., Fronto and Antonine Rome (1980), 21–4CrossRefGoogle Scholar; Duncan-Jones, R., The Economy of the Roman Empire: Quantitative Studies (2nd edn, 1982), 18, 22–4, 323–6Google Scholar; idem, Structure and Scale in the Roman Economy (1990), 126–9, 141–2; Boatwright, M. T., ‘Matidia the Younger’, Echos du Monde Classique/Classical Views 36(1992), 19–32Google Scholar, at 25.

⁴⁴ Rawson, B., ‘Family life among the lower classes at Rome in the first two centuries of the Empire’, CP 61 (1966), 70–83Google Scholar.

⁴⁵ Some of the slave inscriptions from Olympus are striking in their inclusion of (sometimes large) extended families. A female slave of a woman erects this one: ‘Helenous, slave of Aurelia Artemeisia, built the tomb for myself and husband and children and grandchildren and brother-in-law (?γαμβρός), Philoserapis, and my slave Melinne, and her husband, Harpokras, and their children’ (11.967). For a fuller study of slave and freed familial inscriptions throughout Asia Minor, see my Slavery as Salvation: The Metaphor of Slavery in Pauline Christianity (1990), esp. 5–6Google Scholar. The percentages there given for slave families are little different from these for the general population.

⁴⁶ See Yavetz, Z., ‘The living conditions of the urban plebs in Republican Rome’, Latomus 17 (1958), 500–17Google Scholar.

⁴⁷ In the figures for family structures given already, I have taken into account all the persons mentioned in the inscription, whether or not the providers intend to be buried there or not. But even if we count only those people who are actually expected to be buried there, we come up with rather similar proportions: thirty four cases of conjugal couples alone; about twenty one of nuclear families (if we include ‘partial’ nuclear families, that is, any nuclear family members together even if a complete ‘triad’ is not present); and about fourteen of ‘extended’ families. Again, as was the case in counting all relationships mentioned, we see the predominance of conjugal burials.

⁴⁸ See, for example, Martin, E. P. and Martin, J. M., The Black Extended Family (1978), 2Google Scholar; Masnick and Bane, op. cit. (n. 42), 25.

⁴⁹ TAMIV. 126; III.601. Others have a mother, father, daughter, and two sons; or a father, mother, and possibly three sons (IV. 151; 265). Even families as large as two parents and three children are rare in the inscriptions.

⁵⁰ This accords with the conclusions of the study of Egyptian census papyri by Bagnall and Frier, op. cit. (n. 17), 67–8, 134.

⁵¹ David Herlihy remarks about late Roman antiquity that households were too disparate to be easily categorizable by census-takers: ‘No common net could catch them all’ (op. cit. (n. 3), 5).

⁵² In one inscription, a man provides for himself, his mother, his wives (!), both of whom are listed by name, his two sons, a woman named Ammia, and Ammia's husband (11.1101). One (or both) of the wives may be dead, but there is no indication on the tombstone, and given the inclination of inscriptions to avoid the niceties of precise, legal terminology, the reader could take one (or both) of the women to be a concubine. For such situations in Rome, note Jane Gardner's comments: ‘From the city of Rome, 23 inscriptions have been found where a dead woman is commemorated by two living “husbands“’ (op. cit. (n. 41), 82). For concubines in Classical Athens sometimes in the household along with the mistress of the house, see Humphreys, S. C., The Family, Women, and Death (1983), 63–4.Google Scholar

⁵³ Scholars have often supposed that such practices were prejudicial against daughters, but direct evidence has been hard to come by, and the subject has been debated (see Gardner, op. cit. (n. 41), 6). Literary evidence suggests that it was almost a common sense of the ancient world that girls would be exposed much more than boys. See Apuleius, Metamorphoses X.23; Dionysius Halicarnassus 11.15 and XI.34.2; Stobaeus 77.7 (Poseidippos); and the famous papyrus in which a man tells his wife to rear her child if it is a son, but to expose it if it is a daughter: Oxyrhynchus Papyri (Grenfell and Hunt) IV.744. For a survey of literary evidence see Germain, L. R. F., ‘Aspects du droit d'exposition en Grèce’, Revue historique de Droit français et étranger 47 (1969), 177–97Google Scholar. Donald Engels has argued, on the basis of demographic theory, that there was no significant exposure of females in the Greco-Roman world; he argues, in particular, against the use of funerary inscriptions for such research: ‘The problem of female infanticide in the Greco Roman world’, CP 75 (1980), 112–20Google Scholar. His arguments have been countered by Harris, W. V., ‘The theoretical possibility of extensive infanticide in the Graeco Roman world’, CQ 32 (1982), 114–16CrossRefGoogle ScholarPubMed; and Golden, M., ‘Demography and the exposure of girls at Athens’, Phoenix 35 (1981), 316–31CrossRefGoogle Scholar; Emiel Eyben also rejects Engels's arguments, op. cit. (n. 27), 17, n. 44; see also Boswell, op. cit. (n. 25), 18–19. Golden agrees with Engels that funerary inscriptions are a problematic source for the study of this question. But Engels's objections are relevant only to certain kinds of epigraphic material. The nature of the evidence epigraphic material. The nature of the evidence provided by these familial tombstone inscriptions from Asia Minor, particularly the differences between them and Western inscriptions in the manner of recording names, renders his objections irrelevant for my use of these inscriptions. For supporting epigraphic evidence from Hellenistic Greece, see S. B. Pomeroy, ‘Infanticide in Hellenistic Greece’, in Cameron, A. and Kurht, A. (eds), Images of Women in Antiquity (1983), 207–22Google Scholar; Clark, op. cit. (n. 41), 195. See also the evidence from Roman Egypt: Bagnall and Frier, op. cit. (n. 17), 92, 101, 152–3. Interestingly, there is anthropological evidence from a different place and time that also shows a two to one ratio of boys to girls: see Bognár-Kutzian, I., The Copper Age Cemetery of Tiszapolgár-Basatanya (1963)Google Scholar, cited in Uck, P. J., ‘Ethnography and the archaeological interpretation of funerary remains’, World Archaeology I (1969), 262–77CrossRefGoogle Scholar, at 270.

⁵⁴ For one example of an élite woman in a rather independent position of control over her property, see Boatwright, op. cit. (n. 43), esp. 30.

⁵⁵ II.1148. The interpretation of this inscription is not certain, but I think I have got the gist of it. It is uncertain, but likely, that Neike and Artemis are daughters of Gamike, and the two males mentioned their husbands. Finally, it is unclear why Gamike would explicitly exclude any male offspring she might have, if indeed she already has them. For other female providers of extended family inscriptions: 11.1003, 1096, 1103, 1107, 1120.

⁵⁶ 11.1070; see also 1046, 1070.

⁵⁷ See also III.365. Flory, M. B., ‘Where women precede men: factors influencing the order of names in Roman epitaphs’, CJ 79 (1983), 216–24Google Scholar; Flory's study is of Roman inscriptions, but the inscriptions here examined exhibit the same phenomenon for Roman Asia Minor.

⁵⁸ For women as providers of funerary arrangements for large, extended households, see 11.1163 (a woman provides for herself, her husband, children, grandchildren, τρόφιμη, her father, mother, her second husband, who is an σἰκονόμος, of the ‘ethnos of Lycia’, a man identified as her σύντροφος, and his wife); 11.967 (a female slave of a woman for herself, husband, children, grandchildren, son-in-law, her own slave woman, and that woman's husband and children); II.984 (a woman, perhaps a slave of another woman, for self, husband, children, grandchildren, sister and another woman of uncertain relation); 11.990 (a woman for self, husband, another man she claims to have purchased, three women she has reared, her mother-in-law, and her mother-in-law's husband).

⁵⁹ See Segaline, op. cit. (n. 34), 24–7.

⁶⁰ But note that slave couples do sometimes call their children jointly their own even when legally the father had no parental claim: III.541; 815.

⁶¹ K. R. Bradley emphasizes the instability of the Roman family — due to deaths, divorces, remarriage, and questions of which household children would go with — in two articles (‘Dislocation in the Roman family’ and ‘Remarriage and the structure of the upper-class family at Rome’) now collected in his Discovering the Roman Family: Studies in Roman Social History (1901). As Bradley concludes, ‘Because of these factors, the upper-class Roman family certainly has to be regarded as a dynamic entity, but one that in its life course had little regularity of shape. It was, rather, an extremely fluid organism, subject to constant interruption, disruption, and reconstitution. It embraced both kin and nonkin members within a single household and beyond, and combined elements of nuclearity with more extensive associations of vital importance. Perhaps it is not altogether surprising, therefore, that the most common term of reference used by Romans to designate their families was not an abstract noun but simply, as Cicero so often wrote, the open-ended adjectival form, met’ (171–2). My study of the Asia Minor inscriptions, limited and impressionistic as it is, impresses me with the aptness of Bradley's description. See also Corbier, op. cit. (n. 3), 136.

⁶² For examples of others' dissatisfaction with the nuclear/extended dichotomy, see Hareven, T. K., ‘Family time and historical time’, Daedalus 106 (Spring, 1977), 57–70Google Scholar, esp. 69; Wrigley, E. A., ‘Reflections on the history of the family’, Daedalus 106 (Spring, 1977), 71–85, at 73.Google Scholar

⁶³ Wallace-Hadrill, A., ‘The social structure of the Roman house,’ PBSR 56 (1988), 43–97Google Scholar, at 46; see also his ‘Houses and Households: Sampling Pompeii and Herculaneum’, in B. Rawson (ed.), Marriage, Divorce, and Children in Ancient Rome (1991), 191–227. For an interesting study of contrasting structures and familial ideology in the Classical Greek East, see Jameson, M. H., ‘Domestic Space in the Greek City-state’, in Kent, S. (ed.), Domestic Architecture and the Use of Space: An Interdisciplinary Cross cultural Study (1990), 92–113Google Scholar.

⁶⁴ Wallace-Hadrill, op. cit. (n. 63, 1988), 58.

⁶⁵ For a similar critique, see Bradley, K., ‘Remarriage and the structure of the upper-class Roman family’, in Rawson, op. cit. (n. 63), 88–9.Google Scholar

⁶⁶ The number for this category under ‘Termessus’ includes fourteen inscriptions in which a person provided burial for him or herself and included reference to ‘his’ or ‘hers’, but without giving any more precise information about what kind of relations were meant. Since these inscriptions could not be categorized reliably by family type, they have been included in Category B.