SLAVES IN LUSITANIA: IDENTITY, DEMOGRAPHY AND SOCIAL RELATIONS

Na medida em que todas as obras da UC Digitalis se encontram protegidas pelo Código do Direito de Autor e Direitos Conexos e demais legislação aplicável, toda a cópia, parcial ou total, deste documento, nos casos em que é legalmente admitida, deverá conter ou fazer-se acompanhar por este aviso. Slaves in Lusitania: identity, demography and social relations Autor(es): Curchin, Leonard A. Publicado por: Imprensa da Universidade de Coimbra URL persistente: URI:http://hdl.handle.net/10316.2/43447 DOI: DOI:https://doi.org/10.14195/1647-8657_56_3


Introduction
The importance of slavery in the Roman world hardly requires emphasis. Possession of slaves served a dual purpose, not only to perform labour and produce wealth, but also to create leisure and prestige for their owners (Vlassopoulos, 2016, 14-15). In the province of Lusitania there are numerous inscriptions providing information on slaves 1 . However, previous studies of Lusitanian slaves have focused on slavery as an economic rather than a social reality. While no one would deny the importance of unfree labour in the ancient economy, Roman slavery "has to be approached above all as a social institution" (Bradley, 1994, 4). The discussion by Francisco Martín (1989, 187-213) is concerned almost solely with the importance of servile labour in mines, agriculture and as public slaves. An earlier book by Mangas, again concerned largely with slave labour, includes a few pages on relations between slaves and owners and on slave families (Mangas Manjarrés, 1971, 70-74, 130-131); but because of the scope of the work, covering all three Hispanic provinces as well as comparative evidence from ancient literature, he is unable to discuss particular examples from Lusitania. Moreover, both these authors rely on an outdated epigraphic corpus, including a number of inscriptions that we now know do not refer to slaves 2 ; and of course, many new inscriptions have been discovered in the intervening years 3 . The present study is an attempt to explore slave society in Lusitania, first by determining who qualifies as a slave in the epigraphic record, and then by examining the data on demography and social relations.
Before drawing any inferences from the data presented here, it is important to stress the fortuitous nature of the inscriptional evidence, which is dependent on accidents of survival and discovery and on the fact that some sites have been more intensively explored than others. Inscriptions of the lower classes, and slaves in particular, have come to light largely by chance; and many deceased slaves may not have received a stone inscription at all. Nevertheless, new inscriptions will continue to emerge through random discoveries, enlarging our meagre database. Thus, some apparent demographic and social patterns may turn out to be illusory in the light of future epigraphic finds, and some of our suppositions may require modification at a later date. tena Ulali s(erva)"); HAE 2129 = AE 1962, 320 (read "Iolumii Fab(ia tribu) Ser(ani?)", not "Volumii Fab(ii) ser(vus)"); HAE 2130 = AE 1962, 321 (read "M. Licinius Domes(t) icus", not "Eolumii domes(t)icus"); CIL 2, 5173 = IRCP 87 (read "Ploce Ser(vii) f(ilia) Catula", not "Ploce ser(va) E. Catur(ici)"). The slaves Ianuarius and December (CIL 2,5389 = HEp 4,190) and [P]rivatus (CIL II²/7,981), listed by Jiménez Losa, 1997, 743-744 as being from Augusta Emerita, belong in fact to Baetica. The inscription of the slave Chysanthus (HEp 9, 505) from Los Corrales (Sevilla) is wrongly assigned to Lusitania in the Clauss-Slaby epigraphic database. Note also that the supposed "Pullus Iul. Gracilis servus" (BRAH 36 (1900), 7 = EE 9, 79) is actually "[F]uscus [S]incerae [s]ervus" (AE 2006, 592 = HEp 15, 26; my thanks to Dr. J.C. Edmondson for information on this inscription), and "Iud(a)eus" (HAE 752) is more likely "[T]u[r]eus" (CILCáceres 2, 877 = HEp 16,146). Conimbriga, 56 (2017)   Identity: Criteria for determining slave status As Bruun (2014, 607-608) points out, identification of slaves in inscriptions is no easy task, since most persons named in Latin inscriptions neglect to indicate clearly their status as slave, freed or freeborn. Included in this study are persons explicitly designated as slaves by the status markers serva (Table 1), servus ( Table 2), verna (Table 3) or a variety of similar terms ( Table 4). The date of the inscription, when known, is indicated by a Roman numeral denoting the century A.D. 4 Not included are the following: (a) liberti. Although they were formerly slaves, freed persons would in many cases have left their owner's household and made a new life for themselves (Garnsey, 1981). While they may have married other former slaves from the same familia (as shown by the frequent occurrence of spouses with the same nomen gentilicium), they were no longer engaged in the slave experience. Their inscriptions, most often epitaphs composed posthumously, reflect their lives as freed persons, not their previous living condition as slaves. (b) persons with single Greek name. There are more than 100 such examples in Lusitanian inscriptions 5 , and while many, perhaps most, represent slaves, this cannot be assumed in every instance. Although slaves (regardless of their true nationality) were often given a Greek name, not everyone with a Greek name was servile 6 . This is evident in the Greek cognomina of some magistrates and priests, for whom free birth was a requirement 7 . Freeborn persons were sometimes given Greek names because these were euphonious, propitious or salutary (Encarnação, 2011, 307-309) 8 . A Greek name could 4 I am grateful to Dr. Jonathan Edmondson for advice on the probable dates of several of the Mérida inscriptions. Any errors are my responsibility alone. 5 Most of these can be found (intermixed with liberti and persons from the other Hispanic provinces) in Lozano Velilla, 1998, 17-233. 6 On this, see Solin, 1971, passim. 7 E.g. L(?) Iunius Philo and L. Porcius Himerus, duumviri and flamines at Salacia (IRCP 186-187); M. Cornelius Q. f. Gal. Persa, flamen provinciae Lusitaniae (IRCP 7). 8 An instructive example of the danger of drawing conclusions from onomastics is provided by the family of the philosopher and senator L. Annaeus Seneca. His younger Conimbriga, 56 (2017) 75-108 also indicate an immigrant from the eastern provinces, though these were probably few in Lusitania 9 . Finally, a single Greek name is sometimes used to designate a freedman, omitting the nomen gentilicium 10 . (c) persons with single Latin or indigenous name. Although some of these may be slaves, it is impossible to distinguish slaves from peregrines, or from the abbreviated names of freedmen, without a status indicator 11 . A freeborn citizen or freed slave of limited means might state only his cognomen (the distinctive component of the name), omitting the praenomen, nomen gentilicium and filiation or "libertination", to reduce the cost of the inscription (Bruun, 2014, 608). Moreover, in a closed environment such as a workshop, where everyone knew everyone else, it was only necessary to refer to people by a single name. Thus in graffiti recording brick production at Conimbriga and Emerita, we find the workers identified by a single name, such as Agilio, Caletus, Nigellio or Hilarus. While some of these could be slaves, this cannot be taken for brother (an eques Romanus) had the "servile" Greek name Mela, while his older brother was Novatus, a name which according to Lefebvre, 2006, 120 was particularly used by Africans. The son of Mela received the name Lucanus, which should mean a native of Lucania (Kajanto, 1965, 193), whereas he was born in Corduba. In fact, Lucanus might have originally been applied to someone in or from Lucania, but it soon lost its geographical connotation and became a normal Roman cognomen. 9 Rare examples include a Cappadocian at Olisipo (CIL 2, 224), and a Syrian and Phrygian at Emerita (CIL 2, 515 = ERAE 186 = HEp 4, 163;AE 1989, 395 = HEp 2, 39). See Haley, 1991, 38-42. Quintilian, Inst. 1,1,12 recommends that a boy should have a Greek-speaking nurse or paedagogus so that he may learn the language from an early age; but this would only be an important consideration in élite families. 10 E.g. Chresumus Longini lib. (HAE 1112= AE 1967; Hegemon (erected by his conliberti: HEp 2, 38; cf. Edmondson, Nogales and Trillmich, 2001, 138) granted, because we also find the praenomen Titus and the nomen Iulia 12 . While certain names such as Felix, Primus and Ianuarius are particularly common among persons of servile origin (Thylander, 1952, 150;Jiménez Losa, 1997, 745), their use does not necessarily mean that a given individual is a slave rather than a freedman or even an ingenuus 13 . The name Vernaculus, derived from verna, theoretically indicates someone of servile birth, yet it can also be borne by freeborn peregrines 14 . It should also be remembered that the freedmen of peregrines would themselves be peregrines and, like them, entitled to bear only a single name instead of tria nomina 15 . It is noteworthy that several of the slave owners listed in Tables  1-4 have a single Latin name, but this does not mean that they also are slaves. In most cases they are probably either peregrines, or Roman citizens who list only their cognomen in the inscription.
(d) persons of the type "Lupus Vegeti" or "Ianuarius Venusti", where it is unclear whether to understand "(servus)" or "(filius)" after the genitive 16 . As indicated above, a single name is no guarantee that the person is a slave. Lack of status indication does not necessarily denote slave status by default. (e) persons designated "contubernalis". Although this term can refer to a slave partner, it can also be used of any quasi-marital union (Rawson, 1974, 293-299) 17 .
A particularly misleading fallacy, reflected in the slave lists of Mangas, Francisco Martín and Jiménez Losa, is the assumption that persons designated by single name (whether Greek, Latin or indigenous) are slaves. While many of these may indeed be servile, this cannot be taken for granted 18 . Peregrines often had only one name, which could be (but was not always) supplemented with a filiation in the inscription. Even in the case of persons with duo or tria nomina, the nomen gentilicium was often omitted to save space on the stone, especially if it could be supplied from the names of other persons in the inscription (Thylander, 1952, 98). Hence, when a freedman appears to have a mononymous wife or children, it should not be automatically assumed that these are slaves. In the absence of status markers such as "ser(va)" or "lib(ertus)", it is notoriously difficult to determine the status of lower-class individuals named in inscriptions.
The application of the stringent criteria outlined above may seem unduly positivistic, resulting in the exclusion of many possible or even probable slaves. However, this minimalist approach is the only way to guarantee that we have a secure sample of persons who are definitely slaves, uncontaminated by questionable examples. It leaves us with a large sample of genuine slaves that yields interesting information and allows useful conclusions to be drawn.

Demography
Of the 99 slaves catalogued in the tables, two-thirds (66) belong to the conventus Emeritensis (including 31 in the provincial capital, are freedmen. In CIL 2, 561 = ERAE 315, Irene Valentis f(ilia), contubernalis of L. Fuscianus is possibly a slave, as suggested by her Greek name, though the nomenclature of slaves does not usually include filiation (but cf. "Caesia Crusten(i) f(ilia) Severi ser(va)" in AE 1977, 394 = CILCáceres 2, 785 and "Vitalis Messi Sympaeronti f(ilius) et servus" in IRCP 536). Perhaps she is rather the daughter of a freedman. In FC 2, 40, Atimetus, who erects a memorial to an unnamed contubernalis, could also be a slave. In CIL 2, 5331 = IRPToledo 178, Primilla, contubernalis of C. Accius Diodatus, is probably servile; but since the Greek name of her partner suggests that he is a freedman, we should perhaps understand "(Accia) Primilla". Other mentions of contubernales (CILCáceres 2, 669 and 772) are too fragmentary for analysis.
18 "un nombre único podría ser indicio de... condición peregrina o incluso servil": Abascal and Alföldy, 2015, 112. Conimbriga, 56 (2017)   Emerita Augusta). Only 17 slaves are recorded in the conventus Pacensis, and 16 in the conventus Scallabitanus 19 . Mantas (2003, 51) accounts for the relatively small number of epigraphically attested slaves and ex-slaves in the conventus Pacensis by arguing that slaves in this district were employed mostly as miners (with short life spans and little chance of leaving a memorial) and that free, salaried workers predominated over slave labour. This does not explain why there are so few domestic slaves, the group most likely to receive memorials, in Pacensis and Scallabitanus. A more compelling reason for the apparent numerical disparity among the three conventus is that Emeritensis, with a much larger territory, embraced a much higher number of persons in all social classes. An additional consideration is that, with its proximity to the very romanized province of Baetica, Emeritensis had a more developed epigraphic culture, and thus more inscriptions of both slaves and non-slaves, than the other two conventus. But even taking into consideration the smaller population of Pacensis and Scallabitanus, and the fact that not all slaves could afford inscriptions (and of course, that not all inscriptions survive), the slave figures for these two conventus seem minuscule. This paucity of attested slaves throws into question the assumption that every free man owned at least one slave 20 . While the more affluent members of Lusitanian society may have owned large numbers of slaves, there must have been many people (peasants and townsmen alike) of limited means who could not afford to purchase or maintain them.
This problem can be considered more closely if we look at specific examples. Encarnação (1987, 68-69) pointed out that there are no explicit mentions of servi on the south coast of the Algarve, although Speratus, dispensator at Balsa, is undoubtedly a public slave. Does this mean that there was a dearth of private slaves in the district, or does their absence reflect accidents of survival of slave inscriptions or a disinclination of slaves to leave memorials? The last of these explanations may apply in the Civitas Igaeditanorum (Idanha-a-Velha), where epigraphy records only one slave (delicata), yet the prevalence of slavery is revealed by 23 inscriptions of ex-slaves (liberti). In the corpus of 175 19 Idanha-a-Velha, which CIL included in Scallabitanus, is nowadays assigned to Emeritensis. Serpa is considered to belong probably to Pacensis. See Curchin, 2000b, 330. 20 "todo hombre libre tenía, al menos, uno" (Entrena Klett, 1983, 32).
inscriptions from the province of Ávila there is no mention of slaves, though there are at least two liberti (Hernando Sobrino, 2005, n o . 48, 62, cf. 20). And among the more than 1300 inscriptions from the territory of Coria (Cáceres) there are no slaves, but there are five freed persons (CIL Cáceres 4, n o . 1157, 1180, 1184, 1200, 1324. Obviously freedmen were more likely than slaves to be able to afford an inscribed tombstone. Of course, some of the persons with single name in these corpora could also be slaves. The proportion of slaves in the Lusitanian population cannot, in the absence of statistics, be reliably calculated. They are surely underrepresented in the epigraphic corpus, with servile epitaphs comprising only a minute fraction of the roughly 6500 Roman inscriptions from Lusitania. Instructive comparison can be made with other regions of the Roman world. Estimates of the proportion of slaves in the population of Italy during the Early Empire range from 13 to 43 percent (Launaro, 2011, 62;Scheidel, 2011, 291), a spread that does not inspire confidence in the methodology. Census figures for Egypt suggest that slaves made up 11 percent of the population (Bagnall and Frier, 2006, 70). In other provinces, the proportion of slaves may have been between 10 and 20 percent (Scheidel, 1997, 158;Harris, 1999, 65), though this is only a guess. It would therefore not be unreasonable to suppose that slaves comprised 10 to 20 percent of the population of Lusitania.
The attested slaves include 33 females and 62 males. The fact that male slaves outnumber females by a ratio of almost 2:1 is less probably due to a greater tendency to commemorate men 21 , than to a numerical preponderance of males over females in the household 22 . This is readily understandable in view of the need for male labour to undertake difficult physical tasks, though it appears to contradict a recent statement that female slaves in Hispania were nearly always in the majority 23 . Since the jobs available for female slaves in a Roman household were limited, their numbers will accordingly be smaller (Treggiari, 1975b, 58), even though for purposes of reproduction a fairly equal sex ratio would be preferable (Ste. Croix, 1981, 231). Scheidel (1999: 137) argues that, despite the assumption that male slaves were more useful than women for farming, mining, manufacturing and construction, it would have been impossible to maintain an Italian slave population of (say) two million unless there was a huge number of female slaves producing children. Harper (2011, 70) points out that even if, as "cannot be demonstrated", imported slaves were mostly male, slave births would have levelled out the gender balance within one generation. In addition to bearing offspring, large numbers of female slaves may have been needed in such industries as clothing production (Hin, 2013, 36-37).
While a higher ratio of men to women could be achieved simply by purchasing more male slaves, an owner could manipulate the relative numbers of male and female vernae by exposing unwanted girls at birth. Although Harris (1999, 63, 70) claims "powerful evidence" that both imported and home-born slaves were predominantly male, and that the number of girls could be reduced through exposure, Scheidel (2005, 72) finds "no good reason" to suppose that the slave trade favoured males, while Boswell (1988: 100-103) sees little evidence that females were especially liable to exposure. If girls (both slave and free) were more likely than boys to be abandoned, and those found were usually raised as slaves, either for one's own use or to sell to others (Hin, 2013, 140), exposure would actually have increased the proportion of female slaves (Harper, 2011, 70). Treggiari (1975a, 400-401), expressing concern that selling or exposing girls would have a deleterious effect on the morale of the other slaves, suggests instead that girl slaves could be sent to work on the owner's rural estates 24 . However, slave morale was probably less important than the balance sheet to most owners; and a surfeit of female children was, at least in the short term, a financial liability.
Other factors that might explain the smaller number of women include a higher female death rate (assuming that male slaves were in general stronger and more vigorous, and including deaths of women during childbirth), and a greater likelihood of female slaves -mostly employed inside the house, where they were likely to attract the attention of the master or mistress-being manumitted. Indeed the inscriptions show a higher female-male ratio (about 3:4) among Lusitanian freed persons than among slaves (about 1:2) 25 . None the less, it is not apparent how manumission would have resulted in a net reduction in the number of female slaves, because unless these women continued to perform the same tasks as libertae, they would presumably have been replaced as workers by new female slaves.
It has been pointed out that a female slave may have had less chance of being manumitted than a male, since the work of female slaves was menial and unremunerative, and they thus had little opportunity to accumulate a peculium (Flory, 1984, 218). However, manumission could also be granted in recognition of good service (Treggiari, 1969, 15). One possibility is that women were, on average, manumitted at an earlier age than men, perhaps as a reward for bearing several children. Although, as Bradley (1987, 55) points out, no ancient texts calculate the worth of a female slave in terms of the monetary value of the children she produced, it is conceivable that this value was taken into consideration at the time of manumission, as equivalent to the peculium normally paid by slaves for their freedom. For instance, Columella (1,8,19) gave freedom to female slaves who produced at least four children, apparently regardless of how young the mothers were; obviously the value of four new slaves was considered worth the mother's liberflocks on farms. Presumably young girls could also do tasks such as working wool (under the supervision of the vilica: Columella 12,3,6) and picking fruits and vegetables (cf. Bradley, 1991, 114). When older, they might breed new rural slaves. 25 A search of the Clauss-Slaby epigraphic database in April 2016 found 126 libertae and 164 liberti in Lusitania. Disappointingly, neither Serrano Delgado (1988) nor Hernández Guerra (2013) discusses the proportion of women among Hispanic freed persons, nor the actual (as opposed to legal) age at which they may have been manumitted. Conimbriga, 56 (2017) 75-108 ation. Such early manumissions would have transferred more women from servile to libertine rank. However, the opposite view is taken in a recent paper by Lavan (2016, 13), who argues for a significantly later age of manumission for women, particularly those of childbearing age. He adopts the a priori reasoning that owners had an obvious incentive to ensure that the children of these women were born as slaves, adding that "the available evidence suggests that the average age at manumission was significantly later for women." Unfortunately, inscriptions do not record the age at which a person was manumitted, though the age at death does provide a terminus ante quem for the date of manumission. In Lusitania there are 16 libertae who died between the ages of 14 and 40 inclusive, versus 12 liberti; the numbers are small, but suggest that women of childbearing age were not lagging behind in their rate of manumission. Also, Lavan's assumption that owners kept these women in slavery in order to produce more slaves is open to question. Osiek (2003, 259) has challenged the whole concept of Roman "slave breeding", which envisions large numbers of female slaves bought and kept primarily for the purpose of childbearing. She argues that this practice, for which there is no ancient evidence, would have been very expensive, time-consuming, and not particularly profitable. What the sources do show is that slaves who had partners were encouraged to produce children, who as slaves automatically became the property of the owner. Large slave families, argues Osiek, must have been unusual; otherwise it would not have been necessary for an owner such as Columella to offer manumission as an incentive for having more than three children.
Slave inscriptions are mostly short. In many cases, the only information provided is the name of the slave, the name of the owner, and sometimes the slave's age at death. Of the female slaves, 13 have a Greek name, 13 a Latin name, and four an indigenous or foreign name. Of the male slaves, 17 have a Greek name, 37 a Latin name, and 5 an indigenous or foreign name. It is not clear why there is a higher proportion of Latin names among the men. The names given to slaves are sometimes revealing: they include names of good omen such as Felix, Fortunata and Euticus (eutychés), as well as epithets referring to qualities desirable in a good slave (Amoenus, Blandus, Euterpe, Festiva, Gratus, Iucunda, Modestus). The name Primogenes or Primigenia signifies "first-born". Those slaves receiving indigenous names such as Cilea, Medigenius and Corocuta (reminiscent of the notorious bandit Conimbriga, 56 (2017) 75-108 Corocotta) were presumably born in the home and given names current in the region 26 . The presence of servi with Lusitanian names makes it difficult to distinguish slaves from peregrines unless there is either a status marker or a filiation. For a name such as Abimp [---] or Uprila (if the published readings are correct), no satisfactory etymology has been proposed.
The outdated theory that slaves with Greek names are either from the Eastern provinces, or descended from Easterners (Frank, 1916, 700;Thylander, 1952, 159), is no longer credited (see Solin, 1971, 146-158). Slaves with Greek names could have received them from slave dealers, from slave owners, from the steward (dispensator) or overseer (vilicus), or in some cases from the parents (such as Amoenus son of Amoena, Holumpus son of Holumpus, or Herennius son of Herennia) 27 , though possibly it was the owner who gave the slave the parent's name (cf. Bruun, 2013, 32). When it is reported to Trimalchio that 70 slave children have been born in one day on one of his estates 28 , there is no suggestion that he is going to name them all. Although this example is fictitious and exaggerated for comic effect, the point is that a busy owner had more important affairs to concern him than choosing baby names for all his vernae. However, this may have been practicable in a small household. The examples of the vernae Euhodia and Mistiche and the delicium Apolausis show that slaves born at home, not only those purchased at the market, could be given Greek names. Indeed, a study of the nomenclature of 480 vernae throughout the Roman world determined that 37,5 percent had Greek names (Herrmann-Otto, 1994, 414-417). In addition to naming his own vernae, the owner might give slaves purchased at the market a name different from the one assigned to them by the dealer.
Further information about the demography of slaves can be gleaned from their ages of death, although the sample is not large enough to calculate life expectancy, particularly since the majority of slave inscriptions do not provide an age at death. Life expectancy for slaves is in any 26  event controversial 29 . While there was a high rate of infant mortality in the ancient population generally, because of inadequate medical knowledge and the absence of antibiotics, one should expect an even higher rate among slaves as a result of substandard living conditions and arduous labour. Of course, since slaves represented a financial investment, it was in the owner's interest to keep them healthy; a large household would even include a slave medicus. Nevertheless, out of 24 females and 25 males whose age is recorded, 8 of the girls and 7 of the boys died before age 10 30 . A further 7 boys died between age 10 and 18, versus only two girls, both dead at age 15. Only 7 slaves (5 male, 2 female) lived beyond age 30. However, it must be remembered that those who died young are more likely to have had their age emphasized than those who died as mature adults (MacMullen, 1982, 239-240;Rawson, 2003, 350;Mouritsen, 2013, 50); hence a disproportionate number of young people may be represented in the inscriptions. But more importantly, the shortage of attestations of older slaves may be largely attributable to manumission rather than to early death; the epitaphs of those so freed would list them as liberti, not servi 31 . To obtain a more realistic sample of ages at death among persons of servile origin, we would have to include the epitaphs of freedmen. In Lusitania there are 33 preserved ages of liberti, ranging from 3 to 120(!) years, with a mean age of 45; and 28 preserved ages of libertae, ranging from 14 to 100, with a mean of 42. In contrast to the low figures for slaves, 20 liberti and 16 libertae, i.e. more than half the total, lived beyond age 30, and several (9 liberti and 7 libertae) lived to 60 or beyond 32 . Naturally we can reasonably expect manumitted slaves (who were exempt from forced labour and harsh treatment and presumably had higher morale in consequence) to have a longer life expectancy than those unfortunates who remained in 29 Duncan-Jones (1990, 100-104) and Harris (1999, 71) estimate the mean life expectancy of Roman slaves at birth to be 20 years or less. However, Scheidel (2001, 29) argues that the assumption that Roman slaves were short-lived, while intuitively plausible, is impossible to prove. 30 At Rome, by comparison, far more home-born slaves died before the age of 10 than at all other ages combined (Rawson, 1986, 188-189). 31 As Harper (1972, 342) points out, Roman inscriptions record few young freedmen, yet the slave population appears young "because manumissions progressively reduced its numbers as age advanced." 32 I draw my evidence from the Clauss-Slaby epigraphic database. slavery 33 . However, since only a minority of epitaphs of both slaves and freedmen provide an age at death, the data are insufficient to yield a reliable estimate of average life expectancy for either group. The age at death can also provide information on literacy. Illiterate people tend to have an imprecise knowledge of their true age, and therefore "round off" their age. Statistically, 20 percent of the population should have an age ending in "5" or "0". Of 22 female slaves of recorded age (excluding two infants who died at less than age 1), 13 or 59 percent had ages divisible by five. However, of 24 male slaves (excluding one example where the ending of the age is broken), only 6 or 25 percent had rounded ages. This does not necessarily mean that female slaves were more illiterate than males, since in more than half the cases it was a man who set up, and presumably composed or provided the information for, the epitaph. In dealing with larger population groups, it is possible to construct an age-rounding index, based on those who are recorded as dying between ages 23 and 62 34 . Unfortunately, since 30 of our slaves died before the age of 23, this does not leave a sufficient sample to construct a meaningful age-rounding index.

Social Relations
The examples listed in Tables 1-4 include two imperial slaves (Germanus and Ianuarius) 35 , an Augusti verna (Firmus) who was vilicus of the five percent inheritance tax (XX hereditatium), the slave of an aedile of Pax Iulia (Modestus), a dispensator (Speratus) who was a public slave at Balsa, another public slave (Herennius) at Emerita, and four public conservi at Eburobrittium (Óbidos). There is also mention of imperial slaves in the service of the procurator, in the mining regulations 33 Empire-wide, the Clauss-Slaby epigraphic database records hundreds of epitaphs of freedmen who reached age 70 or beyond, versus only a few dozen slaves. However, this disparity could be due in part to libertine families being better able to afford an inscription. 34 Such as I have done for the Lusitanian population as a whole: Curchin, 2000a, 280-284. 35 "Ianuarii Caes(aris) n(ostri)" can only be a Caesaris (servus); if he were a freedman, he would be designated Augusti (libertus).
from Vipasca 36 . Conspicuous among the many private owners is L. Arruntius Stella, apparently a Trajanic senator 37 . The slave Quintio attested at Emerita was the slave of Flavius Baeticus of Conimbriga. Presumably he conducted his owner's affairs at Emerita, and the Coutius Lupus for whose health he dedicates an altar was perhaps a business associate (Encarnação, 2016, 254). Some slaves have more than one owner: "Corocuta Tutilio[rum] Pontiani et Luperci ser(vus)", Phoebus Vibiorum [Po]pillorum Iuventii et Primulae ser(vus)" and "Quintilla M. Curi(i) Quintionis et Curiae Primulae delicata" 38 . In the last two cases the joint owners may be brother and sister; contrary to modern practice, a Roman wife did not assume her husband's surname. Three slaves are identified as vicaria/us, slave of another slave. Since nearly all the slaves in Tables 1-4 name their owner, even if he is a slave, as part of their own onomastic formula, it does not follow, as claimed by one investigator (Encarnação, 2013, 209), that the naming of the master shows that he was an important personage because otherwise his name would have been omitted. Clearly such omission was the rule rather than the exception.
The jobs performed by slaves are rarely specified. Speratus was a public steward (dispensator) at Balsa. Hermes, slave of Aurelia Vibia Sabina, was a marble worker (marmorarius), a slave job paralleled in the household of the Statilii at Rome 39 . His mistress bears two senatorial (though non-patrician) nomina, suggesting that she may have held a high place in local society (Encarnação, 2008, 64). As for personal servants, there is no mention of a pedisequus, paedagogus or ornatrix; and the two attested wet-nurses (nutrices) are of free status, but may have begun as slaves 40 . Wet-nurses might look after not only the owner's 36 CIL 2, 5181 = ILS 6891. 37 Consul in A.D. 102 (PIR² A 1151). He might have been a provincial governor, though he is not otherwise attested in Lusitania. However, senators at Rome often had slaves and freedmen as business agents in the provinces (D'Arms, 1981, 44-45); cf. Digest 40,9,10 ("homines...qui transmarinas negotiationes...per servos atque libertos exercent"). On slaves as business agents in Hispania, cf. Sánchez León, 1978, 263-264. 38 Cf. CIL 2, 773 (Avita, ex-slave of Silo and Vegetus). 39 CIL 6, 6318. 40 CIL 2, 545 = ERAE 249 (freedwoman, but presumably began wet-nursing as a slave); AE 1960, 190 (epitaph set up by Proculinus Proculi (f.) to two wives, Valeria and Amabilis, the second of whom was nutrix to his children). Possibly Amabilis had children, but also those of the slaves 41 . The slave Atimetus was personal physician (medicus) to M. Iulius Rufinus at Emerita, and had a student (discens) Nothus, slave of C. Heius Primus. Nothus was later manumitted, as we learn from an inscription at Olisipo 42 . The Vipasca mining regulations mention domestic slaves who perform barbering services for their owner or for their fellow slaves, as well as slaves engaged in smelting copper or silver 43 .
The religious activity of slaves has generally been neglected (Hodkinson and Geary, 2012, 3-7), yet religion pervades social life, providing cohesion and structure (Durkheim, 1958, 160). In addition to fostering a sense of fellowship and solidarity among the worshippers of each cult, religion furnished slaves with a certain degree of hope and comfort in their oppressed lives. Another possible reason for slaves' attraction to religion was that "few other areas of life were open to them" (Veyne, 1987, 62). In Lusitania, religious dedications were set up by male slaves (none by females) 44 . At least six of these are to indigenous deities: three to Endovellicus, two to Ataecina, and one to Nabia (syncretized as Augusta Nabia) 45 . One slave dedicated to the health goddess Salus, on behalf of his (sick?) master 46 . There is also one dedication each to Mars, Fontanus, and Invictus Deus 47 . The theonym Fontanus, been a slave nurse who was manumitted and married the master after his first wife died, but this is not clear from the inscription. On free wet-nurses see Bradley, 1984, 71. On the status of Hispanic nutrices, see Crespo Ortiz, 2002. I omit mention of a presumably servile doorkeeper (ostiarius) in a Christian inscription from Mértola (Vives, 1969, no. 489 Although the name is presumably Celtic (related to nemeton), the inscription does not include any religious formulas such as "sacrum", "ex voto" or "v.s.l.m." It therefore seems likely that Nemetius is a personal name rather than a theonym. connoting a spring or fountain, occurs elsewhere in Lusitania and once in Baetica 48 ; in Italy and Hispania Citerior we instead find dedications to Fons. Although the title Invictus Deus is sometimes used of Sol or Serapis, it most commonly refers to Mithra.
Much can be learned about slave families from the inscriptions 49 . Several of the slaves (Tables 1-3) were buried by their mother or father, or by both parents. Where only one parent is the commemorator, we appear to have a single-parent family. This could be the result of the other parent dying or being sold. The single parent is most often the mother; this is understandable because, if a slave couple was broken up, the children would remain with their mother. However, it is also possible that the child's father was sometimes the mother's master (or even her mistress's husband) 50 . In some instances, both parents appear to be slaves: for example, Euhodus and Callityche, who set up an epitaph to their teenage daughter Euhodia, slave of Mellinus (Edmondson, 2011, 347) 51 . In other cases, the child was still in slavery while the parent had been freed 52 . Partenis, slave of Accia Marciana, is commemorated by her mother Iulia Fortunata; presumably the mother was freed by the mistress' husband, a Iulius. This manumission must have occurred after the birth of Partenis, since a newborn takes its mother's status. Whether Fortunata's mate was a slave or a freedman cannot be known. Callaecio was buried by his parents, who were both libertine, and Corocuta was interred by his mother who was a freedwoman. Herennius, a public slave, is commemorated by his mother, Lucceia Herennia. Vitalis 48 CIL 2, 150 = IRCP 438;HEp 7, 124;AE 2004, 740 = HEp 13, 84. 49 Cf. for the city of Rome: Flory, 1978. 50 On tension between spouses over the husband's sexual relations with slaves, see Saller, 1987, 78-79. 51 Crespo Ortiz (2003 reads the owner's name as Mellinius, a known nomen gentilicium. However, Solin and Salomies (1988, 361) recognize Mellinus as a cognomen in this inscription. 52 In AE 1967, 189 = ERAE 468 ("Aphrodite an(nis) [---], Mallia Macaria matri piissimae f(ecit)") we appear to have the opposite situation, where the daughter has been freed and the mother is still a slave: see commentary in Edmondson, Nogales and Trillmich, 2001, 126. But it is possible that the mother is also a freedwoman and her nomen gentilicium omitted as redundant, since it would be the same as Macaria's. Omission reduces the length and cost of an inscription, but creates ambiguity. If the text had been worded "Malliae Aphrodite an. [---], Macaria matri piissimae f.", we might have concluded that the daughter was still in slavery! is said to be the son of the freedman Messius Sympaeron. The formula "Messi Sympaerontis f(ilius) et servus" does not mean that he was both the son and slave of his father, but rather the slave of Sympaeron's former master, also named Messius (Encarnação, 2008, 64-65). The slaves Amoenus and Ruga, in different towns, belong to families of Iulii and coincidentally have mothers named Amoena: in both cases, is she a slave, or a (Iulia) Amoena with nomen omitted? The verna Faustinus was buried by L. Sempronius Faustus, who is apparently his father as the similarity of names suggests. That the child remained in slavery after the parent had been freed is not surprising. Under the Lex Aelia Sentia of A.D. 4, slaves could not be formally manumitted before the age of thirty, though there is ample epigraphic evidence for children being manumitted informally 53 . However, Crespo Ortiz (2003, 129) raises a different possibility: instead of being a freedman, L. Sempronius Faustus may be the owner as well as the father of this child. Masters had the legal right to have sex with their slaves, and any resultant offspring would follow the mother's status.
The relation of the dedicant to the deceased is not always stated, but can sometimes be postulated as parental. Myrinus, age 12, is commemorated by Claudia Marciane, who could be his mother. Hispallus, age 3, is buried by Euhodus, possibly his father. Vicarius, age 3, receives a dedication from Cornelia Corinthia Anna, who is probably his mother. His name suggests that he was born the slave of a slave; therefore his owner, Iuventius Vitalis, may be a recently manumitted slave. Lucius Sempronius Faustus, who erects a monument to the verna Faustinus, is presumably his father (a freedman?); the mother must be a slave, since children follow their mother's status. In the absence of parents, slaves might be buried by their siblings. Caesia was buried by her brothers (fratres), which suggests her parents were already deceased. Gratus is said to be the brother of L. Iulius Secundus, a money-changer (nummularius).
Slaves frequently entered quasi-marital unions (contubernia) with mates of the opposite sex. Such relationships were permitted and indeed encouraged by owners, not only to improve morale but also to produce 53 On informal manumission of children, see Weaver, 2001, 103;cf more slaves (Treggiari, 1975a, 396). Those engaged in such relationships sometimes adopted the terminology of legal marriage (Bradley, 1984, 49). Although a slave could not be a husband or wife under Roman law, such a relationship was permissible under the ius gentium. Catulla is called uxor merens by her partner Primitivus, while Felicia is buried by her anonymous vir p(ius). More often we have to guess at the relationship. The ancilla Maura is buried by one Saturninus, and the verna Salvius buries Amoena, daughter of Caenicus: are these contubernales? Phoebillus, age 27, is commemorated by Venusta: is this his mother or a contubernalis? A similar question surrounds the verna Felicio, age 18, named on a stone set up by Patricia uxor. Crespo Ortiz (2003, 58) assumes that she is Felicio's wife. However, the top of the inscription is missing, and the phrase "[P]atricia uxor f(aciendum) c(uravit)", which by epigraphic custom should come after the name of the deceased, precedes the mention of Felicio 54 . It is therefore reasonable to assume that Patricia's dead contubernalis was named in the lost part of the inscription, and that their son Felicio died later and was added to the epitaph. Another inscription records the deaths of T. Marius Secundus, age 40, Leuche, slave of Atilia Tertulla, age 24, and the verna Mistiche, age 6 months. It is tempting to see Leuche as the mother of Mistiche, and Marius Secundus (perhaps a freedman of Atilia's husband) as the father. Vernacla, slave of Trebia Musa, is commemorated by Q. Licinius Catullus, possibly a freedman and her contubernalis. Contubernium between slaves and freedmen in the Roman world was commonplace, in spite of their disparity of status, as "there was always the possibility that the slave partner in a union could attain the status of the other partner" (Rawson, 1966, 72). However, there is no example in Lusitania of contubernium between a freedwoman and a male slave. Edmondson (2004, 350) has remarked that owners at Emerita erected an abnormally large number of epitaphs to their slaves. Specifically, he identifies four commemorations by master to slave (Edmondson, 2000, 308). This is a relatively large number, since Saller and Shaw (1984, 148-151) found only one example of owner-slave commemoration in the epigraphy of Hispania, and only two examples 54 Museo Arqueológico Nacional, inv. 34473; photograph viewable on the Centro CIL II website.
(out of 119 slave inscriptions) in the province of Noricum 55 . However, Edmondson's four examples, while plausible, are not completely certain 56 . Indeed, I cannot find any Lusitanian inscription unequivocally set up to a slave by his dominus. However, there are numerous epitaphs of the type "X, slave of Y, lived z years", followed by the closing formula "h.s.e. s.t.t.l.", with no clear identification of the dedicant. Can we assume that the owner Y was the one who erected the tombstone, and that he cared enough about his slave to have it inscribed? Or was it erected by the fellow slaves of X from their peculium, omitting their names to reduce the cost of the inscription? This raises the interesting question of whether Lusitanian slave owners were humanitarian like Pliny the Younger, who allowed his slaves to make a will, or cruel like P. Vedius Pollio, who threw slaves that offended him into a pool of lampreys 57 . In most cases, of course, the truth will lie somewhere between these two extremes, though some owners may have been kinder or crueller than others. Since some owners manumitted slaves out of affection, it is reasonable to infer that some would have commemorated slaves for the same reason. But affection was a less common motive for manumission than profit (Hopkins, 1978, 127-128;cf. Badian, 1982, 55 Mangas, 1971, 70 cites three epitaphs erected by owners to their slaves, all from Tarraconensis. 56 Prof. Edmondson has kindly informed me of the four commemorations in question: Salvianus to Fortunata, "ser(va) fidelissima conservatrix et amatrix domini" (EE 8,43); Claudia Marciane to Myrinus, age 12 (ERAE 334); L. Sempronius Faustus to the verna Faustinus, age 3 (CIL 2, 5269); epitaph of the verna Felicio, age 18, possibly dedicated by Patricia who commemorates her dead husband (name lost) in the same inscription (HAE 697). Edmondson reasons, perhaps rightly, that the intimate terms describing Fortunata's relationship with her master make him the likely dedicant (though the single name Salvianus seems more appropriate to a fellow-slave; cf. AE 1966, 194). Claudia Marciane could be the owner of Myrinus, or she might be his mother, manumitted after his birth. Similarly, L. Sempronius Faustus could be the owner of the house-born slave Faustinus, who bears a diminutive form of his name, or again we could have a libertine parent. It is also possible that Faustinus is an illegitimate child of his master (but would the master advertise the slave's paternity in this way?). The relationship of Patricia to Felicio remains uncertain; if she is the dedicant, she may be his mother. The fact that a slave is labelled "verna" does not necessarily mean that the commemorator is the slave's owner. 57 Plin. Ep. 8,16. Plin. Nat. hist. 9,39;Seneca De ira 3,40;Dio Cass. 54,23,2. 168), and there was no financial gain in burying slaves 58 . Even if we discount ancient characterizations of slaves as tools (instrumenta) or possessions (res) rather than persons, they were clearly regarded as inferior beings (Veyne, 1987, 51-52) and were liable to be beaten for the smallest offences 59 . It is doubtful whether many owners would have incurred the trouble and expense of having a tombstone engraved for them. Providing some kind of funeral honours for the deceased might have been good for the morale of the other slaves, but from the owner's perspective, constant fear was more effective than good morale in motivating slaves to work harder. It is also doubtful whether the owner would have permitted the bodies or ashes of his slaves to be deposited in the same vault or mausoleum as his own family, or whether they would have been consigned to the Roman equivalent of the potter's field. It is noteworthy that the monumental tomb of the Voconii at Mérida contained only four cinerary urns, corresponding to the four members of the nuclear family named on the accompanying inscription (Bendala Galán, 1972, 229 and Fig. 2); there were no urns or epitaphs for their slaves. Admittedly, a domestic slave was far more likely to attract the attention of the owner than one working in the fields, and hence (if the attention was favourable) might have a better chance of being commemorated by him in a sepulchral inscription. Yet "proximity also put slaves in the urban household closer to the slave-holder's whip" (Joshel, 2010, 191;cf. Saller, 1991, 158-159), and slaves whose performance was unsatisfactory or marginal could hardly expect the owner to provide them with a memorial when they died. In my view, those who are most likely to have ordered and paid for slaves' tombstones were their conservi and 58 Owners profited from manumission in several ways: (1) They benefited from increased productivity by slaves who worked harder in the hope of manumission. (2) Slaves had to purchase their freedom by surrendering their savings (peculium) to the owner. The amount varied, but in some cases was sufficient to purchase a replacement slave. (3) Freedmen continued to owe work or services (operae) to their former owner. (4) In the case of slaves freed informally, any property amassed by the freedman reverted at his death to his former owner; and formally freed slaves who were childless had to leave him half their estate (Gaius,Inst. 3,41). 59 A rescript of Antoninus Pius (Digest 1,6,2), while forbidding cruel or excessively severe treatment of slaves, upheld the master's traditional right to inflict punishments.
contubernales, not their owners. Slaves felt a bond of loyalty and affection toward other slaves in the same familia, owing to a feeling of community and shared servitude (Flory, 1978). They are the most logical people to have erected the majority of epitaphs, albeit anonymously, to their fellow slaves.
Inscriptions do not reveal the kindness of some owners or the brutality and tyranny of others. A few at least seem to have earned the affection of their slaves: Modestus erected a memorial to his master, the aedile Clodius Quadratus, at his own expense (de suo), while Primogenes and Felix did the same for "Secunda nostra", who was probably their mistress rather than a fellow slave. A more unusual relationship is seen in the epitaph of Fortunata, who is called the fidelissima conservatrix et amatrix domini, thus not only a loyal retainer but also lover of her master. This could mean that Fortunata was his concubine (Edmondson, 2009b, 271), though a more innocent interpretation is not excluded. Slave owners had no qualms about using their slaves for sexual purposes, and some of the attested vernae could be children of the owner. Some degree of owners' affection is suggested by the word delicium or delicata applied to two infant girls. These terms referred to a cute slave child who was coddled and played with, like a favourite pet, by the master or mistress (Laes, 2003). In some instances this might be an illegitimate child of the master (Veyne, 1987, 79). In the case of childless couples, the delicium might receive the affection that they would otherwise have lavished on their own offspring.
MacMullen (2014, 1) has stressed the need for historians to take into account the affective as well as the cognitive determinants of human behaviour in past societies. Certainly slavery and slave-owning were emotional experiences. While slaves suffered from anxiety, misery, pain, hopelessness and fear of punishment, owners felt contempt, insensitivity, fear of slave revolt, anger sometimes engendering brutality, and in rare instances compassion (Dwyer, 2012, 10-21). Behind the rigid phrasing of Lusitanian epitaphs we can appreciate the grief and affection felt by slaves for deceased loved ones, and occasionally tender feelings between owners and slaves. Affective response to death may indeed have been more important than "epigraphic habit" as the motivation for slaves to erect inscribed memorials to their fellows, particularly since, with the exception of paedagogi and clerks, most slaves would have been illiterate (Harris, 1989, 257-258) and not in the "habit" of composing or reading inscriptions. The stock formulas used in slave epitaphs were probably suggested by the stonecutters to their customers (Susini, 1973, 47-48).

Conclusions
This study began by determining criteria for identifying slaves in the epigraphic record. Although the inclusion of all persons with a single Greek, Latin or indigenous name would have added hundreds of additional examples, they would be inadmissible as evidence because there can be no certainty as to their servile status. A single name could represent a peregrine or the abbreviated name of a freedman. Therefore our analysis has been restricted to those instances, ninety-nine in number 60 , where an individual is explicitly labelled servus, verna or the like.
Even though limited to definite slaves, the study shows the shortcomings of an evidence-based (i.e. epigraphic) approach. Inscriptions tend to represent slaves employed in the household or in business, to the exclusion of those involved in agricultural labour or mining. For instance, there are no slave epitaphs from the mining village of Vipasca (Aljustrel). Inscriptions also do not provide a complete sample of slaves' ages at death, and are biased in favour of the young, largely because many older slaves would have changed their status to that of liberti. Finally, inscriptions often fail to indicate the relationship between the deceased and those who commemorate them.
Whatever the geographic origin of imported slaves, it is clear from indigenous nomenclature, the term verna, and votive dedications to local deities such as Ataecina, that some slaves were born in Lusitania. Male slaves greatly outnumber females in the epigraphic record. The limited number of female slaves, consonant with the restricted range of jobs available for them, was probably due to a combination of factors, which may have included a higher female death rate, earlier manumission of some women, and exposure of unwanted girls. Attestations of age suggest a high mortality rate: 30 out of 49 slaves of recorded age died before age 23, and only 7 lived past age 30. However, these statistics exclude slaves who were manumitted and subsequently died as freed persons.
Data on the persons who erected epitaphs show that slaves were often buried by their parents, who in some cases had already been manumitted while the child remained in slavery. Slaves could also be buried by siblings or fellow slaves, but there is no explicit example of a slave buried by the owner. The naming of the owner on a slave's epitaph shows only whose slave he was, not who set up the inscription, and there are numerous examples where we know the dedicant was not the same as the owner. Expense rather than modesty is the likely reason for not always naming the slaves who erected the memorial. The inscriptions also provide evidence that slaves entered into quasi-marital unions, sometimes adopting the terms vir (husband) and uxor (wife) as if legally married. In the case of mixed unions between slaves and freedmen, it is not possible to determine whether these occurred before or after the one partner was manumitted. Some slaves, at least, seem to have had amicable relations with their masters.