The Greek Census Inscriptionsof Late Antiquity'

KYLE HARPER

I INTRODUCTION

Trustworthy numbers, reliable glimpsesof rural life,and documentarysources from outside Egypt are rare in ancient studies.A dossier of evidence, then, which records measured quantitiesof land and labour in theEastern countryside officially of theLate RomanMediterraneanshouldbe of tremendous interest. Such evidenceexists,in theform - Astypalaia, Chios, Cos, Hypaipa, of Greek census inscriptions fromeleven cities Meander, Miletos, Mylasa, Mytilene on Lesbos, Perissa on Thera, Magnesia on the Samos, and Tralles.' Nevertheless,thecensus inscriptions presenta forbidding jumbleof and bureaucraticjargon,a fact which has rendered thestonesof abbreviations, fractions, use outside thespecialistliterature on Late Roman fiscal A. H. M. Jones little assessment.2 ofmaking offered documentsin I953.His study has thevirtue thestandardanalysisof the of theinscriptions and it thedry,technical information meaningfulforeconomichistory, foroverhalf a century.3 has held thefield Despite itsenduring value, Jones'sstudyis flawedin smallbut significant ways.More from Thera radicallychanges the importantly, the recent discoveryof a new fragment not tomention that economicprofile of the Greek countryside offered by thiscollection,
like to express my gratitude to those who have commented on this paper at various phases, especially and Scott Johnson. I also wish to thank the Editor, the anonymous referees for JRS, and Duncan-Jones Chris Wickham, whose responses have greatly improved the article. Above all I am indebted to Christopher Jones, and Brent Shaw, who have patiently helped me develop this paper from the time itwas a Michael McCormick, slavery. chapter inmy dissertation on Late Roman 1 IG XII.3, nos 180-2; A. D?l?age, La capitation du Bas-Empire ASTYPALAIA: (1945), 190-4. CHIOS: D?l?age, und Funde op. cit., 182-6, pi. III. COS: R. Herzog, Koische Forschungen (1899, repr. 1983), no. 14; M. Segre, I would Richard

Iscrizioni di Cos (1993), ED 151. HYPAIPA: J. Keil and A. Premerstein, 'Bericht ?ber eine dritte Reise in Lydien und der Wissenschaften, den angrenzenden Gebieten Klasse, Ioniens', Kaiserliche Akademie philosophisch-historische 57 (1914), nos 85-7; IGSK 17.2, R. Meri? et al., Die Inschriften von Ephesos 7.2 (1981), nos 3804-6. Denkschriften 122. no. am Maeander I.Milet 3.1389-90. MILETOS: O. Kern, Die Inschriften von Magnesia MAGNESIA: (1900), IG XII.2, nos 76-80; IGSK 34.1, W. Bl?mel, Die Inschriften von Mylasa MYLASA: (1987), nos 271-81. MYTILENE: tesLesvou: sympler?ma (1968), no. 17; E. Erxleben, 'Zur Katasterinschrift Mytilene S. Charitonides, Hai Epigraphai tax assessment from 'A fragment of a Diocletianic IG XII 2, 77', Klio 51 (1969), 311-23; R. Parker and H. Williams,

IG XII.6, G. Kiourtzian, Recueil des inscriptions EMC 39 (1995), 267-73. SAMOS: 2.980. THERA: Mytilene', (2000), nos i42a-g; E. Geroussi-Bendermacher, grecques chr?tiennes des Cyclades 'Propri?t? fonci?re et inventaire in V. Anastasiadis and P. Doukellis d'esclaves: Un texte in?dit de Perissa (Thera) tardo-antique', (eds), Esclavage IGSK 36.1, F. Poljakov, socio-culturelles (2005), 335-58. TRALLES: Inschriften von antique et discriminations 'Estates and the land in Late Roman Asia Minor', Chiron 37 Tralleis und Nysa (1989), no. 250; P. Thonemann, and Astypalaia, (2007), 435-77, provides a new edition of the inscriptions from Tralles along with some new for generously making his work available to me Thera, and Lesbos. I thank Dr Thonemann Magnesia, readings for in advance, and I have benefited from his careful study. 2 The basic discussions are D?l?age, op. cit. (n. 1), 163-96; A. H. M. Jones, 'Census records of the Later Roman Das Finanzwesen Staates (1958), 43-53; des fr?hbyzantinischen Empire', JRS 43 (1953), 49-64; J. Karayannopulos, A. Cerati, Caract?re annonaire et assiette de l'imp?t foncier au Bas-Empire (1975), 244-60; W. Goffart, Caput and Colonate: Towards a History of Late Roman Taxation (1974), 113-21; R. 'Note sur l'inscription Roman Economy (1990), 199-210; G. Kiourtzian, E. Magnou-Nortier (1993), vol. (ed.), Aux sources de la gestion publique 3 Consequently, most historians have followed his presentation: e.g. M. Europe

Southern Argolid from Prehistory to the Present Day (1994), 112; C. Wickham, Framing the Early Middle Ages: slave supply', in K. Bradley and P. and theMediterranean, 400-800 (2005), 277; W. Scheidel, 'The Roman World (eds), The Cambridge World History of Slavery, 1: The Ancient Mediterranean (forthcoming). Cartledge dissent. op. cit. (n. 2), 199-210, offered an under-appreciated Duncan-Jones, (?World Copyright Reserved. JRS 98 (zoo8), pp. 83-II9. Exclusive Licence to Publish: The Society for the Promotion of Roman

Structure and Scale in the Duncan-Jones, in cadastrale IG XII 3,N? 343 de Th?ra', 1, 35-44. Jameson et al., A Greek Countryside: The

Studies 2008

84

KYLE

HARPER

our knowledgeof theLate Roman economyhas expanded since the I95OS.4 All of this a reappraisal justifies of theinscriptions.' This articlebeginsby reopening thequestionof the date and contextof the inscriptions, challengingthe conventional wisdom which ascribes themtoDiocletian or his immediate successorsand arguingthat they belong to thelaterfourth century (SectionII).Then thearticleconsiderstheeconomicdata preserved in the inscriptions. The collectionoffers valuable documentary intothestructure insights of wealth (III) and thedeployment of labour (iv) inLate Antiquity.A finalsection (v) considers thedemography of a large slave population recordedin the remarkable new fragment from Thera. The goal of thisarticle is to reframe what the inscriptions can and cannot say about economichistory.6 The interpretation which emergesstands incontrastto thatpresented who saw in the inscriptions a structural by Jones, crisisof Late Roman agriculture. The inscriptions should be seen, rather,as artefactsof a dynamicperiod in theEastern which thecompositionof thearistocracy Mediterranean,during was mutating and the laboursystem was complex.Landedwealthwas stratified but fragmented, ina ruralsector was deeply influenced, that not utterly though dominatedbyurban landowners. The prac ticeof assigning head-tax liability(capitatio)to thelargestlandowners was common,but thedossier suggeststhat land leasing,freeof fiscalties,remaineda prominentstrategy among smaller landlords. Most importantly, thesedocuments insistthat slave labour played a vital role inagricultural on elite-owned production land.The new fragment from Thera provides incomparable evidencefora slave-basedestate inLate Antiquity. Demo graphicanalysisof this population suggests reproductive successandmeaningfullevelsof male manumission. The demographic findings have far-reaching implicationsfor the natureof estate managementand thecomplexcharacter of social relationships in therural economyofLate Antiquity.
II THE DATE AND CONTEXT OF THE INSCRIPTIONS

The inscriptions come from elevencities,all of themin theeastern Aegean islandsor the mainland of far-western AsiaMinor. The geographyis interesting, both administratively and economically. Nine setsof inscriptions come fromtwoprovinces, Asia and Insulae.7 Both Asia and Insulae fell within theproconsulateofAsia, an administrative unit inde pendent fromthenormaldiocesan hierarchy and under thecontrolof a proconsulwho was equal in rankto thevicarofAsiana.8The onlycensus inscriptions outside originating of theproconsul's jurisdiction are those from Miletos andMylasa, in theprovinceof Caria. The provincialgovernment ofCaria was under theauthority of thediocesan vicar ofAsiana normalpost-Diocletianic the More thana 'momentary organization. regional fad', the inscriptions should be the products of an officialadministrative action that
4 To name only a few recent contributions on the rural economy of the Late op. cit. (n. 3), Empire: Wickham, and Society in the Age of Justinian (2006); A. Chavarria and T. Lewit, especially 259-302; P. Sards, Economy research on the late antique countryside: a bibliographic 'Archaeological essay', inW. Bowden, L. Lavan and C. Machado in (eds), Recent Research on the Late Antique Countryside (2004), 3-51; J. Banaji, Agrarian Change Late Antiquity: Gold, Labour, and Aristocratic Dominance (2001). 5 This study is reliant upon the published editions. Thonemann, op. cit. (n. 1) has shown the value of closely re examining the stones. 6 The reflections of A. Bowman, in theHermopolite nome in the fourth century A.D.', JRS 75 (1985), 'Landholding land registers could be applied in this case: 'the result may be a greater rather than a 137-63, on the Hermopolite lesser degree of uncertainty about many important issues... the picture ismore complex, the developments more subtle and ambiguous than might once have appeared'. 7 For Insulae, see J.Marquardt, R?mische Staatsverwaltung (1881), vol. 1, 348-9; Hierocles, Synecdemus, 686-7 (Ed. G. Parthey (1866), 26-7). 8 Notitia Dignitatum Or. XX.5-8 (Ed. O. Seeck (1876), 45-6); A. H. M. Jones, The Later Roman Empire, 284-602: A Social, Economic, and Administrative Survey (1964), 375.

THE

GREEK

CENSUS

INSCRIPTIONS

OF

LATE

ANTIQUITY

85

includedthe provinces under thecontrol of the proconsulaswell as thoseunderthe power of thevicar.9 The inscriptions are keydocuments of post-Diocletianicfiscality, with itsdual system of liability on land and persons, iugatioand capitatio.10 It is important to recognizethat of individuallandowners;they theseinscriptions recordthe tax liabilities are not a com in thesenseof amodern census.Large landowners could be plete countof souls or things liable for thecapitation taxesof theirregistered dependents, whether slaves, tenants, or A fundamental lawofA.D. 37i drew a distinction betweentenants permanent employees.1" own property with their who leased extra land and coloniwho exclusively worked the The landlord property of an individuallandlord.12 was only responsible forthehead tax in thelatterinstance, when the workerhad no landof his own; theextent of unregistered renderedinvisiblein thesestones. The inscriptions likewise tenancyis thereby measured of itsfiscal burden (assessedinunitsof arable,vines,olives,and pasture). theland in terms We thussee therurallandscapethrough a particular prism,thefiscal assessment of specific landowners fortheir registered labourers and taxable land. in their ormethod ofmeasurement.Ithas been The inscriptions arenot uniform format are two subsetsin thegroup, 'primary' and 'secondary'registers.13 So argued thatthere recorded called primaryregisters quantitiesof land and persons in raw physicalunits.14 reflected theconversion of rawphysicaltotalsintofiscal Secondaryregisters, by contrast, units isunknown units, iugaand capita; thescheduleused to convert physical intofiscal are said to reflect a sourceof uncertainty.15 The differences of format the and represents the and thenitscalcula fiscal processat different stagesofmotion, first plainmeasurement is tooneat.17 tionintotax liability.16 But this divisioninto and secondary primary registers or at exactlythe No twocities memorialized thefiscalliabilitiesinexactlythesame form same instantin the cycle.This diversity within an envelope of similarity points to a combination of imperial stimulus and local controlover thefiscal process. The inscriptions have eluded a precisedating.Jonesclaimed thatthey'wereprobably or earlyfourth century A.D.,whenDiocletian and his colleagues engravedin thelate third out censuses'.'8 On this logic, and successorsare known to have been active incarrying
9

The phrase isGoffart's, op. cit. (n. 2), 121. The province of Phrygia and Caria was separated from Asia already in the third century', JRS 71 (1981), 103-20. It is 'Rome, Asia, and Aphrodisias by the 250s A.D.: C. Rouech?, interesting that possibly in the mid-fourth century, and certainly in the early fifth, the positions of proconsul and vicar were combined: D. Feissel, 'Vicaires et proconsuls d'Asie du IVe au VIe si?cle: remarques sur l'administration au bas-empire', Antiquit? du dioc?se asianique tardive 6 (1998), 91-104. The inscriptions could originate from a moment when the offices were combined, or they could tell us that combined jurisdiction reflected an underlying

administrative coherence; see CT 7.6.3 (a.D. 377), cited below, which also insinuates joint financial administration. 10 J.-M. Carri?, 'Diocl?tien et la fiscalit?', Antiquit? Tardive 2 (1994), 33-64. 11 del Basso-Impero": See Section iv, for the status of the labourers. J.-M. Carri?, '"Colonato la resistenza del e contadini delVlmpero romano B. Sirks, in E. Lo Cascio mito', (1997), 75-150; (ed.), Terre, proprietari 110 (1993), 331-62. 'Reconsidering the Roman colonate', ZRG 12 CT 11.1.14; Carri?, op. cit. (n. 11), 100: 'un testo fondamentale'; C. Grey, 'Contextualizing colonatus: the origo of the Late Roman Empire', JRS 97 (2007), 155-75, at x^9- See O. Seeck, Regesten der Kaiser und P?pste f?r die Jahre 311 bis 476 n. Chr. (1919), 27, for the reasons itmust be A.D. 371. Iwill argue below that the inscriptions and this law are connected to precisely the same census. 13 op. cit. (n. 1), 169, 181?2. op. cit. (n. 2), 46-7; D?l?age, Karayannopulos, 14 and Mylasa. The primary registers are Thera and Lesbos, along with Hypaipa, Miletos, 15 The secondary registers are from Astypalaia, Chios, Cos, Magnesia, Samos, and Tralles. A. H. M. Jones, 'Capitatio and Iugatio', JRS 47 (1957), 88-94, is a lucid discussion of fiscal assessment. The size of the iugum is

considered in Section in. 16 Erxleben, op. cit. (n. 1), 315. 17 alone lists farms alphabetically declarations. Magnesia by district. The Hypaipa uniquely records household Thera inscription may include both primary and secondary elements. A block from Lesbos provides the only indication that land might be graded into first and second class, as suggested by the complex schedule outlined in 106c: Ed. W. Selb, Das the Syro-Roman Lawbook, two levels of arithmetic within the fiscal units. 18 Jones, op. cit. (n. 2), 49. syrisch-r?mische Rechtsbuch, 3 vols (2002). Astypalaia includes

86

KYLE

HARPER

a Tetrarchicdate.'9Recently, most have favoured Thonemannhasmade a strong case that theinscriptions shouldnotdate beforeaboutA.D. 3IO,allowing timeforthecensusprocess He also claims thatthestones tobe completed. must date beforethe3zos, arguingthat one of thepropertiesatMagnesia may have been in thepossession of thegoddessArtemis. This reading were temple of thestone isnot certain, and,moreover,there propertiesin the eastern Mediterranean intothelatefourth century.20 of Instead,thelater yearsof thereign antequem fortheinscriptions: Theodosiusmark theterminus thepagan religious officials are impossibleto imagine recordedin theregisters afterthe380s.21 The assignment to the that it isworth reconsidering the entireperiod between Tetrarchicperiod is so flimsy C.A.D. 3io and 390 as a possible context. Within thisrange,thereisnothing which con are reasons toprefer a laterfourth-century clusively demands a particular date, but there originfortheinscriptions. The inscriptions can be analysed fromtheperspective of fourdating criteria:scripts, and prosopography. onomastics,language, Unfortunately, many of thepublishededitions are inadequate.The inscriptions of the census inscriptions from Thera, however,have received carefulre-edition. The editornotes theappearanceof cursive delta, a letter form which is first attestedin the of thefourth-to-fifth inthe Cyclades around theturn centuries, todate the inscriptions Melos.22He prefers catacombsof nearby from Thera sometimein thecourseof thefourth have been struck century. Many observers of by the inconsistency letter forms and irregularities of spacing throughout the inscriptions.23 Keil and Premer of the late imperial steindescribedthe 'irregular and oftenindistinct letters period' in the with another local inscription Hypaipa inscriptions, contrastingsharply that can be with variable letter dated toA.D. 30I.24 forms and poor lineation narrowly Scripts could be or laterfourth-century consistent with an earlier date, thoughina setofpublic inscriptions more comfortably from Roman Asia thesecharacteristics fit with a laterdating.25 The

description of the farm's location), but this does not make the goddess the owner. Above all, temples owned land late into the fourth century. See R. Delmaire, sacr?es et res privata: l'aerarium Largesses imp?rial et son administration du IVe au Vie si?cle (1989), 641-5. 21 line di3. The priests at Tralles and Magnesia do not imply an early fourth-century date; cf. the e:g. Magnesia, in the municipal album from Timgad, dated to the 360s A.D. (Ed. pagan priests among the civic dignitaries A. Chastagnol, Valbum (1978)). On the official suppression of paganism under Theodosius, municipal de Timgad see J. Curran, Pagan City and Christian Capital: Rome in the Fourth Century (2000), 209-17. 22 C. Rouech?, Aphrodisias in Late Kiourtzian, op. cit. (n. 1), 215. The letter form is late also at Aphrodisias, Antiquity (1989), 332, with cautions about using late letter forms as a dating tool. 23 The Astypalaia inscriptions are said to have be written in letters 'of a late age', without reference to specific IGSK 34.1, no. 271: 'mauvaise gravure, lignes irr?guli?res ... aspect cursif.' letter forms: IG XII.3, no. 180. Mylasa: Lesbos: Parker and Williams, op. cit. (n. 1), 268: 'crudely cut'. 24 Keil and Premerstein, op. cit. (n. 1), 67-8: 'unregelm?ssige und vielfach undeutliche Buchstaben der sp?ten Kaiserzeit.' The fragments of A.D. 301 have been shown to concern a trust belonging to an association of wool sellers: Th. Drew Bear, 'An act of foundation at Hypaipa', Chiron 10 (1980), 509-36. See also G. Fagan, Bathing in Public in the Roman World (1999), 344. 25 cf. Rouech?, op. cit. (n. 22), xxii. On the other hand, the census inscriptions are unique expressions of public op. cit. (n. 1), 444, notes that the inscriptions may epigraphy, leaving us without direct comparanda. Thonemann, reflect documentary practices.

19 op. cit. (n. i), 163, implying a Diocletianic date; Karayannopulos, op. cit. (n. 2), 45-6, for A.D. 289 or D?l?age, 'Roman Asia Minor', in T. Frank (ed.), ESAR 4 (1938), 914-15. Erxleben, op. cit. (n. 1), 298; T. R. R. Broughton, 314, argued for the years A.D. 307?313 on the basis of the Hypaipa inscriptions, which he argued show the registration of the urban plebs; these seem to be village household declarations. Cerati, op. cit. (n. 2), 255, offered the first dissent, arguing for a date in the late fifth or sixth century. Kiourtzian, op. cit. (n. 1), favours a date in the course of the fourth century. 20 as [xco(piov) Apjx?ui?o? Thonemann, op. cit. (n. 1), 438-9. Line a3 records a property described jtpo? toponyms throughout the census inscriptions are derived from the Gl)vop(ioi?) novo7t?pyou HpaK?Axoi). Many names of the pagan gods; the 'chorion of Artemis' names the place, not its owner. It is unusual (but cf. di6-i8) that this entry did not list an individual declarant (though perhaps there was not room on the line after the verbose

THE

GREEK

CENSUS

INSCRIPTIONS

OF

LATE

ANTIQUITY

87

inscriptions also use abbreviationsymbolsthatare not attested epigraphically beforethe fourth century and are unusualbefore A.D. 350.26 The onomasticevidenceis inconclusive butmay also suggest a later date. In additionto a scatter of individual names, there are two largesetsof names: a groupof I52 slavesfrom from Thera and about 65 landowners Magnesia. Most of thenames are common Greek names,albeit with a lateantique flavour.27 The admixture of Jewish andChristiannames is distinct.Among the slaves, Theodoulos, Eustathia, Theodote, and Sambatia are A landowner on Astypalaiawas also called Theodoulos, a name borne exclus attested. ively byChristiansand Jews.28 The relationship betweenonomasticchangeand religious conversioninLate Antiquityis controversial and farfrom clear.Bagnall has uncovereda surprisingly quick pace of onomasticchange inEgypt,but it would appear thattherateof in naminghabitswas slower in the core areas of theGreekMediter transformation Itwould not be unreasonable to questionwhether flagrantly Christiannames ranean.29 in a public record shouldbe so prominent of slavesduringa periodof persecution.30 The language of theinscriptions is a highly abbreviated dialectofLate Roman bureau ofC1y67 cratese. and KcoaXai indicates a period contemporary The technicalterminology with or later thanDiocletian. As Cerati noted, the presence of a neologism like could argue that the terms ofDiocletianic policyhad evolved.31 The term, UWyoKSC'aXOv used at Astypalaia, does not appear in other documentsuntil a praetoriandecree of A.D. 480 and thenan imperial constitution issued inA.D. 498.32 Cerati's imputation was A are later,perhapsmuch later,than theDiocletianic reform. that these inscriptions intheseinscriptions, a term which with theuse of the word JtapoLKoL parallelcan be drawn also enteredofficialparlance rather late.33Its appearance in the census inscriptions Roman usage, but therelative documentsin predatesanyotherofficial paucityof official makes itdangerousto speculateon thebasis of language. Greek fromthefourth century Prosopographyholds themost hope for dating. The sixty-five landowners in the or sixmembersof thesenatorial includesix decurions,five Magnesia inscription order,

'The small superscript Greek letters as abbreviations. Thonemann, op. cit. (n. i), 443-4; A. Chaniotis, Namely, new evidence and old problems', SCI 21 (2002), 209-42, at 215. Jews of Aphrodisias: 27 Geroussi-Bendermacher, op. cit. (n. 1), 345-9, has a good analysis of the names on Thera. Only some villagers at Hypaipa attested carried the name Aurelius, which is perhaps unsurprising for the sort of modest households here. Over the fourth century M. Aurelii disappear: cf. J. Reynolds and R. Tannenbaum, Jews and God-fearers at (1987), 20. Aphrodisias 28 IG op. cit. (n. 1), 347. XII.3, no. 182. Geroussi-Bendermacher, 29 R. Bagnall, change in Early Byzantine Egypt', BASP 19 (1982), 105-24; 'Religious conversion and onomastic E. Wipszycka, 'La valeur de l'onomastique pour l'histoire de la christianisation de l'Egypte. A propos d'une ?tude a reply', ZPE 69 (1987), 243-50. de R. S. Bagnall', ZPE 62 (1986), 173-81; R. Bagnall, 'Conversion and onomastics: The bank of names preserved from Late Roman Aphrodisias (see the name index of Rouech?, op. cit. (n. 22)), would also suggest a slower pace of change throughout the fourth century. 30 in for instance, the name index from the second half of the third century in Phrygia available Compare, E. Gibson, The 'Christians for Christians' Inscriptions of Phrygia (1978), which shows a thoroughly normal set of in the immediate pre-Tetrarchic period. It is also worth noting Greek names among a group of known Christians the extreme frequency in the census inscriptions of names ending in -lo?, a form not unusual from the late second op. cit. (n. 1), 349; Kiourtzian, op. cit. century, but very popular from the fourth century. Geroussi-Bendermacher, (n. 1), 215; B. Salway, 'What's in a name? A survey of Roman onomastic practice from c. 700 b.c. to a.d. 700', JRS are more traditional, which might be expected among the 84 (1994), 124-45, at I3^- The names at Magnesia conservative, landowning classes of a provincial town. Still, names ending in -lo? are well represented. 31 at Thera is rightly expanded, If the abbreviation k? = K(?<|)a^o)?(uya)? the use of two separate portmanteau terms would suggest that they were the product of evolution, not an imperial formula dictated from the centre. Kiourtzian, op. cit. (n. 1), 142a, lines 4, 5, 11. See below, n. 67. 32 en Carie (ierAo?t 480)', TM 12 (1994), 263-97; inscrite ?Mylasa D. Feissel, 'L'ordonnance du pr?fet Dionysios CJ 10.27.2 (a.d. 498); Banaji, op. cit. (n. 4), 57-9. This point involves extraordinarily thorny problems about the development of Late Roman fiscality, but iuga and capita do not appear together in Latin laws until themid-fourth century. See Jones, op. cit. (n. 15), 88. 33 op. cit. (n. 1), 175; Kiourtzian, op. cit. (n. 1), 225. The earliest official use is CJ 1.34.1 from the reign D?l?age, of Anastasius. On the status of paroikoi in the inscriptions, see Section iv.

26

88

KYLE

HARPER

None of the individuals two perfectissimi, of senatorialrank can be and an Asiarch.34 Around A.D. 300, the identified, yet the prominence of senatorsis itself thecrucialdatum.35 Westerners.Over the fourth Senate claimed only c. 6oo members, the bulk of them century,thenumberof senatorsexpanded into the thousands; the late 350s-360s saw were the main recruiting grounds for thenew feverish growth.36 The Eastern provinces
Senate

Duncan-Joneshas shown that providedby theinscriptions is truly striking, especiallysince We are forcedto believe in one of two scenarios. 60-75 senators in thecomplete list.37 under Diocletian happened toown landat of all Eastern senators Eithera highpercentage Magnesia, or the inscription postdates the expansion of theEastern Senate and the are an unrecognized artefact of devaluationof thesenatorialtitle. Surely,the inscriptions Eastern elite.38 a crucial fourth-century transformation: thereconfiguration of the If the census inscriptions date to sometimeafter the 350SA.D., it requiresus to re one of the census evaluate theiroriginal purpose.39 Though it has gone unremarked, was behind theact of inscribing inscriptions preservestherevealing detail thattheimpetus This is in itself to census 'books' and an imperial unusual: references order to do so.40 whereas thereisno clearparallel to thepracticeof 'pages' inLate Antiquityare common, tax liabilities on stone.44 of thecensus inscriptions inscribing The veryexistence begs for
34 are XauTtpoxaxoi, i.e. clarissimi: see Jones, op. cit. (n. 8), 528-9. Jones convincingly The senators atMagnesia should be decurions argued that the <()i>.OG8?aaxoi (though the word does not have this sense in the High Empire, in near-contemporary is called f| (|>iA,OG8?ao~xoc ?ouA,f|); otherwise the curial inscriptions the council ofMagnesia class ismissing entirely. The size of their properties matches other curial-scale holdings. 35 A The members of the senatorial order are Priscillianus, Capitolinus, Hermonactiane, Eutyches, Aristocleia. senator, Attalus, appeared in a patronymic on Thera. These should be Easterners, since the landholdings of even the Western provinces: see especially Western Senate in the fourth century were limited to the wealthiest members of the D. Vera, 'Simmaco e le sue propriet?: Struttura e funzionamento di un patrimonio aristocr?tico del quarto sec?lo sur Symmaque inAtti del Colloque Genevois d.C, (1986), 231-70, at 243-5. 36 et ses For the creation of the Eastern Senate, see G. Dagron, Naissance d'une capitale: Constantinople institutions de 330 ? 451 (1974), 119-210; A. Chastagnol, 'Remarques sur les s?nateurs orientaux dans le IVe si?cle', 'Senators and Senates', inA. Cameron and P. Garnsey (eds), 24 (1976), 341?56. See especially P. Heather, AAntHung

at Constantinople.

The

number

of senators

in the small but random

sample

we have at most

8 per cent of theMagnesia

register. This

leads us to project as many as

for which see B. Salway, 'Prefects, patroni, and per cent of the total, itwould imply the right order of magnitude: in A. Cooley decurions', (ed.), The Epigraphic Landscape of Roman Italy (2000), 115-71, at 127. Indeed, the small number of attested decurions strongly reinforces the inference that the original number of senators was also originally much larger. 38 The number of Greek senators grew throughout the second and especially the third centuries, as a coterie of old aus den see H. Halfmann, 'Die Senatoren ?lites from cities like Ephesus entered the order: municipal Provinzen des r?mischen Reiches vom 1. bis 3. Jahrhundert', in Epigraf?a e ordine senatorio (1982) Kleinasiatischen 2, 603-50; H. Halfmann, Die Senatoren aus dem Chr. (1979); B. Remy, Les Carri?res s?natoriales J.C.-284 apr?s J.C.) (1989). Even assuming that claimed that there were still only 300 Eastern recorded sixty senators,

CAH2, vol. 13 (1998), 184-210. 37 op. cit. (n. 2), 137-8. The stone for farms starting with the letter beta has thirty-seven farms. This Duncan-Jones, is the nearest-complete stone, although another fragment shows that we are missing some 'betas'. Using the CIG as a database of place-names, Duncan-Jones estimated that beta should account for about 3.5 per cent of all place names. Thus we have, at most, 8 per cent of the original, possibly less. There were also six to eight representatives of the curial order listed. The size of town councils varied, but if the number visible in the extant fragments was 8

then one out of every five Eastern

?stlichen Teil des Imperium Romanum bis zum Ende des 2. Jh. n. au haut empire (31 avant dans les provinces romaines d'Anatolie fully half of all 600 senators were Eastern by A.D. 300 (Themistius senators as late as A.D. 357), and that the inscription originally senators held land at Magnesia, a ratio which seems

problem of the function of the stones seriously. 40 /^?0]oi) npOKofwrjaioi). IGSK 34.1, no. 275: ?]v<|>i Kaiaap[o? / oia]xayu?xco[v Mylasa: 41 op. cit. (n. 2), 47, with references to libri censuales, censuales paginae, diagrapha, polyptycha, Karayannopulos, = Lit. 3.1, no. 17) is a possible parallel, but see the apt cautions of inscription (CIL X.407 KG??l?, etc. The Volcei Goffart, op. cit. (n. 2), 113-14: we do not know what the Volcei register is.

implausibly high. 39 On the importance of the material function and 'overall physical appearance' of inscriptions: A. Cooley, 'Introduction', The Epigraphic Landscape op. cit. (n. 1), 439, argues that the of Roman Italy (2000), 1. Thonemann, census inscriptions were permanent memorials census. It would weaken of the Diocletianic the case ifDiocletian account is the first to treat the intended a regular census cycle (see the references below, n. 48), but Thonemann's

THE

GREEK

CENSUS

INSCRIPTIONS

OF

LATE

ANTIQUITY

89

explanation,and indeed thereare administrative patterns in the later fourth century, specifically in the reignofValentinian and Valens, which may offera solution.42 Their tenure was marked by aggressive efforts toestablish central controlover thefiscal process, in part by compiling lists,breviaries,and registers of public obligations.43 Notably, ThemistiusandAmmianus both observed that Valens managed theEmpire as thoughit were a private with theabilityto see its household, budget,income and expenses, province by province.44 This metaphorpoints towardsthefiscalreforms whichmay have createda political motive and an administrative contextforthecreationand displayofmonumental tax records. Valentinian and Valens implemented sweepinginstitutional reforms aimed at central izingcontrolover the fiscal process. Imperialtaxation was a delicatebalance of imperial standards and local action.45 While Julianraisedcurialautonomyto a principle ofhis rule, ValentinianandValens dramatically reversed this policy in favour of imperial oversight. The emperors made an abortiveeffort to substitute retiredimperial officialsforcurial agents in theprocessof tax collection.46 The tensioninherent in this policy isvisible,for example, in an Italian inscription which shows that the emperorshad to order their governorsto audit theactual tax chargesagainst thearchivedliabilities.47 Their central izingpolicymade the intersection of the tax archivesand thedistribution of annual chargesa flashpoint of dispute.Cyclical imperial censuses were rarely conducted in the fourth so thecontrolof thetax registers century, gave thetowncouncilsa real measure of informal power.48 Valentinian andValens intervened at thissensitivejuncturein the tax process, a move which provides a possible political motive behind the creation of epigraphic archives.49 If thiscentralizing policy provides thepolitical background, itmay be possible to connectthestones with a specific action.Convergingevidencepoints to a administrative AsiaMinor in theyearA.D. 37I. In an inscribed census in letter to theproconsulofAsia, Valens describeda seriesof imperialactswhich occurredon thebasis of progressively betterfiscal data from Asia. Initially Valens had acted 'on thebasis of an assessment', but
42

12.6.9 (a.d. 365); CT 12.1.97 (a.d. 383); D?l?age, op. cit. (n. 1), 34-7. 47 inGiardina and Grelle, op. cit. (n. 42), especially 263-5. Cf. also CT n.4.1 Analysed (a.d. 372). 48 For the absence of a regular census: Jones, op. cit. (n. 8), 454. The absence of evidence is striking, for when a census did happen, it is audible (e.g. Lactantius, De mort. pers. 23.1-2 (Ed. J. Creed (1984), 36); cf. the Cappadocian fathers below). Seeck thought there were regular five-year censuses conducted throughout the fourth century, but the evidence he collected is unconvincing after the 320s A.D.: O. Seeck, 'Zur Entstehung des Indictioncyclus', 12 (1894), 279-96; noted by T. D. Barnes, The New Deutsche Zeitschrift f?r Geschichtswissenschaft Empire of Diocletian and Constantine (1982), 226-37. 49 A series of laws which mandated the careful recording of annual receipts against tax assessments emanated from the East over the next century. See Feissel, op. cit. (n. 32), 288.

Roman Empire (1972), 161-7. For regional variety: D?l?age, op. cit. (n. 1), 34-7, 224-5, 24?~5> R- Bagnall, Egypt in Late Antiquity (1993), 157-9. The imperial centre provided standards of tax assessment and channels of appeal, while town councils retained material power through control of the process: the maintenance of census books, the distribution of annual charges, and the act of collection itself.When Symmachus needed to know who actually ? not the governor ? and asked who had been paying possessed a disputed property, he went to the town council taxes on it. Symmachus, Rel. 28 (Ed. O. Seeck, MGH AA 6.1 (1883), 302-3). For the role of town councils in distributing charges, see, e.g., CT 8.15.5 (a.D. 366). 46 CT 12.6.5 (a.D. 365); CT 12.6.7 (a.D. 364). The reform was perhaps not carried out inAfrica and Egypt. Cf. CT

A good general account of their rule: N. Lenski, Failure of Empire: Valens and the Roman State in the Fourth and F. Grelle, 'La tavola di Trinitapoli. (2002), 264-307. Jones, op. cit. (n. 8), 138-54. A. Giardina Century A.D. Una nuova costituzione di Valentiniano I',MEFRA 95 (1983), 249-303. 43 Lenski, op. cit. (n. 42), 272-4. 44 Themistius, Or. 8.114 (Ed. H. Schenkl and G. Downey (1965), 174): 7t poo p?v (bent?p oiK?a? ua?? xf|? xoaauxr|? See P. Heather and J.Matthews, The Goths in the exou?, xi 5? ?va^o?xai. ?pxf|?, xi ju?v 7tp?CT?iaiv ?Kaaxou Fourth Century (1991), 13-26. Cf. Ammianus Marcellinus, Res gestae 31.14.2: 'ut domum propriam'. 45 Curial role, see A. H. M. Jones, The Greek City from Alexander to Justinian (1940), 144; Jones, op. cit. (n. 8), 456-7; Goffart, op. cit. (n. 2), 7-26 (but cf. the review of R. Duncan-Jones, JRS 67 (1977), 202-4, for a balanced in the Later opinion of imperial oversight); J. H. W. G. Liebeschuetz, Antioch: City and Imperial Administration

90

KYLE

HARPER

of theinscription, aroundA.D. 371,he claimed tohave 'amost complete in the presenttime The letterimpliesa vigorousevaluation of account' of taxable iugaof thecivic class.50 would requirean from Caria, however, fiscalliabilitiesinAsia.51The census inscriptions evidencepoints to ofAsia. Indeed,thelegaland literary proconsulate act broader thanthe Asiana and Pontica inA.D. 37I.52 The legalevidenceconsistsof a a fullcensus throughout Modestus. He was equippedwith broad authority dossiergiven to thepraetorian prefect, audit public liabilities, and see that 'thecensus is stabilized to renovatetax assessments, He receivedlandmarklaws sorting decurionsfromsenators are ended'.53 and arguments The itinerary and with theirlandlord.54 which coloni shouldbe registered and determining legal briefofModestus inA.D. 37I show thata major reallocationof tax liability was Pontica vividly neighbouring Literaryevidencefrom carriedout underhis supervision.55 to thiscensus.56 testifies must date betweenc. A.D. 3io and 390, a strong case can In sum, while the inscriptions Tetrarchicdate. The bemade forplacing themseveraldecades laterthan the traditional argumentin favourof a Magnesia inscription is a compelling numberof senatorsin the of the date after theexpansion of theEastern Senate in the 350SA.D. If the suggestion period aroundA.D. 37I is right, shouldbe relatedto thecentralization of the inscriptions and thefullrevamp tax collection (which made thearchivalrecordsa siteof contention) context. ingof the census inAsia. This date provides a motive and an administrative thisis a circumstantial case, and an earlierdate remainsa possibility. Ultimately,though, If the conjectureof Valens' reign is correct,however, the stones become important of fiscal but in the not only in the documents assessment, politicsof taxationat the history and local power. pointof contactbetween imperial
III THE INSCRIPTIONS AS ECONOMIC DATA: LAND OWNERSHIP

on landed exact date, the inscriptions preservevaluable information Regardless of their The stonesallow of labour in thefourth-century countryside. wealth and thedeployment quantitative insights into the composition and scale of individual properties, the
in A. Chastagnol, 'La l?gislation sur les biens des villes au IVe si?cle ? la lumi?re d'une inscription d'Eph?se', (1986), 77-104. His text is used. Lines 2?3: '[Quod ex red]itibus fundorum iuris re[i publicae, quo]s intra civitatibus ad instaurandfam mojenium diversis quibusque faci[em <c. 10 spaces> pr]o certis | [partibu]s Lines 12-13: 'Ha(n)c sani (sic) quia ratione plenissima, quod intra Asiam rei concensimus.' habita aestimatione publica iuga esse videantur.' 51 Valens asked for even more complete registers. Lines 19?20: 'sane quia rerum omnium integram cupimus habere 50

AARC Asiam

notitiam et ex industria nobis tuam expertam diligen[ti]a[m | confit]emur, plena te volimus (sic) ratione disquirere per omnem Asiam provinciam fundos iugationemque memoratam.' 52 The geography of the stones would argue that Asia and Asiana were administered similarly (as does their see Feissel, op. cit. (n. 9)). Moreover, in a law of A.D. 377, levies were combination under a joint vicar-proconsul, applied at different schedules around the Empire. The law bundled the dioceses of Asiana and Pontica together as though they also cohered in terms of theirmode of fiscal assessment (and itdid not need to mention the proconsulate separately): CT 7.6.3. 53 CT 13.10.7 (a.d. 371); CT 13.5.14 (a.d. 371); CT 13.10.7 (a.d. 371): 'exhibitis partibus secundum fidem rerum coram cognoscant ac stabilitatem census finita altercatione component.' 54 CT 11.1.14 (a.d. 371); CT 12.1.74 (a.d. 371). 55 PLRE IModestus Code often descend from specific administrative laws in the Theodosian 2, 606-7. The contexts and were not, originally, 'general': J.Matthews, the Law: A Study of the Theodosian Code Laying Down

(2000), 66?71. The itinerary ofModestus might thus argue that the census was limited to Asia, Asiana, and Pontica. was made consul in a.d. 372. Modestus 56 T. A. Kopecek, Fathers and civic patriotism', Church History 43 (1974), 293-303, collects 'The Cappadocian some of the evidence, but does not acknowledge the role of the census behind the campaign. See Basil, Epistulae (Ed. Y. Courtonne (3 vols; 1957-66)), Letters 36, 37, 83, 88, 104, 284, 299, 309, 310, 312, 313, 315. In a.d. 371, as part of was divided in two? a well-known event that has not been situated the province of Cappadocia the reorganization, reforms of a.d. 371. Basil struggled against this division within the momentous (Letters 74, 73, 76). See W.-D. Or. 19 (PG 35, cols Basilius von Caesarea: Hauschild, Briefe I (1990), 138-41, 208-9; Gregory of Nazianzus, 1044-64).

THE

GREEK

CENSUS

INSCRIPTIONS

OF

LATE

ANTIQUITY

9I

of landedwealth, and the social profileof land ownership.Four sets of distribution inscriptions areparticularly informative: Thera, Lesbos,Tralles, and Magnesia. However, units. The inscriptions thedata are recordedindifferent from Thera and Lesbos are prim ary registers, so thatlandand labourare tabulatedin raw totals. The data from Magnesia andTralles, bycontrast, aremore complex,forin theseinscriptions therawnumbers have been convertedto fiscal units.The scheduleused to convert physical iugerainto iuga is us toconsiderthepossible ratesof unknown, and a full understanding of the data requires conversion. The Magnesia inscription poses specialproblems, but it remains, despite the uncertainties, among the most revealing documentsfor the structure ofwealth inLate
Antiquity.

The inscriptions from Thera andLesbos furnish and directevidenceforthecomposition a basis scaleof several properties. The sample is smallbut straightforward and establishes On bothThera and Lesbos, the registers more complexdocuments. for interpreting the An owner's totalproperty, were organizedby individual proprietor. typically fragmented was added up in terms of iugera of arable land, iugera under into numeroussmallerfarms, An individualowner's farms vines, and the number of olive trees.57 were classified were theproperties geographically by district, village,or ruralspace. In the inscriptions, in terms toKf7Ot, '6tOI, and named and locatedprimarily of xcpit, with some reference The inscriptions aypoi. The Xxpiov was thedominantgeographicunit everywhere.58 in traditional not an attempttodescribethecountryside terms presenta fiscal geography, As an overriding was flexible such as Kdogatl. bureaucraticcategory,the termXcopiov were located, as well as independent enough to encompass villages inwhich farms give the overall The inscriptions nucleated settlements comprisedof single estates.59 of highlyfragmented with occasional glimpsesof villages,on impression landownership, estate-based settlements on theother.60 theone hand, and dispersed, thatpropertiesin thisregionof The inscriptions from Thera and Lesbos demonstrate theEmpire could reach considerableproportions.From Lesbos, four stones survive. thenames and headingsare lost, so it is unclearwhere each registration Unfortunately, We do not knowwho owned thisland,nor therelationship betweenthe beginsand ends. a separateestate,there On theassumptionthateach stone represented individual stones. is information different about four properties:
LESBOS XII.z.76 XII-z-77 XII.z.78 XII.z.79 in iugera ARABLE" 15I4 267 52 594 VINE" I09 20 I5 19 OLIVE (# TREES) 55II Iz8I 248 2000 # PLOTS i6 i8 5 5

alone, land given over to pasture was included as a separate category. see Thonemann, op. cit. (n. i), 454. Except the inscriptions of Tralles: 59 Thonemann, op. cit. (n. 1), 454-7, has an extremely useful discussion of the terminology. See Kiourtzian, op. cit. are analysed in N. TO 'TO ITYPPAIf?N OPOS from Lesbos (n. 2), 219-21. The place-names Spencer, an archaeological niTYilAES: and epigraphical approach to a topographical problem', ZPE 12 (1996), 253-62. The structural difference between the Eastern and Western countryside is explained in Wickham, op. cit. (n. 3), 442; 58 On Lesbos

57

Chavarria and Lewit, op. cit. (n. 4), 16-17. 60 The largest contiguous properties are attested atMagnesia (a 7$-iuga estate) and on Lesbos, where farms of 430, and literary 305, and 294 arable iugera are listed, probably all of the same owner: IG XII.2, no. 76. Archaeological see J. J. Rossiter, settlement patterns into parts of Greece: evidence for rural villas and the intrusion of Western 'Roman villas of the Greek East and the villa inGregory of Nyssa Ep. 20', JRA 2 (1989), 101-10; Wickham, op. cit. (n. 3), 462-6; Sarris, op. cit. (n. 4), 121.

92

KYLE

HARPER

The inscriptions from and yield comparable results.It is Thera are similar in format thetotaldimensions possible to reconstruct of three distinct properties.
THERA ARABLE" VINE* OLIVE TREES # PLOTS

I4zA (Paregorius) I4zB (Attalou) 14zC

?in iugera

504 6I4 528

79 i 6z I20

554 I4z0 586

J0 I6 I7

The fragmentation of landholding is notable and has been emphasized by all com was built fromsmallerplots. It is possible that land from mentators.61 Each property was physicallyadjacent, and it is equally possible that landwithin a different districts so thattheinscriptions was further district may over-or understate theactual fragmented, degreeof fragmentation. and itconfirms Nevertheless,thepatternisbroadly indubitable, is not exceptional.The inscriptions that theEgyptianevidence for fragmentation from Thera and Lesbos also provide a valuablemeasure of elite land-allocation strategy. The estateson Thera specializedin wine production, while oliveswere an agricultural priority on Lesbos.62 was limited, and evenon theselarge Nevertheless,thedegreeof specialization properties, was thedominantagricultural arable farming operation.This would suggest thatelite-driven was a phenomenon specializationin staplecrops,at least in thisregion, thatoccurredalong the marginsof a landscapedominatedbygraincultivation. of theselandholders, butcomparative socialprofile data allow us tocontextualize thescale of these the properties.In everyrespect, Egyptian land registers from mid-fourth-century Hermopolis are theessentialcomparandaforthecensus inscriptions. These papyrirecord of hundreds of urban-basedowners,and theythusallow us to sayhow the land-holdings from Thera andLesbos compare to their the proprietors With only Egyptiancounterparts. on Lesbos (no. 78), the theexceptionof theincomplete third property Aegean landowners wealthiestof theproprietors atHermopolis,where would have rankedamong thevery arourai or more (= 2i9 iugera)belonged only to the top 3.6 per cent of landowners.63 attestedin the inscriptions held land elsewhere,they deserve Especially iftheproprietors to be classifiedas large-scalelandowners. The small sample preservedfromthese two islands is thus,itappears, a snapshotof thehighesttier of landholding. This knowledge is valuable as we turn to the secondaryregisters from Astypalaia, Tralles, andMagnesia. These documentsrecordedtax liabilityin terms of fiscalunits, around ioo iugeraof arable land,based on an apparent 'conversion'from large iugum, physical to fiscalunitspreservedat Thera; Duncan-Jones lateradduced several reasons
iuga. The schedule used to convert raw iugera into iuga is unknown. Jones favoured a holdings of i,ooo arourai (= I,095 iugera) were exceedingly rare and holdings of zoo The appearance of a senator in a patronymic on Thera (I4zB) is the lone clue to the

why the iugum must be smaller, closer to iz iugera of arable.64 Thonemann has now pro vided arguments for a conversion rate similar to the one advocated by Jones. Thonemann's case must be broken down into two separate claims, one about themethod of accounting,

61 op. cit. (n. i), 475. These figures generally follow Paton's tabulations for Recently emphasized by Thonemann, Lesbos with Erxleben's additions to no. 77 and Thonemann's emendations for Thera. 62 Thonemann, op. cit. (n. 1), 467: on Thera land was allocated to arable/vine/olive in the range of 80/17/3 per cent. On Lesbos, the ratio was c. 88/6/6 per cent. 63 See, for instance, Table 4B of Bowman, op. cit. (n. 6), 159. Other evidence from Egypt broadly confirms the data: see the properties analysed by D. Kehoe, Management and Investment on Estates in Roman Hermopolite Landowners Egypt during the Early Empire (1992), especially 75; J. Rowlandson, Social Relations of Agriculture in the Oxyrhynchite Nome (1996), 123. 64 op. cit. (n. 2), 199-210. Jones, op. cit. (n. 2), 50; Duncan-Jones, and Tenants inRoman Egypt: The

THE

GREEK

CENSUS

INSCRIPTIONS

OF

LATE

ANTIQUITY

93

the otherabout thespecific conversion scheduleinuse throughout Asiana. In a realbreak through, he has shown that the inscriptions deploy a relatively 'restricted rangeof frac tions... based on productsof primesno greaterthanfive'.65 With thisinsight, he argues thatthefractions attestedin theinscriptions shouldbe relatedto theconversion rates.66 He = ioo thus works out a 'satisfyingly straightforward seriesof conversionrates': i iugum iugeraarable = I5 iugeraof vine = 300 olive trees. The interpretation of theaccounting method is elegantand convincing, but the specific conversionschedule, particularly the 'large'arable iugum, presentsinsurmountable difficulties. In thefirst place, thereis no reason why thisspecific schedule (ioo arable/I5vine/300 trees) must be correct.It is based entirely on one of the inscriptions from Thera, which may preservethecalculationof raw intofiscal In three units.67 places, thestoneinquestion XK(?xakXo)4(uya). reads ?%0o1Gv Not only is theterm kephalozygacompletely unattested, thedispositionof the 'calculations' on thestone ishard to reconcile with the theory that theyare calculations.68 It seemshighlyplausible that theseare not in factconversions. Moreover,Duncan-Jones'scase fora small iugumisnot easilybrushedaside.He cites the contemporary work of Epiphanius, whose book on measurements claimed thata iugum was equivalentto iz.5 iugera of arable land; thisisa compelling sourceoverlookedin most impliesa rateof return on estates, in terms of solidi per iugum,thatrequiresthe lower were ioo iugera,theseestateswere producingunbelievablylow estimate; if the iugum revenues.70 Another reason todoubt thisschedule (ioo arable = i5 vine= 300 olive trees) is thatthe impliedtaxeson vinesand olives are implausibly high: thezo per centof land allottedtograpesand oliveswould bear over two-thirds of thetotaltaxburden.71 The smalleriugumalso allows fora more persuasivereconstruction of landholding at Astypalaia,Tralles, and Magnesia. The registers from Tralles andAstypalaia preservefive complete properties.72 Assumingforsimplicity thatthelandwas only inarable,and using
accounts.69 Duncan-Jones also invokes the letter of Valens to the proconsul of Asia, which

the rates of relative taxation implied in the Syro-Roman Lawbook, iugatio. Compare op. cit. (n. 17), where olive trees were taxed at a rate of 11 trees = 1 iugerum arable (not 3 trees = 1 iugerum). In the West, taxes were apparently assessed by area regardless of land-use: Jones, op. cit. (n. 8), 453. 72 I follow the superior editions of these texts provided by Thonemann, op. cit. (n. 1). Another property from I of the Tralles Column inscription, which seems to have totalled over 38 iuga and 86.5 capita (although it is in the visible portions, is so fragmentary that it is omitted impossible to tell if these are slaves, animals, or paroikoi) from discussion here but would have been one of the larger properties. Likewise, the fragmentary properties on Samos and Cos are omitted. The only property on Samos appears to have been an estate totalling 13 iuga. Thonemann, op. cit. (n. 1), 470-1, reports soon-to-be-published fragments from Cos pertaining to two estates, totalling 4.56 and 27.3 iuga respectively.

Thonemann, op. cit. (n. i), 466. It isworth noting that there are, however, a number of exceptions. For instance, in light of the common fraction 1/300, he proposes that 300 olive trees = 1 iugum. 67 IfThonemann's reading is correct (op. cit. (n. 1), 464, lines 4-5), since he reads k? rather than Kiourtzian's k?u, ? itwould allow for other interpretations, such as K?<|)aA.ai ? cov a category attested in the inscriptions. 68 For the new edition, Kiourtzian, op. cit. (n. 1), no. 142a. The first 'calculation' is preserved at the end of the third farm on an estate comprised of four farms. The second 'calculation' is at the end of the fourth farm on this same estatel The third 'calculation' comes at the end of the first farm on another estate. Jones and Thonemann hold that the first calculation is perfectly candid about this (preserved in line 3) is a total of the first four farms. Thonemann problem, op. cit. (n. 1), 464, at n. 104. 69 It isworth noting, too, that in op. cit. (n. 2), 201, although as he admits Cyprus lay in Oriens. Duncan-Jones, sixth-century Petra, a iugum was equal to 10 iugera of arable land: P. Petra 1.7-10. See also L. Koenen, 'Papyrology, Ptolemaic Egypt, and Byzantine Palestine', in B. Palme (ed.), Akten des 23. Internationalen Papyrologenkongresses (2007), 5-13, at 12-13. 70 op. cit. (n. 2), 206-7. The inscription implies that these estates were desirable (they are described Duncan-Jones, as opima), at a yield of just under 2 solidi per iugum. 71 Thonemann, op. cit. (n. 1), 467, argues that the 80 per cent of sown land bore only 30 per cent of the total 66

65

94

KYLE

HARPER

= I1 iugera,theproperties with theestateson compare a conversionscheduleof i iugum Thera and Lesbos.73

TRALLES ARABLE IUGERA ASTYPALAIA ARABLE IUGERA

Tatianus Kritias Latron Fulvius

6i6 25I 205 39

Heracleides

I29

of were closer to ioo iugera of arable, thelandowners We shouldnote thatiftheiugum Tralles and Astypalaia far outclassed the proprietorsof Thera and Lesbos. This is small ofThera and Lesbos included,in a rather forthe landowners intrinsically unlikely, Moreover, wealthy enough to own I52 slaves. and someone sample, a senator's relative Tralles andAstypalaia from why thefewattestedlandowners thereisno obvious reason Hermopolis, a of fourth-century shouldbe offthechartsincomparisonto thelandowners largeandwealthy town.74 valuable and it isespecially from Magnesia is also a secondaryregister, The inscription on overninety farms. Late Empire,foritpreservesinformation fora study ofwealth in the location;onlypartsofA, B, E, by thename of their The farms are groupedalphabetically, At first of amuch largerlist. glance, so thisisa randomsegment and anotherletter survive, over sixty-five of a broad base of urban landowners: the inscription gives the impression are attested more thanonce, and ownersare attested. Twelve of thelandowners different citiessuch as Ephesus, are registered. neighbouring fourteen out-of-towners, mostly from held landup and down thefertile whosewealthierfamilies aristocracy We thussee a regional were not of any special civic rank. Meander valley,but the majorityof local landowners
MAGNESIA IUGA 75 z6
25 20-I I2-15 8-II

# OF PROPERTIES
THIS SIZE

I 2
I 2 2 6

4-7
2-3 1-1.5 < I

6
I5 I4 i6

data from howwe understandthe Magnesia. If the deeply influences The sizeof theiugum large iugum (c. ioo iugeraarable) were correct,itwould make the top landownersat
rate represents only an order of magnitude. I think i iugum - c. 10 iugera is equally possible. for grapes and olives, larger if pasture/fallow. But the of iugera would be smaller ifwe accounted inscriptions of Thera and Lesbos show how little land was actually allotted to vines and trees, and it is unlikely that an inordinately large percentage of the tax burden (much less over half of it) fell on grapes and olives, so that the simplifying assumption of arable iugera should not radically distort the scale of the properties recorded in iuga. 74 Given the (apparent) size of the estates on Cos and Samos (13, 4.56, and 27.3 iuga), the small iugum would make the large iugum all of the attested owners from Tralles, Astypalaia, Cos, and Samos large landowners, whereas IV. would make them all exceptionally large landowners. Cf. Bowman, op. cit. (n. 6), Table This conversion The number 73

THEGREEK OF LATE INSCRIPTIONS CENSUS ANTIQUITY

95

Magnesia exceedingly prosperous, far richerthan theircounterpartsin Egypt or the as a singleplot, one of the Aegean. The largest at 75 iuga, would constitute, property, largest this holdings known from theeastern Mediterranean. With a small iugum, property would stillrankamong theverylargest The propertiesin the Hermopolite land register. at There were numerous small proprietors bottom end of the scale is also revealing.
Magnesia,

atHermopolis or faroutclass them. proprietors The median holding in the Egyptiandata was around io arourai (ii iugera).75 The median holdingat Magnesia was z iuga;this was,
depending on the schedule we use, on the order of zo or zoo iugera. Surely, the median landholder atMagnesia was not the sort of rich proprietor with zoo iugera. At the top and for the small iugum. The Magnesia

in the range of i-z

iuga. These

owners either closely resemble the small urban

forindividual bottomof thescale, the orderofmagnitudeimplied argues holdingsstrongly

register can also be used to analyse the stratification of

wealth in thisregion.
0.6
0.5

X=0.4
'u0.3 0.2

-I

0.1

Docile
FIG. i. Land ownership by decile.

measure of inequality were included,the would not appear ofEgyptianvillagers property

nearly so drastic.78 This raises a fundamental question about theMagnesia register: does it include the holdings of peasants and villagers, or is this only a list of urban proprietors? for our understanding of The answer to this question carries significant consequences the data. If the register originally recorded all taxable land in the territory of Magnesia,

and Mag across Roman history.76 The two most stratified samples come fromHermopolis nesia.77 But in the case of Hermopolis, Bagnall has noted a crucial caveat. The Egyptian land register recorded only urban landowners, exclusive of peasants and villagers. If the

of landed wealth Duncan-Joneshas gathereda numberof data-setson thedistribution

Roman world. The most stratified known from the reflects the thenthisinscription society
75 See Bowman, op. cit. (n. 6), 158-9. At a median of 200 iugera, these could hardly be family farms. A law of A.D. 342 even suggests that in some cities a property of 25 iugera was sufficient to qualify for curial service: CT 12.1.33. 76 op. cit. (n. 2), 121-42. Duncan-Jones, 77 ? At Hermopolis, the raw figure is .815 I calculate a Gini coefficient of .677 at Magnesia. extremely high:

in (see below). R. Bagnall, Bowman, op. cit. (n. 6), 150. This number should be adjusted downward 'Landholding Late Roman Egypt: the distribution of wealth', JRS 82 (1992), 128-49, at 131, fig. 2. 78 Bagnall, op. cit. (n. 77), 133-5, especially 138; a corrected Gini of .560, including villagers. The largest owners, such as the senators, held property on a larger geographic scale, with estates in other cities and provinces. So the it understates the largest owners' holdings, but overstates inequality of graph is distorted in two directions: distribution in this delimited space by excluding rural proprietors.

96

KYLE

HARPER

would requiresuch a readingof the large conversionrate for the iugum (ioo iugera) exceed plausible bounds. The inscription, lest the implied size ofMagnesia territory scenario seemsmore coherent:a small iugum,a register that includedonly alternative ofwealth. This case issupported by and severe but credibleinequalities urbanproprietors, which only Hermopolite registers, with otherdocuments,such as the formal comparison InEgypt,villages were assignedthetaskof of townresidents.79 recordedthetax liabilities and rendering thesums to thetowns;a sixth-century collectingtaxeson theirinhabitants inscription fromtheprovinceof Caria shows thatvillages inAsia Minor could also be grounds, thecity.80 On administrative liable forcollectingtaxesand paying themthrough in thecity's territory.81 listedall proprietors then,it isunlikelythatthe Magnesia register register also allow us to project The assumptions of a small iugumand an urban-only ofMagnesia. plausible ordersofmagnitude forthepopulation and overall civic territory unique We have c. 8 per cent of theoriginal list.The extantportion includessixty-five and fifty-one Magnesians. Thiswould leadus topro landowners: fourteen out-of-towners in theentireterritory of thecity. The urbanpopu jectaround 6ooMagnesian proprietors at I2,500 souls.82 This estimateisquite pos lationofRomanMagnesia has been estimated was exceptionally high; it is easier to believe unless theurbanizationrate siblytoo large, in a city population in therangeof 5-i0,000,with a ruralpopulation at severaltimesthis in theterritory Six hundredis surelytoo small as a numberof totalproprietors number.83 - but itseemsplausible estimates of thepopulationare correct -especially ifthehigher around thedistrict as a number of urbanowners, inparticularif manywere small-holders of thecity itself.84 calculationscan be carriedout forthe totalcivic territory Similarorder-of-magnitude at 47i kmzof cultivableland.85 If we have c. 8 per of Magnesia, which has been estimated cent of the register, portionwould implyan the 366.8 recorded iuga on the surviving iuga. If I iugum= ioo iugeraarable, then the total original total of 4,585 registered If,however, would farexceed thesize ofMagnesian civic territory.86 registered property were around iz iugera,it impliesthaturban landowners held 55,ozo iugerain the iugum of total.This would constituteroughly z9 per cent of all taxable land in the territory Magnesia. There are severalpossible sourcesof error in thiscalculation,not tomention owned. But it is note we have no way of knowinghow much landwas imperially that ownershipatMagnesia of 29/71 is almost precisely worthy that a ratioof urban/rural In sum,by assuming Hermopolite nome, 30/70.87 equivalent to theratiocalculated in the a small iugumand an urban-onlyregister, the yields an imageof Magnesia inscription which both thescaleof individual and theoveralldistribution of land wealth in properties can be closelycompared with contemporary data fromtheothercensus inscriptions and the Egyptianpapyri.
list recorded only urban owners. Jones, op. cit. (n. 2), 54-5, argued on this logic that theMagnesian For Egypt: Jones, op. cit. (n. 8), 454. For the Carian example, see D. Feissel, 'Un rescrit de Justinien d?couvert ? Didymes', Chiron 34 (2004), 285-365. 81 There is another internal argument: virtually none of the smaller holdings have registered capita. The in the next section, but the immediate conclusion must be of this distribution of labour are discussed consequences that these do not represent village or peasant households, which would have been liable for the head taxes of their 80 family members. 82 R. M?rchese, The Lower Maeander Flood Plain: A Regional Settlement Study (1986), 317. 83 M?rchese, (over op. cit. (n. 82), 317, employs a rank-size approach which implies very high rates of urbanization rates in B. Frier, 'Demography', CAH2 30 per cent). Cf. the population densities and urbanization (2000), 787-816. 84 The Hermopolite data, crucially, exclude the pagus nearest the city, but see Bowman, op. cit. (n. 6), Table VI, and the exponentially greater number of owners in the 5th, 6th, and 8th pagi, the next closest areas. 85 op. cit. (n. 82), 317. M?rchese, 86 If the tax liability were only in arable, itwould imply 458,500 iugera, a number which could be reduced by folding some of the liability into vine-land and olive trees (indeed, it is only by putting over half the tax burden on for all land in the territory, that the scale becomes 79

vines and olives, and by assuming that the register accounted in the large-iugum scenario). manageable 87 See Bagnall, op. cit. (n. 77), 137.

THE

GREEK

CENSUS

INSCRIPTIONS

OF

LATE

ANTIQUITY

97

Finally,the Magnesia inscription providesa social profile of the landowners.

CURIAL CURIAL Paul Tuchichos Mandrogenes IUGA 8.78 1.05 2.I 7.2 AND SENATORIAL CAPITA 5-.5 LANDHOLDERS AT MAGNESIA IUGA z.65 x 75.I5 5z.88 CAPITA Iz.z8

SENATORIAL Priscillianus Aristocleia Farm C3

Pollio
Phanius Heracleides Farm Farm ii i3

2I.I5
4.07 9.5 .37 2.I5 I.58 5.02 I0.36

Capitolinus
Hermonactiane Eutyches

.74
3.I7 I.I5

TOTAL

49.I7

29.4I

TOTAL

8z.86

65.i6

The riseof a new aristocracy serviceisa classic theme based on imperial ofLateAntiquity. This class graduallydisplaced an older, locally-rooted municipal gentryin theEast. To was a re-labelling ofpre-existing someextent, this process elites,as theentrenched nobility But the imperial was able tomaintain its statusundernew circumstances. centrerecon matrix of loyalties, figured the and resourcesthrough privileges, which theupper-class maintained itsstatus. The census inscriptions provide a static imageof thisprocess in a of the concrete which thereis little evidence. TheMagnesia inscrip sub-region Empire for tion shows that the imperial, was intruding on the senatorialstratum municipal aristo reserves of economic clout.88 cracy,thoughthe latterretainedsignificant The inscription were new men with new money or former cannot tellus if the senatorial landowners status. decurions Either castsvivid light on the promotedto imperial way, the inscription which is so audible in thesourcesof theperiod. plightof thetowncouncils
IV LABOUR IN THE CENSUS INSCRIPTIONS: DENSITY AND STATUS

The census inscriptions intothedeployment also allow quantitativeinsights of labour in the Late Roman countryside. The study of labour in Roman antiquityisbeset by intract able problemsin theempiricalrecord: what evidencesurvivesis typically vague,unrepre or ideologicallycoloured.89 The random,documentary data of the census sentative, inscriptions can be a corrective, and they deserve wider notice than they have received. Nevertheless,caution is in order. Jonesdrew two conclusionsabout labour fromthe He arguedfora lowdensity of labouron thelandand an overall ratioof free inscriptions. men to slavesaround 5:I. These claims are notmethodologicallysound,and itwill be a of thedata. Despite theconsiderableuncer goal of thissection to emphasize the limits tainties, however, solid evidence about the use of labour on elite-ownedland can be rescuedfromthestones. a lessurgenttheme now than in I953. The The density of labouron the land is surely
bleak image of a landscape scarred by agri deserti is simply no longer tenable in light of

88 As Thonemann, op. cit. (n. i), 473, points out, the only obvious 'absentee' landlord is at line a4, a farm owned but declared by (an apparent slave manager) by Quadratus Syneros. 89 11 (2002), 195-213; See, e.g., U. Roth, 'Food rations inCato's De agri cultura and female slave labour', Ostraka W. Scheidel, 'The most silent women of Greece and Rome: rural labour and women's life in the ancient world', G&R 42 (1995), 202-17 ana1 43 (1996)> 1-10; W. Jongman, 'Slavery and the growth of Rome. The transformation in C. Edwards and G. Woolf of Italy in the second and first centuries BCE', (eds), Rome the Cosmopolis (2003), 100-22.

98

KYLE

HARPER

of which showsa dense and vibrant pattern theever-accumulating archaeologicalevidence in the More profoundly, theeffort toextract measurements settlement Late Roman East.90 recordonly of labour densityfromthesedocuments is flawed.The census inscriptions - workers for was liable. Jones whose capitation tax the landowner registered labour recognizedthat free, contract-based land-leasing would not be detectible, but he never His account shows the theless proceeded to diagnose a criticalshortageof farmlabour. was not registered, subtle influence of thecolonate, forhe seems to assume thatiflabour thatland itprobablydid not exist. Yet revisionist work on thecolonatehas demonstrated Late Empire; thepos leasingremained a prominent element of theruraleconomyintothe sibility labour isentirely realistic and indeedfinds of non-registered supportin theinscrip of labourwas inseparablefrom tions (see below).Moreover, Jones'scase fora shortage was on theorderof ioo iugera. his belief that the iugum The downward revision of the on the land.9' of labourdensity iugum drastically alters theimpression Ifwe begin the investigation explicitlyas a search forpatternsof registered labour, The secondary from interesting results emerge. registers Astypalaia,Tralles, and Magnesia But aswith theiugum, all offer data about theratiosof registered iuga. capita to registered the conversionscheduleused to convert individualhumans intounits of tax liability, We seem to be fortuitously capita, is uncertain. well-informed by a contemporary law A constitution ofA.D. 386 adjusted thenumberofmen and schedule. about theconversion of the Ponticdiocese; itclaimed that several previ women ina caput throughout provinces women had equalled one unitof head-tax liability, It is thecaput.92 ouslyoneman or two were not quite so simple, but as a best estimate,thescheduleof i man = clear thatthings z women = i caput seemsplausible.93 This yields,forinstance, thedensity of labouron a mid-sizedestateon Astypalaia:94
ASTYPALAIA Fiscal Physical LAND IO.8 iuga 130 iugera LABOUR I3.I3 capita I3-z6 workers

measurementsto bemeaningful, we would have toknow theaveragenum For thedensity in berofworkersneededper iugerum. Our knowledgeofmanning-ratios Roman farming writersand landgrants. These yielda rangeof average labour comes fromtheagricultural of arableperworking somewhere between5 and 25 iugera densities, male, but thesefigures must have varied significantly scaleof estate, by region, natureof cultivation, mobilization At theleast,they can providea roughideaof labour requirements. of femalelabour,etc.95 With one worker (presumably including women) forevery5-IO iugera,thedensityof

90 to Justinian', in A. Dunn, W. Bowden, 'Continuity and change in theMacedonian countryside, from Gallienus L. Lavan and C. Machado (eds), Recent Research on the Late Antique Countryside (2004), 535-86; S. Kingsley and M. Decker in the East Mediterranean (eds), Economy and Exchange (2001); Jameson, op. cit. during Late Antiquity (n. 3); S. Alcock, Graecia Capta: The Landscapes of Roman Greece (1993). Not to mention that agri deserti should terms: C. R. Whittaker, I. Finley (ed.), Studies in be interpreted in fiscal, not demographic, 'Agri deserti', inM. Roman Property (1976), 137-65. Most recently, C. Grey, 'Revisiting the "problem" o? agri deserti in the Late Roman Empire', JRA 20 (2007), 362-76. 91 As noted already by Duncan-Jones, op. cit. (n. 2), 207. 92 CT 13.11.2 (a.d. 386), issued to the Prefect of the East. Equivalent fiscal schedules were apparently in use for Asiana and Pontica: CT 7.6.3 (a.d. 377). See in general Jones, op. cit. (n. 2), 50. 93 The Astypalaia inscription included a column of 'human capita'. But the figures recorded in this category include bizarre fractions like 1/10, 1/200, 1/300, and 1/750. See Thonemann, op. cit. (n. 1), 477-8. 94 I multiply the capita by 1-2, to reflect an all-adult male or all-adult female population. The truth should be somewhere in between, unless large numbers of children were included as small fractions. 95 M. Spurr, Arable Cultivation in Roman Italy, c. 200 B.C.?c. A.D. 100 (1986), 133-46; R. Duncan-Jones, Roman Economy: Quantitative Studies (1974), 327; Jones, op. cit. (n. 2), 56. The

THE

GREEK

CENSUS

INSCRIPTIONS

OF

LATE

ANTIQUITY

99

registered labouronAstypalaia falls within therange of sufficiency. On this estate thereis little need to adduce thepossibility of extensiveleasing thatdid not include the land owner's responsibility forhead taxes. Similardata can be extractedfrom aremore vari Tralles, althoughin thiscase there Tralles inscription, likethatatAstypalaia, is that ables.The advantageof the we are able to see the tax liability foran owner's entireproperty in thiscase, fourowners.But unlike at Astypalaia, the labellingof human capita is not limpid.Some lines total the capita of 'animalsand slaves', others the capita of 'animals', and othersplain capita. Assuming,as did Jones,thatthefirst represents slaves,thesecond represents animals,and thelastrepresents paroikoi,we would see thefollowing results.96
TRALLES SLAVES PAROIKoI LAND

Fulvius
Tatianus

8-i6

4-8
56-iiz

39 iugera
6i6 iugera

Kritias Latron

2-4 4-8

9-I8 0

z5i iugera 205 iugera

if uncertain,results. With the exception of The data fromTralles yield interesting, was supplied with enough labour to be at or near suf Latron's property, each landowner labour.Latron, perhaps, rentedsome of his ficiency, without theneed forunregistered may have been not only land,and it is easy to imaginethathis slaves workers,but effect most non-slave Tatianus can claim tohave theestate with the agentsormanagers.97 ively but it isnoteworthy labourersinanyof the Greek census inscriptions, thatevenhe could - a pheno coreworkforce claim around a dozen slaves,perhaps as part of a permanent menon attested elsewhereinLate Antiquity.98 a challenging, once again represents sourceof The Magnesia inscription yet significant units (capita), This documentrecordedlabouronly in fiscal without specify information. measures of ing thestatusof the workers,norwhetheranimalswere included.It reveals a structural of labour. At the which at first labourdensity glance seem to reflect shortage same time,thisdocumentrevealsa pattern which offers clues to thedynamics important of registering labour in theLate Empire.AtMagnesia the tax dues were organizedgeo so we do not know the owner.The graphically, whole tax liabilityforany individual was not spreadevenlyacross the land.99 labour make itclear thatregistered inscriptions on with scattered As we might expect holdings, itwas normal to concentratelabourers certain as domicilesforthe workers.Jonescalculatedaround 263 plotswhich functioned in which both categoriesare visible.'00 iugaand zIz capita forthepartsof theinscription This indicates a lowerdensity of registered labour thanelsewhere. Possibly this is a fluke was domiciled),or possibly there was a we lack registered sites where the workforce (if of rural labouratMagnesia. But perhaps some landowners atMagnesia leased shortage fortheir land tovillagersand peasantswithout assumingresponsibility capitationtaxes.
96 Some entries recorded ?o v K(e<|>aA.ai), others oo?taov Kai ?obcov K(e<|>aA.a?), others simply K(e<|)aA.a?). The op. cit. (n. i), 458, that the capita registered as 'slaves and livestock' were possibly mobile argument of Thonemann, work-gangs, while the plain capita were assigned to a plot (but not indicating legal status) is possible and throws into doubt the arguments of Jones for the ratio of free to slave labour. 97 'The Roman slave supply' (forthcoming). For this sort of relationship, see Also suggested by W. Scheidel, P. Oxy. 2474 (third century, possibly late). 98 e.g. P. Lips. 97 (a.D. 338); CT 9.42.7 (a.D. 369); Basil of Caesarea, Homil?a in divites 2.2 (Ed. Y. Courtonne

(1935), 46-7). 99 On Lesbos, the slaves were concentrated on no. 76, line d6 and no. 78, line C2. At Tralles, one location (line 22) farms a6, ai2, b7, b8, bi2, bi8, C3, en, and g2 held concentrations had 15.56 capita on .95 iuga. And atMagnesia, of labourers. 100 op. cit. (n. 1), give a slightly higher number of capita. Jones, op. cit. (n. 2), 54. The new readings of Thonemann,

100

KYLE

HARPER

is theonly example in theentire Significantly, the Magnesia inscription dossierwhere numerous petty landholders are recorded. The inscriptionsfromLesbos, Tralles, Astypalaia and Thera are heavilybiased towards largecurial-scaleproperties. Perhaps small and middling landowners were themost likelyto lease theirsmall plots in the tovillagersand peasants. If suchvillagers or peasantsowned their countryside own land, was not responsible thentheurban landlord forthecapitationtaxes,and thelabour would not be registered under thelandlord'sliability.101 In support of thisreconstruction, liability forcapitation taxes atMagnesia was strongly on threetypesof land: the concentrated largest properties, thefarms of citizens with highstatus, and thefarms of landowners from other towns, who were presumablylargelandowners.102 This would argue thatfree, con tract tenancyand the fiscalregistration of dependent workers co-existed,but on two planes. There was both a sectorofmiddling landlordsleasingextra fieldsto ruralsmall holders and a levelof largerlandowners with their own (fiscally 'dependent' that is, registered) labour force. The inscriptions thatallow comparisonbetweenregistered workersand registered land do not seem to show an acute shortage of labour.103 What thecensus inscriptions reveal is a countryside on elite-ownedland, givenover to intensive polycultural exploitation with a labour forceconstituted of a mix of tenants, slaves,and free dependents. Of course, the in relationships registration of free dependents(paroilkoi) thatincludedthe landlord's forcapitationtaxes does appear as a widespread practice in thisregion. responsibility Most of theproperties are curial-scalefortunes, recorded and only at Magnesia do we get a fuller picture thathelps us understandhow smallerowners could continue to use even as largeowners became increasingly contract-leasing responsiblefor thecapitation taxesof their workers.This divergence how largelandowners helps us tounderstand may have been able to exploitpublic laws restricting movementfortheir fiscal own benefit.104 In Jones'sanalysis,theshortage of labour was one part of a story whose main drama was theriseof thecolonate.The colonatenot only ledhim tounderestimate the possibility of non-registered labour, it seems to have encouraged him to downplay the striking evidenceforslaveryin these documents.105 The discovery of thenew fragment from Thera, over i50 slaves on a singleestate,demands that attesting we paymore attentionto the positive role of slaveryin thissample. Moreover, theRoman colonate has come under

labour on the farm. 104 See especially J. Banaji, 'Lavoratori liberi e residenza coatta: il colonato romano in prospettiva storica', in Lo zum sp?tantiken Kolonat in der kaiserlichen Cascio, op. cit. (n. 11), 253-80; D. Eibach, Untersuchungen unter besonderer Ber?cksichtigung der Terminologie Gesetzgebung: (1980). 105 Jones, op. cit. (n. 2), 57, claimed that, excluding the gangs on Lesbos, slaves were 12-13 Per cent o? the rural ? and this excluded the Slavery at 12-13 Per cent of the rural population would be a remarkable figure population. two samples available to him where slavery was most prominent. Careful study of the Roman slave system has downsized the plausible extent of slavery, so the discovery of prominent agricultural slavery in this region of the is immediately striking. See W. Scheidel, (outside the traditional heartland) Empire 'Quantifying the sources of slaves in the Early Roman Empire', JRS 87 (1997), 156?69; W. Scheidel, 'The slave population of Roman Italy: speculation and constraints', Topoi 9 (1999), 129-44; Scheidel, op. cit. (n. 3); Jongman, op. cit. (n. 89), 100-22.

format, and other districts on this stone do not list names, making it probable that all of these properties belonged to one owner. The named individuals were probably his actores, conductores, or tenants. This stone alone records land of different classes, a clue that the taxes on these particular farms needed to be carefully specified. The names ? ? not the total cf. Bowman, op. cit. (n. 6), 142 probably signal a tax assignment separate from legal ownership

101 CT 11.I.14 (A.D. 371). 102 This pattern is difficult to quantify precisely, but the distribution is overwhelming. the twenty-one Among largest registrations of capita (which account for nearly all the total capita), only three owners were apparently not senatorial, curial, from other cities, or owners of multiple estates. 103 The only other place where Jones found a shortage of labour was no. 79 from Lesbos. In this inscription, four of the properties included a personal name underneath the heading for each district. This is the only stone with this

THE

GREEK

CENSUS

INSCRIPTIONS

OF

LATE

ANTIQUITY

I0I

In a seminalarticle, Late Roman labour relations.106 critiqueas a paradigmfor withering on coloniwas fiscalinnatureand could not anchor a Carrie showed that the legislation as thedominant labour system.107 narrative of transition 'from'slavery'to' tied tenancy and it led in land-leasing, His work allowed scholars to recognizethedeep continuities thatcoloniwere not only tenants, but oftenpaid recognition to the striking ultimately has agreed of therevision Ifnot everyone with the more radicalversions workers.'08 estate on thecolonatehas at leastpreparedus forthe of criticalscholarship istview, a generation attestedin theinscriptions. diversity of thelaboursystem record liabilityforcapitation taxes on two classes of rural The census inscriptions paroikoiwith thestudiously slavesand paroikoi.Throughout I have translated labourers: inorder toavoid prejudging how theselabourers were organized, vagueword 'dependent', or estateemployees.In a strict as tenants sense, the inscriptions tellus only forinstance was responsible capita were of freelegalstatusand thatthelandowner fortheir thatthey with If theinscriptions date to thecensusofA.D. 37I, thenit ispossible to say tiontaxes.'09 own andwho were thus were labourers who owned no landof their thatthese certainty estate.The inscriptions only record under thelandowner's inscribed on thefiscalregister - a fact of Jonestoextractthe which immediately disqualifiestheeffort registered labour in therural documents. Moreover, population fromthese proportion of slavesand tenants thatour documentsare biased towards theupper tiersof it is imperative to remember as a whole. Never of thecountryside wealth. They are thusnot representative landed of theorganizationand statusof theless,it ispossible to obtain a controlledimpression establishedbeliefs forceus to reconsider labouron elite-ownedland,and thesefindings period and region. about theroleof slave labour in this data about the legal statusof the labourers: Three of thecensus inscriptions preserve problematic Tralles, Lesbos, and Thera."0 FromTralles we have a small and extremely of fiscalunits. Jonesargued that the ratioof freetenantsto sample,preservedin terms The ratio is 5:i, so thatslaveswould be slaves could be extractedfromthisdocument. thenumbers may be com labourforce. Unfortunately around I7 per centof theregistered of animals,and, as Thonemannhas argued, it isnot clear that promisedby theinclusion freeparoikoi. The glimpsesof the the categoryof plain capita necessarilyrepresents The largefarm owned by thedecurion atTralles are onlyusefulqualitatively. countryside on the order withworkers who were likely but employed free, Tatianuswas well-equipped whichwas worked by a few, held a small farm of eightto sixteenslaves.Fulvius,a priest, a larger Kritias had perhaps two to fourslaves to complement freelabourers. apparently workers.This small staff of free workers.Latron had four to eight slaves but no free in terms of at leastpoints to a diverselaboursystem, with all of itsuncertainties, sample, workers. and thestatusof its itsownership patterns
un mythe historiographique?', Selectively: Eibach, op. cit. (n. 104); J.-M. Carri?, 'Le "colonat du Bas-empire": (ed.), Opus 1 (1982), 351-70; D. Vera, 'Forme e funzioni d?lia rendita fondiaria nella tarda antichit?', inA. Giardina Societ? romana e impero tardoantico 'Padroni, contadini, contratti: realia del (1986), vol. 1, 367-477; D. Vera, in Lo Cascio, colonato tardoantico', op. cit. (n. 11), 185-224; W. Scheidel, 'Slaves of the soil', JRA 13 (2000), in the Roman Empire (2007), especially 163-91; Grey, op. cit. (n. 12); 727-32; D. Kehoe, Law and Rural Economy B. Sirks, 'The colonate in Justinian's reign', JRS 98 (2008), 120-43. 107 Carri?, op. cit. (n. 106). 108 For continuities in land leasing, see the work of Vera, op. cit. (n. 106). For the recognition that not all coloni were tenants, see Banaji, op. cit. (n. 4), 209?11; Sarris, op. cit. (n. 4), 128-9; 0/ 11.48.19 (Anastasius). 109 See Sarris, op. cit. (n. 4), 151; Kiourtzian, op. cit. (n. 1), 225. For a fourth-century definition, Basil, Homiliae super Psalmos 14.1 (PG 29, col. 252): TQarcfip o? Tc?poiKoi, ?Moxp?av ?K|xia0o?^i?voi yfjv, Ttpo? x? ?oOATina xo? ?K???cokoxo? YficopyoOai xf]v xc?pav. 110 whose There are other important clues that slavery was regionally significant. A proprietor at Magnesia, op. cit. properties were named Barbaria and Barbariane, held little land but enormous capita liability. Thonemann, (n. 1), 474, has made the suggestion that these were slaves used to quarry emery, an important local resource. In Chios two census blocks survive. The second block listed the owner's liability in at least ten different villages. Under each village the liabilities were listed in iuga and 7iap(o?K(ov) K(?<)>a?uxi), but also in oou^(cov) K(?(|>aA<ai) and ?q)(cov) not filled inwith numbers. See also n. 112, below. K(?(j)aA,ai). Unfortunately, on the stones from Chios, this template is 106

102

KYLE

HARPER

Lesbos and Thera, recordedin raw head counts,we are on With theevidence from neat ratiosof freeto slave stateof theevidencethwarts firmer ground.The fragmentary roleon unequivocallythatslavery played a structural labour,but thestonesdemonstrate sowe All fourof thestonesfrom Lesbos are incomplete, elite-ownedland in thisregion. forany landowner. This iswhy Jones removedtheevidence from lack the total liability to workers.Though it is impossible Lesbos incalculatingtheoverall ratioof slaves to free was composed of slaves, it is importantto saywhat proportionof the labour force on Lesbos. Two of thefourstones forruralslavery acknowledgetheconcretetestimony stone recordsland in sixteen The first conserveinformation about the labour force. plots butmentions labouron only one.'11The farmin that localewas sizeable: 9i iugeraof oxen, fifty arable land, zo iugeraof vineyard,352 olive trees,land forpasture, twenty on thisstone,theownerhad no slaves. In theotherdistrictslisted sheep,and twenty-two slaveor free.It is likelythatsome of his landwas worked by tax liabilities forlabourers, the slaves, some let out to tenantsresponsiblefor theirown personal taxes,or some on the lostpart of thestone.The onlycertainty about this worked by labourersrecorded slaves.112 landowner'slabour forceis thatit includeda gang of twenty-two Lesbos mentionspropertiesin five different Another stone from plots. In one of these Yet thefarmin this slaves.113 district places, theownerwas liable fortaxeson twenty-one of arable land, some vines,and 132 olive trees.Itwas a smallplot included only 5 iugera more than house theslaves, who presumably worked on the other whichmay havedone little to slave labour on Lesbos,we shouldtry ratios toestimate, If we cannotextract of free farms. were integral ormarginal toagricultural whethertheslaves as away of assessing exploitation in thissample, of theregistered land theslavescould haveworked.Using what proportion ofDuncan-Joneson thefarmsrecordedin IGXII.Z, no. 76: the manning-ratios
LABOUR RAW AMOUNT REQUIREMENTS, MANNING-RATIO PROPERTY i, LESBOS REQUIRED LABOURERS

arable 15I4 iugera vine I19 iugera IIO iugera olives

man per 25 iugera 8 iugera man per man per 25 iugera

6o.6 I3.6 4.4 78.6 labourers

with thisexercise. It ispossible thatfallowarable There are a numberof complications of taxableproperty.'14 on was counted in theregister Moreover, the manning-ratios rely was the norm for monocultural farms.115 treatisesdiscussing imaginary Polyculture a strategy which reducedclimaticandmarket riskbut alsomade Mediterranean farming,
IG XII.2, no. 76, c-d. 112 In four places, individual names, listed with animals, are This stone presents another interesting problem. recorded between villages. Jones thought these were owners of animals grazed on a third party's land. But it ismore and Cyzicius, shepherds. Their names, Elpidephoros, likely that they were servile or dependent Philodespotos, Aristotle, are consistent with a group of slaves, though names are rarely probative. Aristotle had a patronymic, so is as convincing as name evidence gets. Cf. Bagnall, possibly he was free or freed and the others slaves. Philodespotos op. cit. (n. 45), 126. Parallel evidence for slave shepherds in Late Antiquity is not lacking, e.g. Augustine, Contra Cresconium 7.22 (Ed. Rochefort 1.30.35 (Ed. M. Petschenig, CSEL 52.2 (1909), 355); Julian, Orationes (1932-1964), inNorth Africa: B. Hitchner, evidence for estate-based pastoralism vol. 2, 75). Archaeological 'Image and reality: in the Tunisian High Steppe', in J. Carlsen in the Roman Empire the changing face of pastoralism (ed.), Landuse (1994), 27-43. 113 IG XII.2, no. 78, c 114 On the taxation of fallow land, Bagnall, op. cit. (n. 45), 116. The register from Lesbos, however, uniquely recorded pasture. On the relationship between fallowing and pasture, see Grey, op. cit. (n. 90), 369-70. Cf. CT 9.42.7 (a.D. 369), a complete fiscal valuation of a confiscated estate included what was currently and had been 111

previously cultivated. 115 op. cit. (n. 95), 327, is the best discussion. Duncan-Jones,

THE

GREEK

CENSUS

INSCRIPTIONS

OF

LATE

ANTIQUITY

I03

efficient use of thecalendar.116 Finally,the manning-ratios assumeworking men,whereas the slave counts included women. But it is likelythatslavewomen were worked hard, because slave statusoverrodeculturalnormsof genderedlabour.This has been a major sourceof efficiency forslavery.117 Factoringall these considerations, thetwenty-two slaves a quarter to a third of this landowner may have been able to handle, roughly, of the necessary work on thevisible registered land. The calculationscan be repeatedforIGXII.z.78:
LABOUR REQUIREMENTS, PROPERTY 2, LESBOS RAW AMOUNT MANNING-RATIO REQUIRED LABOURERS

arable 52 iugera
I5 iugera vine 5 iugera olives

man 25 iugera per

8 iugera per man 25 iugera per man

z.o8
I.88 o.z

4.i6 labourers By any estimate,the twenty-one slaveson this property provided fartoomuch labour which are visible. Most likelytheownerheld other forthepart of theowner's holdings which were lost in thedamage to thestone,and it is evenpossible thatthesame farms on both stones. We can only say thatslave labour,ina smallbut ownerheld all thefarms to theexploitation randomsample,appears structurally integral of elite-ownedland on Lesbos. which offers The inscriptions from Thera are thefinal set information about thestatus in the same of agriculturallabourers. Unhappily, these inscriptions defyquantification are fragmentary, sowe do not know thetotal liability of way as thosefrom Lesbos: they who had two slaves, three free any landowner except theheirsof Paregorius, dependents, and probably leased some of theirland.Four other significant but inscriptions survive, aggregate ratiosof freeto slave labourare unattainable.
THERA I4zb LAND 6I4 iugera arable i6z iugera vine i,4z0 olive trees I42C 528 iugera arable
I20 iugera vine

FREE ?

SLAVE

586 olive trees I42d New ? i6 ? I52+

of the total labour forceon Thera was Although it is impossibleto saywhat percentage The attestation remarkable. composedof slaves,thenew inscription is,by any reckoning, owner ranksas the most concrete, credibleartefact of over I5z slavesbelongingto a single
116 op. cit. (n. 95), 36?7. Even in the U.S. South, cash-crop farms like cotton plantations allowed less Duncan-Jones, than half the year to cotton production. See G. Wright, Slavery and American Economic Development (2006), 86. 117 see J. T. Toman, Scheidel, op. cit. (n. 89), 213. On the efficiencies of using female slave labour in theNew World, in Economic History 42 (2005), 310-23. Cf. Stobaeus, 'The gang system and comparative advantage', Explorations (second century A.D.), Anthologium 4.28.21 (Ed. O. Hense and C. Wachsmuth (1894-1912), vol. quoting Hierocles koivcove?v ii?vov xa?? 0?pa7caivai?, ?XXa Kai x v aXX&v 5, 699): cocjxE \ir\ xfj? xataxcjia? ?pycov x v ?7tav?pox?pcov.

104

KYLE

HARPER

Roman Empire.This property must have been fromtheentire of large-scale ruralslavery largerthan the farmsassociatedwith gangs of slaves on Lesbos, themixed estate of Tatianus atTralles, or thelargesenatorial domain at Magnesia. The imageof thecountry side from Thera, where large landownersspecialized inwine productionand employed more diversity to thelaboursystem attestedin the Greek of slaves,adds even vastnumbers census inscriptions. that Despite their limits, thecensus inscriptions with a highdegreeof credibility, testify, to the elite land-use strategy was structurally of the fourth-century slavery important in thisregion. The presentation of Jones, who was, tobe fair, without aristocracy working of thedramaticnew fragment, has allowed someof the most valuable evidence thebenefit Roman period to be marginalized.Finley, forruralslaveryin the Whittaker, and others have longemphasizedthattheliterary evidenceforslaveryin the Late Roman countryside is as abundantand credibleas foranyperiod of antiquity."'The evidenceforslaveryin A variety Mediterraneanhas beenparticularly theeastern of authors,including neglected. make plausible statements Basil ofCaesarea, Libanius, and John Chrysostom, to theeffect in thefourth was important At the least, the inscrip thatagricultural slavery century.'19 tionssuggestthattheobservations of these authorshad a basis in reality. As documentary evidence foragricultural slavery,thereis littleinRoman historythatcan match these Ifonewanted tomarginalize theevidenceforslaveryin the Greek census inscriptions, it on thebias of thesample towardsislands.It ispure accident mightbe possible to insist with good data about labourare fromislands, thattwoof theinscriptions but it is inter microeconomicparameters of a slave estate.120 From the esting.Islands could fosterthe an island slavemanager's perspective, mighthave reducedtheriskof flight. i9 km long, Thera was a natural cage for the slaves forced towork its land. Secondly, an island environment market. With a limited could aggravatethe risksof theopen labour catch ment area, an islandcould have an inflexible of extra labour.Slaverythusinsulated supply costsof finding Of course, these thelandowner against theriskand transaction labour.12' are emphases, not exceptionaltraits, determined features by islandecology. Mediterranean islandsare not abnormalsocial habitats,and Lesbos, afterall, is at i,6oo kmznot a small Itwould be unwarranted topress thecase of islandexceptionalism island.122 veryfar. enhanceour knowledge Such quibbles shouldnot diminishthefactthattheinscriptions of how slavery toRoman agriculture. In theproperties contributed documented,slavery took a varietyof forms:a few slaves on small farms, mid-sized gangs in singleunits,a core ofworkerson a farm workedmostlyby freelabour,and truly permanent vast slave of slavery The prominence based estates. emphasizesthattheLate Roman labour system cannotbe describedin terms of a struggle betweenslavery and tenancy as twocompeting modes of production.123 This claim builds on Carrie's original insight that thecolonate
inscriptions.

inM. Finley (ed.), Classical 'Circe's pigs: from slavery to serfdom in the later Roman world', Slavery (1987), 88-122; M. Finley, Ancient Slavery and Modern Ideology2- (1998; orig. 1980), 191-217. 119 In Matthaeum InMatthaeum 24.11 (PG 57, col. 63.4 (PG 58, col. 608); John Chrysostom, John Chrysostom, In acta apostolorum in divites 2.2 (op. 32.2 (PG 60, col. 237); Basil of Caesarea, Homil?a 319); John Chrysostom, cit. (n. 98), 46?7); Basil, Homil?a in illud: Attende tibi ipsi 5 (Ed. S. Rudberg inmartyrem (1962), 31); Basil, Homil?a (Ed. R. Foerster, Opera (1903-8), vol. 4, 370); Libanius, Or. 47.28 Julittam 1 (PG 31, col 237); Libanius, Or. 62.46-8 (Ed. Foerster, vol. 3, 417-18); Libanius, Or. 14.45 (Ed. Foerster, vol. 2, 103). 120 I. Finley's Ancient Slavery and Modern cf. B. Shaw, '"A Wolf by the Ears": M. Ideology in historical context', foreword toM. Finley, Ancient Slavery and Modern Ideology* (1998), 3-74, at 18. 121 For a vivid illustration of a desperate landowner hunting harvest labour (in the form of hired slaves) in fourth Cf. Wright, op. cit. (n. 116), 117-19. century Egypt, see P. Lips. m. 122 P. Horden and N. Purcell, The Corrupting Sea: A Study ofMediterranean (2000), 224-30, 390. History 123 On modes of production, see J. Banaji, 'Modes of production in a materialist conception of history', Capital and Class 2 (1977), 1-44. Cf. C. Wickham, 'The other transition: from the ancient world to feudalism', PP 103 (1984), and Late Roman 'Marx, Sherlock Holmes, commerce', JRS 78 (1988), 183-93, especally 187; 3-36; C. Wickham, Wickham, op. cit. (n. 3), 260.

118 C. Whittaker,

THE

GREEK

CENSUS

INSCRIPTIONS

OF

LATE

ANTIQUITY

I05

andmore directly itconfirms theneed to think was not a displacement of theslave system, of labour of farmlabour in terms of a 'logicof deployment',inwhich multiple forms on individualestates.'24 The Late coexisted in the agrarian sector,even side-by-side detail thevitalityand diversity of estate Roman papyri demonstratein extraordinary management inEgypt, where a burgeoning curial class assertedcontroloverproduction, the of staplesforthe market,bydirectly managing amixed labour particularly production The inscriptions was vital in the force.125 offera parallel imageand insistthat slavery Aegean region. matrix of elitecontroloverproductionin the The inscriptions cannot tellus precisely how the labourof slaves and paroikoiwas organizedormanaged. Significantly, however,there are signs of directestate management. The patternsof habitation reflect centralorchestration: theconcentration of labour in individualvillages at Magnesia, Tralles, and Lesbos argues that theworkers were not dispersedon the land as tenants.126 The demographicanalysis controlledcentrally, weremanumittedin their adultyearsat structurally below argues that male slaves signifi cant levels.This pattern impliesthatslaveswere managed fromabove and that manu missionwas a toolof domination;equally, it militatesagainst a readinginwhich slaves were leftto organizeproductionand private lifeon their own.127 Literarysourcesof the were the impresarios Con of slave-basedagriculture.128 period assumed thatlandowners intensive circulation of agri temporary papyrireflect management,and thereinvigorated cultural manuals, in Greek,with an emphasison the production of staplecrops, isanother On theotherhand, theabundance clue thatlandowners directly managed rurallabour.129 would argue thatslaveryin thisregion of femaleslavesand the weight of arable farming mode of plantation labour. But did not catalyse a shift towards a transformative more sensitive to thecomplexity has become of ancient revisionist work onRoman slavery betweenplantationslaveryand hands-off land-usestrategies, while thestrict dichotomy land-leasing has been undermined.'30 an effort If theinscriptions reflect direct by landownersto controlproductionthrough estate management, on multiple scales and in varyingdegrees, it underscores the in the labour sector.13' was an of theverticalrelationships complexity implicated Slavery was onlyone form but it of socially-engineered accessoryto elitecontroloverproduction,

124 Carri?, op. cit. (n.n and n. 106). See Banaji, op. cit. (n. 4), 200-2: 'once one construes patterns of labour use in terms of a logic of deployment it is unnecessary to have to suppose that the agricultural institutions of the empire rested predominately on one type of labour force or category of labour.' 125 and Rural Society in Third-Century A.D. Egypt: The Heroninos Above all, D. Rathbone, Economic Rationalism Archive and the Appianus Estate (1991). P. Lips. 97 (a.D. 338) is a fourth-century example (though in Upper Egypt and no wine production) where slave and free workers were used on the same estate. Cf. CT 9.42.7 (a.d. 369) for a legal example. Banaji, op. cit. (n. 4), on patterns of land ownership. 126 As Jones already noticed; this very fact likely prompted him to employ the loaded word 'gang'. Jones, op. cit. ... slaves were (n. 2), 57: 'these farms were evidently each a centre from which a group of farms was worked sometimes employed in large gangs.' On estate-based residency, see Rathbone, op. cit. (n. 125), 32. 127 This is an important point, since it is sometimes claimed, in the face of considerable evidence for slavery in the Late Empire, that slaves were organized in a tenant mode of production: especially D. Vera, 'Dalla villa perfecta alla villa di Palladio: sulle trasformazioni del sistema agrario in Italia fra principato e dominato', Athenaeum 83 (1995), in Late op. cit. (n. 3), 262, who allows for more direct control over production 189-211, 331-56. Cf. Wickham, ? a crucial in the fifth and sixth centuries that many slaves (especially in the West Antiquity but maintains distinction) were tenants. 128 por example5 Libanius, Or. 14.45 (Ed. R. Foerster, Opera (1903-8), vol. 2, 103), where a master's absence from Greece encouraged his slaves to be lazy and his estate to decline. 129 'The Byzantine agricultural tradition', DOP 25 (1971), 33-59, especially 42. J. Teall, 130 (ed.), Italy', in J. Carlsen Spurr, op. cit. (n. 95); W. Scheidel, 'Grain cultivation in the villa economy of Roman the radical in the Roman Empire Landuse (1994), 159-66; Roth, op. cit. (n. 89). A contrary view emphasizes character of republican villa slavery, e.g., A. Carandini (ed.), Settefinestre: una villa schiavistica nell'Etruria romana (1986). 131 Emphasized romaine', TOO '? propos by J. Andreu and J.Maucourant, 9 (1999), 48-102; Grey, op. cit. (n. 12). de la "rationalit? ?conomique" dans l'antiquit? gr?co

io6

KYLE

HARPER

It is impossibleto say how theparoikoiwere deployed and how much intensification. a fiscal not a type could assertover them. category, of controllandowners They represent The prominence othersestateemployees. agricultural labour.Perhaps somewere tenants, was remunerative and supportsthe of slaveryipso facto impliesthatcontrolover labour suspicion that free workerswere vulnerable to coercion,a situationexacerbatedby the state's new fiscal regime.132 levelsof manu Moreover, if the argumentfor significant mission is sound, some of theparoikoiwere probably freedmen. The idea thatpatron freedmen relationsinfluenced forms of rural dependency has seemedattractive but difficult The inscriptions theconstruction toprove.133 could argue thatelite-driven intensification, of slaves,and new forms of state-sanctioned of estates,a limitedinflux fiscal patronage were symbiotic processes. in of rurallabourpreservedin thecensus inscriptions is significant Finally,thesnapshot itsregional context. The ideaof an Easterncore region capturestheparticular qualitiesof In contrastto the Western core, this was not a district thearea coveredby theinscriptions. In contrastto thesurrounding a rentier-style of ownership.134 Easternprovinces, through was a region to this with a long-established municipal aristocracy situated inproximity The parallels with Egyptare closest, keyexchangenetworks.135 while those with Syriaand where thedegreeof elitecontrolover agriculture isdebated,are seemingly more Palestine, of slaveryin theEastern core,mirrored to a lesserextent in distant.136 The prominence Egypt,shows thatoutcomes in thelabour system were influenced of factors: by a variety tocommodities tenurial structures, proximity markets,access to slave markets,thedensity a new phase in theconstruc of thelocalpopulation,etc.137 The fourth century represented tionof estates in theeastern Ifother forms Mediterranean.138 of ruraldependency gained and sixthcenturies,it appears that slavery over the fifth was structurally vital in this of elitecontrolover agriculture, in theregion seminalphase, along theunstablefrontiers documented by the inscriptions.
V A NEW FRAGMENT FROM THERA: SLAVERY AND DEMOGRAPHY

dominated

by an ancient senatorial class that may have controlled vast quantities of land

from The new inscription Thera is a unique document that invites stillcloser attention.
There

statistical manipulation,and thisinscription should take itsplace as one of the few that providegenuinequantitativeinformation on theslave system.139 The literary evidencefor rural slavery is notoriouslyexiguous, and documentarysources are practicallynon so theinscription from Thera has specialvalue.The inscription existent, offers a profile of
Banaji, op. cit. (n. 104). Kehoe, op. the ability of landowners to use social 133 L. Foxhall, 'The dependent tenant: Shaw, op. cit. (n. 120), 37-8. 134 Vera, op. cit. (n. 35); Vera, op. cit. 135 For a prosopography of the ?lite 132 cit. (n. 106), however, has shown how Roman institutions also constrained coercion, into the fourth century. land, leasing, and labour in Italy and Greece', JRS 80 (1990), 97-114, at 103; (n. 127). of western

are only a handful of data-sets

in the history of Roman

slavery that allow

any

North Syria', in Kingsley and Decker, op. cit. (n. 90), 69-86. 137 For the use of paidaria in Late Roman Egypt, and the convincing arguments that these must be slaves, see economy: new Bagnall, op. cit. (n. 45), 126 and Sarris, op. cit. (n. 4), 86?7. P. Sarris, 'The origins of the manorial 119 (2004), 279-311, at 287. P. Lips. 97 (Hermonthis, a.d. 338) and P. Oxy. insights from Late Antiquity', EHR 58.3960 (a.d. 621) are important examples. 138 Sarris, op. cit. (n. 4). 139 We might mention, on this very short roster, the Egyptian census records, the manumission inscriptions of Delphi, and the senatorial family tomb inscriptions of the Early Empire.

Asia Minor, Halfmann, op. cit. (n. 38), 'Senatoren aus den Provinzen'. See p. 616 for the proprietary networks of this ?lite, extending deep into the Aegean, Kleinasiatischen 'LR2: a container for the military annona on the Danubian including Thera. Trade networks: O. Karagiorgou, border', in Kingsley and Decker, op. cit. (n. 90), 129-66. 136 por t^e debate, seeWickham, 'Food for an empire: wine and oil production in op. cit. (n. 3), 445-6; M. Decker,

THE

GREEK

CENSUS

INSCRIPTIONS

OF

LATE

ANTIQUITY

I07

moment.The owner, in a small space, at a given a slavepopulationbelongingto a single questions,suchas Thera can speak to fundamental and contested demographic data from pat Moreover, thedemographic thepresenceor absenceof femalesin thecountryside.140 forthe natureof estate tern exhibitedin the populationof slaveson Thera has implications management and the dynamic relationshipsbetween slavery and other modes of The demographicinvestigation is thusclosely dependencyin the Late Roman countryside. explored in theprevioussections. relatedto theeconomic themes recordsindetail theslaves for whom one owner,unnamed in the The new inscription covers two stones. On thefirst stone, extantfragment, was fiscally liable.The inscription Farms', followsa listof over I52 names and, beside under theheading 'AndSlaves on the invites This information demographic analysisof the each name, theage of the slave.141 are hazardous.First,there slavepopulationon Thera, althoughtheexercise is inevitably as demographic We are at the data.142 mercy of using census returns theusual difficulties Secondly,thefactthatthisis a of thethoroughness and accuracyof thecensusofficials.143 to see a clearly sword. It isa uniqueopportunity single micro-populationis a double-edged of smallpopulationscan be undulyinfluenced delimited groupof slaves,but thedynamics variations inmortalityor fertility byminor factorsthatare invisibleto us. Short-term mention thatthesamplesizesare too small tooffer statistical coulddistortthe data, not to is itself confidence.144 Moreover, this micro-variability part of the 'kaleidoscopeof local RomanMediterranean.145 the What we have in which characterized demographic regimes' isnot a representative thenew fragment sampleof 'TheDemography'ofRoman slavery, of slaves in thisparticularplace at thisparticular but a snapshotof thedemography moment. A slavepopulation isboth a bio The variableof slavery poses itsown setof problems. of a slavepopulationwas shapedby thenormal The structure logicaland a social entity. - inadditionto a fourth, - fertility, andmigration mortality, parameters ofdemography biological factors such as may have determined critical manumission. Unfree legal status mortalityand fertility.146 etc., inways thatcould affect nutrition, marital opportunity, more profound effect the could havehad an even Slave status upon the population through migration is other twoparameters, migrationandmanumission. In thecase of slavery, - slavescould be imported, or through eitherliterally alternative mechanisms involuntary - sold offor trans Slaves could be exported like theenslavement of exposed children. other thandeath, that may have ferred. Likewise,manumission is a means of attrition, The groupof slaveson Thera was structured the populationas we see it in theinscription. a stablepopulation inequilibrium, a proprietary group,not necessarily freefrominterfer was constructed. Given how this ence.There isnoway toknowwith certainty population we will be forcedto juggle scenarios. all theuncontrollable multiple interpretive variables, thisfindis.The numberof slaves in Still, it isworth emphasizingjusthow important of Egyptian is greater than thenumber recoveredfromthreecenturies this inscription
140 See W. Scheidel, 'Columellas privates ius liberorum: Literatur, Recht, Demographic Einige Probleme', Latomus 53 (1994), 513-27; Scheidel, op. cit. (n. 89); Roth, op. cit. (n. 89). 141 Geroussi-Bendermacher, op. cit. (n. 1), 340: Kai ?ouXou? STti xf|? xcopa?. Paralleled on the estate of Paregorius, no. 142, line 16, which allows us to understand how this list fits into the template used on Thera. 142 See R. Bagnall and B. Frier, The Demography of Roman Egypt (1994), 40-52, on the Egyptian data. 143 The Egyptian data show that the Roman fiscal apparatus was capable of administering a sophisticated and ? land and capita were reasonably accurate census. The extraordinary precision of the Greek census inscriptions ? some of the apparent this bolster counted down to tiny fractional amounts further may impression. Moreover,

in the data are comprehensible, and we may hope to gain some control over possible distortions. To problems anticipate, justwhere we might expect tax evasion and age rounding to cause problems, the data from Thera appear distorted. 144 ^ Scheidel, 'Roman age structure: evidence and models', JRS 91 (2001), 1-26, especially 5. See also R. Sallares, The Ecology of the Ancient Greek World (1991), 120?5. 145 Scheidel, op. cit. (n. 144), 18. 146 inW. Roman in Roman Scheidel See W. Scheidel, (ed.), Debating 'Progress and problems demography', (2001), 1-81, at 24-9; Sallares, op. cit. (n. 144), 118-19. Demography

io8

KYLE

HARPER

of ancient demographyare forced to use census returns (n = I,8).147Practitioners which may have applied to a small amountsof data to extrapolatepatterns extremely millions of people. This inscription population on theorderof several basicallydoubles with itsown quirks.This inscrip theevidenceat our disposal, even ifit is a smalldata-set tion may not unveilfor us thetrue ofRoman slavery, but itrepresents a docu demography of ancientslaverythathas proven inaccessibleto mentarysourcefora social environment At least, it can join thehandfulof stray remarksinColumella, empirical treatment. influenced ofRoman slavery. Appian, et al., that have inordinately thestudy Realistically, it of a ruralslavepopulation might standas ourmost objectivesource forthe demography in the Roman period. The data can be sortedfor information about the sex and age structures of the slave An initiallimitation is thatthetwo stonesare damaged.Of the i5z names, population.148 most straight 87 lines retain legibledata on the slave's age, and the age data are the forward. Only ii9 names can be assigned a sex,most of these with a high degree of There are 76 lines with both sex and age data. The inscription certainty.149 provides invalu able data about thesex ratioof an actual rural slavepopulation. In the ii9 names that were 63 females there and 56males. The sex balance isprima facie provide information, and supportstheview thata predominantly masculine ruralslaveryisa mirage of striking Of course, it ispossible, even likely, that males are under-reported theliterary sources.150 reason:taxevasion.151 But allowingforsomedistortion fora verycomprehensible does not alter theessentialfactthatfemaleslavesexisted inabundanceon Thera.152 The inscription providesdetailed information about the age structure of this slave The number of childrenis immediately striking. Only one slavechildunder population.153 was recorded,thoughages under tenarewell-represented. Infant three mortality is not in the table.154 reflected Before exploring the meaning of thisage structure, we should

147 Bagnall and Frier, op. cit. (n. 142), 342-3. 148 1 for a list of the slaves' names, my reading of their sex and age data, and all the information on See Appendix which these conclusions are based. 149 All names are in the accusative. This makes it impossible to identifywith absolute certainty the gender of names ending with -lov. Most of them will have been masculine names ending in -10c, which became extremely popular in the onomasticon of Late Antiquity (see above, n. 30). But it is not possible to rule out that some were neuter-form aus Paros', in Kleine Schriften 2.3, feminine names, for which, see A. Wilhelm, 'Die sogenannte Het?reninschrift this should not introduce a significant source of error. See the published in SAWW 679 (2000), 441-72. Nevertheless, 1 (the Aegean), LGPN is attested eighty-three times as a masculine name, statistics, where for instance A(|)po?iaio? but never in a neuter-form feminine or even an ambiguous case. 150 Bagnall and Frier, op. cit. (n. 142), 94, found a female-majority slave population. To find a female majority in an estate-based structure on Thera contrasts with the masculine, army setting ismore unexpected. The population barracks model of plantation 'Villa romana e la piantagione slavery in republican Italy: e.g. A. Carandini, in E. Gabba and A. Schiavone schiavistica', (eds), Storia di Roma (1989), vol. 4, 101-200, at 106. See below, n. 152. 151 See below. Scheidel, op. cit. (n. 144), 14; Bagnall and Frier, op. cit. (n. 142), 97-8; R. Bagnall, 'Missing females in Roman Egypt', SCI 16 (1997), 121-38, at 124-5. F?r a modern comparison, M. Karasch, Slave Life in Rio de

1808-1850 (1987), 64. can answer the obstacles suggested by W. Harris, 'Demography, geography, and the sources of Roman ? at least at this later slaves', JRS 89 (1999), 62-75, at ^9>tnat a skewed sex ratio afflicted the rural slave population date. E. Herrmann-Otto, 'Modes d'acquisition des esclaves dans l'Empire romain. Aspects juridiques et socio in 26e colloque du GIREA M. Garrido-Hory (ed.), Routes et march?s d'esclaves, (2002), 113-26, has ?conomiques', also been sceptical of reproductive success. In any case, if the slave population of the fourth century was still extensive, the system was in a largely self-reproducing stage. One could argue, then, either that the Thera data represent only this more mature phase by positing change over time), or that 153 The second digit in the age of one 20-24 and 25-29 bracket. 154 cf. the similarities to the Florentine

Janeiro, 152 This

of the slave system (thus reconcilable with contrasting demographic models, the inscription provides better evidence than the earlier sources. male in his 20s is illegible, so he has been distributed equally between the catasto of 1427-30, where the 'first two or three years of lifewere the most Family, and Ritual in Renaissance Italy (1987), 98. Contrast Bagnall

C. Klapisch-Zuber, underreported': and Frier, op. cit. (n. 142), 44.

Women,

THE

GREEK

CENSUS

INSCRIPTIONS

OF

LATE

ANTIQUITY

1o9

AGE 0to4 5 to 9 10 to 14 15 to 19 20 25 30 to 24 to 29 to 34

MALE 4 5 3 I .5 3.5 I I 3 I 2

FEMALE 4 9 4 2 5 I 5 2 4 I 4

SEX UNKNOWN 0 z 0 I I z 0 0 4 I 0

TOTAL 8 i6 7 4 6.5 6.5 6 3 II 3 6

35 to 39 40 45 50to to 44 to 49 54

55 to 59 6o+ Total

3 29

4 2 47

0 0 II

5 5 87 slaves

consider likely sources of distortion in the recording of the data. Here a comparison with the breakdown of slaves by age cohort in the Egyptian data is helpful (Fig. z).

large number of slave children but a steep drop in the teens. On Thera, the decline after twelve years was calamitous. Bracketing the possibility of anomalous factors like a local 'baby boom' or a boatload of imports, this pattern should be explained by one of two causes. Perhaps the data truly reflect the age structure of the population on Thera, in which case an age-specific mortality factor ravaged adolescent slaves. Such a scenario is entirely possible in particular demographic niches of the ancient world.'55 But it seems 30 25
E Thera

There is one overriding parallel between the twodata-sets.Both populations show a

ID20

8 Egypt

0 E
z 3 10

O to9

tol19

20 to29

30 to39 Age

40 to49

50 to59

60+

FIG. z.

Age cohorts

inThera

and Egypt.

155

Scheidel,

op. cit. (n. 144), 8.

II0

KYLE

HARPER

more likelythattheseverecohortdecline isdue, at least inpart, todistortion motivated is filtered a countconductedforfiscal by tax evasion.All of theinformation through pur poses.Given thesimilarity of the patternin Egypt (where pathogenicand social conditions were different, environment but the fiscal comparable),and the truly catastrophic drop after age twelveon Thera, tax evasionmakes an attractiveexplanation.This would ispresumed when tax liability of ages forslaves in their explain theunder-reporting teens, to have begun.156 in thedata from There are two salientdifferences Egyptand Thera, and these may be The first is theextent relatedto a distortion introduced of age through recording practices. While theEgyptiancensus recordsshow surprisingly rounding. littleage rounding,the from Thera shows a distincttendency for rounding, fromtwenty inscription years and The ratiosof ages divisibleby five up.157 per totalnumberof ages in each bracketare as follows:
AGE 0-9
10I9 20-z9

DIVISIBLE BY 5 3
7 7

TOTAL 24
II 13

30-39 40-49 50-59 6o+

6 II 5 4

9 I4 II 5

tobe rounded. were by fartheleast likely The ages of children under ten This could either a need to track untiltheownerwas liablefortaxeson thechildren or reflect ages carefully simplythegreaterease with which theage of childrencan be guessed at a glance.Age with levelsof illiteracy, so it isnot intrinsically correlated surpris roundingis commonly was prominent ing to findthatage-rounding among ruralslaves.158 Moreover, ifthepat ternisdrivenby an awarenessofwhen tax liability the influence of the began, itconfirms fiscal contexton therecording of this data. to age-rounding The tendency may partiallyaccount for thesecond important diver Thera and Egypt: the largenumberof old slaves on Thera. gencebetween thedata from Partof theanswer lies ina particular with thedata from Thera: sevenslaves anomaly were 'about forty'(spotqji), theonly age recordedas an estimate.Some of theseshould be in theirthirties. contributesto over-estimation Moreover, age-rounding of age, a pheno menon called 'age shoving'.159 in theadultpopulation The high incidence of age-rounding has probablydistortedtheobservedage structure on Thera in favour of old age.Another to overstate ended at some age, providingan incentive possible factoris thattax liability But these distortions theage of a slave.160 and may onlyaccount for partof thedifference, in cities,probably suf it will be necessaryto considerthat Egyptianslaves,concentrated of excessurbanmortality, feredtheadverseeffects whereas thepopulation on Thera was relatively healthy.'16
156 Jones, op. cit. (n. 2), 51; Bagnall and Frier, op. cit. (n. 142), 97. Although in Egypt only males were taxed, in our region females were taxed, but at half the rate of males: CT 13.11.2 (a.d. 386). In the Florentine catasto, young males op. cit. (n. 154), 102. mysteriously disappear at the age of tax liability: Klapisch-Zuber, 157 Bagnall and Frier, op. cit. (n. 142), 44-5. 158 See Duncan-Jones, op. cit. (n. 2), 79?92. 159 Bagnall and Frier, op. cit. (n. 142), 44. 160 Dig. 50.15.3; Bagnall and Frier, op. cit. (n. 142), 107. 161 in cities: Bagnall and Frier, op. cit. (n. 142), 70?1. Concentration

THE

GREEK

CENSUS

INSCRIPTIONS

OF

LATE

ANTIQUITY

III

distortionsintroducedby significant The data fromThera, then,probably reflect young of males, particularly practices.These would includean under-count recording distortions, Despite these theages of adult slaves. toover-estimate males, and a tendency measure of thedemo of theslavepopulationon Thera is an invaluable theage structure Thera from toexploretheage structure worth trying of a groupof ruralslaves. It is graphy plausible that the slave and freepopulations had structurally in detail. It is intuitively marriage environmental stress, hard labour, poor nutrition, profiles: demographic distinct mortalityand may have altered privacy,and exposure to pathogens patterns,restricted to controlforflowsof the Moreover, while it is difficult along linesof status.162 fertility Egyptiancensusdocuments,theslavepopu areas in the free populationbetweendifferent - represents - inand out a factorthat is which 'migration' lation is a social group for Manumission repre to observe in theevidenceas it is preserved.163 difficult particularly of a slave population. variable in the age structure sents a unique and complicating Thera allows us to explore,cautiously,each of these from Nevertheless,the inscription parameters. demographic Roman mortalityin the higher The assumptionthatslavesexperienced First, mortality. Frankly, plausiblebut remainsimpossibleto substantiate'.'64 world 'mayseem intuitively world, were thecommon lot in theancient nutrition poor health,and imperfect poverty, The overall age withoutprejudice. to approach thedata onmortality and it is important and tax of the slave population on Thera- even assumingsome age-shoving structure brutal mortalityregime of an excessively - would argue against thesupposition evasion forruralslaves.The steepdrop in theio-i9 bracketcould be relatedto an age-specific amongadult slaveson but,on theotherhand, theslow rateof attrition pattern, mortality was not cruel, of slavery thattheinstitution This isnot to imply Thera isequally striking. of domina The relationship of ancientslavery. ideasabout thehumanity nor to resurrect was a subtleand complex searchforcontrolthatcould call upon pain or rewardas tion and fiscaldistortioncause the record to overstatethe Even ifage-shoving necessary.165 on slavery Thera would suggestthatagricultural of old slaves, theevidencefrom number New some for known regimes mortality to the devastating thisislanddoes not compare thatloca healthyandmay confirm The population looks relatively World slave systems. of health.166 thanstatusas a determinant was more important tion In thecase of slaves, migra Migration isanotherfactorina population's age structure. of theslavepopulationon Thera could be explainedby The age profile tionis involuntary. thepurchaseof a largenumberof veryyoung slaves.The slavemarket is likelyto have of urban slaves transfer as is thetemporary and such a scenario is imaginable, fluctuated and nine male child children female Therewere thirteen forrearing.167 to thecountryside were and perhaps some of them exposed as infants were disproportionately ren; females - via 'social which should intotheslavepopulation,a scenario migration' transferred Emigrationcan also not be ruledout. If thereproductive seemdistinctly possible here.16' owner's slavesexceededhis need forlabour,hemight selloffthechildren capacityof this - perhaps - theperiod ofmaximum value on the market as theyreachedadolescence

162 age structure as typical. Scheidel, op. cit. (n. 144), 11, warns against taking any singular model of Roman the work of Bagnall and Frier, op. cit. (n. 142), on Egypt provides the basic point of reference. Nevertheless, 163 Scheidel, op. cit. (n. 144), 21-4. 164 Scheidel, op. cit. (n. 146), 29. 165 (1932-1964), vol. 2.1, 163-4); Lactantius, perfectly clear in the ancient sources: Julian, Or. 9.15 (Ed. Rochefort in the Roman Empire: A Inst. div. 5.18 (Ed. S. Brandt, CSEL 19.1 (1890), 461). See K. Bradley, Slaves and Masters as modes of domination, but (1987). Pain and reward incentives are not mutually exclusive Study in Social Control a model', Journal of Economic History 44 see S. Fenoaltea, 'Slavery and supervision in comparative perspective: (1984), 635-68. 166 As argued by Scheidel, op. cit. (n. 146), 15. 167 Such children were supposed to be numbered among the urban slaves: Dig. 50.16.210; Dig. 32.99.3. 168 in the Roman Empire', JRS 84 (1994), 1-22; Bagnall, op. cit. (n. 151). W. Harris, 'Child-exposure

112

KYLE

HARPER

even to otherownerson the island.169 But theabundance of adult femaleslaves and the close associationbetween thechildren on the inscription and the females(demonstrated the impression below) reinforce thatthis was largely a self-reproducing populationwith out dramaticinterference through purchaseor sale. Manumission isanothercrucialdeterminant in theage structure of a slavepopulation. The frequency ofmanumission in the Roman slave systemis a matterof controversy. An older tradition argued that the Romans manumittedslaveswith such regularity that the whole institution of slavery was an improvised means of assimilation.This has been replacedby amore plausibleparadigm in whichmanumission was commonenough to be an effective means of social control, but not thefateofmost slaves.170 The census returns of Egypt provide the most objectivedata on manumission in theRoman Empire; they displayan important pattern bywhich somemales weremanumitted around thirty, while females were kept in slaveryinorder to reproduce new slaves.17' Nevertheless,the inci dence of rural manumission remains almost a complete due to lackof evidence. mystery, The usual assumptionis thatratesofmanumission in thecountryside were negligible.172 The data from Thera may at lastallow us to testfor manumission in an agrarian social The age and sex structure of thepopulation can be searchedforpatternsof attrition. The sample size is small,but thedata from Thera parallel theEgyptianevidence.The imbalancetowards women isdistinctly more pronouncedafterthirty yearsof age (iz:zz = .55), where there are nearlytwiceasmanywomen asmen, thanbeforethirty (I7:25 = .68). Since tax evasion encourages theunder-counting ofmales, thisdistortion usually abates with age;we mightexpect theretobe a relatively greater number of oldmales, rather than fewer.173 Bracketingthepossibility of higher male mortality,thepatternsuggeststhat,as inEgypt, there was a gendered manumission system which freed more men than women, in order to exploit the slavewoman's reproductive potential.174 Justas importantly, it provides some of our best evidence that rural slaveswere manumittedat structurally significant levels. Even thelimited use ofmanumission inagricultural would argue slavery forinternal hierarchies on theestateand slaves inpositionsof responsibility.'75 It may also mean thatsome of theparoikoiwere freedmen.176 But the 'long tail' of theage curveon thisisland,therelatively large of old slaves,reminds proportion us that manumission was
not a sure shot as a slave grew older. Only environment.

of overallattrition, supportsthehypothesis ofmale manumission. is thehardest Fertility demographic parametertogauge fromtheinscription. Fertility is theoutcomeof a variety of factors, including age atmarriage, modes of reproductive con of remarriage, trol,incidence etc.The reproductive successof the Roman slavepopulation

the shift in the sex ratio, not an abnormal

rate

169 See K. Bradley, 'The age at time of sale of female slaves', Arethusa n (1978), 243-52. Cf. Edictum de pretiis rerum venalium 29 (Ed. M. Giacchero (1974), zo8). 170 The high estimates of G. Alf?ldy, 'Die Freilassung von Sklaven und die Struktur der Sklaverei in der r?mischen 2 (1972), 97-129, have been received with Kaiserzeit', Rivista storica dell'antichit? realistic general scepticism. More estimates are available in T. Wiedemann, 'The regularity of manumission at Rome', 35 (1985), 162-75; CQ K. Hopkins, and Slaves (1978), 115-32; Bradley, op. cit. (n. 165), 81-112; K. Conquerors Bradley, Slavery and Society at Rome (1994), 154-65. 171 Bagnall and Frier, op. cit. (n. 142), 71, 156-8. 172 e.g. Bradley, op. cit. (n. 170), 163; Wiedemann, op. cit. (n. 170), 162-3; Duncan-Jones, op. cit. (n. 95), 25. 173 cf. Bagnall and Frier, op. cit. (n. 142), 98-9; Scheidel, op. cit. (n. 144), 14. The most anomalous sex/age grouping on Thera is the near-total absence of males in their late teens and early twenties. 174 Male and female mortality is generally balanced in under-developed populations: Bagnall and Frier, op. cit. (n. 142), 95-9; Scheidel, op. cit. (n. 144), 21. 175 Mediterranean so itwould agriculture required 'care' in addition to 'effort', in the formulation of Fenoaltea, a hallmark of care-intensive slave labour, make sense to find manumission, responsibility, and worker hierarchies: and the slave market in Louisiana, Fenoaltea, op. cit. (n. 165). Cf. S. Cole, 'Capitalism and freedom: manumissions 1725-1820', Journal of Economic History 65 (2005), 1008-27. 176 See above: this would encourage us to believe that manumission was part of the developmental cycle of rural in the fourth and fifth centuries. dependency

THE

GREEK

CENSUS

INSCRIPTIONS

OF

LATE

ANTIQUITY

II3

is a controversial matter,again due to sheerlackof evidence. But Scheidel'sargument that ifthe Roman slavepopulation fromthelate Republic intothe Empirewas on theorderof five million ormore slaves, then natural reproduction had to be thedominantsourceof new slaves, has seemed reallyunassailable, even given that it restsentirely on careful modelling ratherthanpositivedata.177 Despite its limits, the Thera inscription canmake to this an important contribution discussion. Though here the ground becomes less firm,itmay be possible to discern nuclear relationships among the slaves.The slaves could have been recordedin any numberof ways: oldest toyoungest, alphabetical,at random. But, as theeditorhas noticed,they are grouped 'manifestement par families, chaque famille par lemembre le plus commendant The othercensus inscriptions from Thera recordthefreelabourers list by family, age'.'78 ing theoldestmember first and descending by age. In thecase of free workers, thefamily were explicitly was a wife, son,or daughter. stated. The censusnoted ifa person relations For example, 'Paroikoi:Theodore from the location called Politike, thirty years old, twoyearsold, one cow, Zosima, hiswife, twenty yearsold, andTheodora, their daughter, A slavegroupwas listedin this one donkey,five manner: 'Eutychos, sheep'.179 forty years old, Theodoula, twenty-five years old, Lampadion, x years old, Eutychos, fouryears was thesame as in thecase of the old'.180 By all appearances,theorganizational principle free betweenindividual slaves workers,but therelations were notmade explicit. The ordering principle behind thenames is in itself a significant factand suggests that of slaveswas organizedaccordingto smaller,family private lifeamong thiscommunity was entirely under thecontrolof the type groups.The slave'sprivate life master, in legal terms;inmaterial terms,thispowerwas mediated by a numberof complex economic, The dominationof slaves,particularlyin and managerial considerations. disciplinary, isconstrained can nature,and concessionstoprivate life agriculture, by itsends-oriented ofmaintainingeconomiccontrol. be an important component There is seriouscompara is thegreatest tive of reproductive evidencethatfamily determinant success in opportunity ratesseen in a slavepopulation.The reproductive NorthAmerican slavery were thepro duct of a rathersimple recipe.Space, physical independence, privacy,and opportunity, American slave ratherthandeliberatebreedingpolicies,were behind the fertility of the It isof tremendous thattheslaves'namesonThera were organ importance population.181 izedby family-type groups. Nevertheless,given the constraintsthat status imposed on the slave family,it is which thecensus presumptuousto assume thattheslavescarriedon normal familylives We should scrutinize to recognize out of juridical official refused the inscription purism. forquantitative familial women patternsin thepresumptive groupings.It is certainthat inthis and children were prominent population.This makes itintuitively plausible thatthe naturalreproduction of theslavepopulationwas important on thisestate. A distinct pat ternin theregister servesto linkthechildrento theadult females and thus may strengthen of natural reproduction. In twelveinstances where children(< fifteen theinference years) followan adultwhose sex isknown, sevenof theadultswere females.In three cases, the - a adultswere males. In two cases, the sequencewas adultmale, adult female,child plausiblenuclear family.182
177 The lack of evidence is a recognized problem: Scheidel, op. cit. (n. 3): 'unfortunately, our sources do not permit any empirical assessment of this issue.' 178 Geroussi-Bendermacher, op. cit. (n. 1), 345. 179 Kiourtzian, op. cit. (n. 1), no. 142a. 180 Geroussi-Bendermacher, op. cit. (n. 1), 343. 181 R. Fogel, Without Consent or Contract: The Rise and Fall of American Slavery (1989), 114-53. 182 For my judgements on the family groupings, see the Appendix. It is important to note that where legible, the ? children always appear in descending order of age and are always preceded by a plausible adult there are no to assign 'IA,apav and 'E?,7tioa as adult female-child sequences. The it possible stranded children. This makes sequence E?xu^ov Aa^7ta?iov Euxuxov has been grouped as a nuclear family, even though the age 0?O?ou?/r|v is illegible, because the identity of the father-son name would make it likely that this is a father of AauTld?lov mother-son-son group.

114

KYLE

HARPER

FAMILIAL

PATTERNS

IN THE

SLAVE

POPULATION

OF THERA

- Child AdultFemale Male - Child Adult

Adult Male - Adult Female - Child

7 3
z

child'pattern The numbersare verysmall,but the 'adultfemale, appears twiceas oftenas the 'adult outnumber'fathers' This male, child'pattern.'Mothers' plus 'nuclearfamilies'. not necessarily In twoof three caseswhere motherhood.183 suggests practical,if biological, thereis an 'adult male, child' sequence,thechildhas thesamename as theadult, support a book-keeping ingthethesisthatthese are fathers and thatthe whole patternisnot simply must have died young,of course,and it isnot surprising to find accident.184 Many mothers what appear to be father-child patterns.185 and over 'families'is encouragedby the of 'mothers' over 'fathers' The predominance whole population contained a greaternumberof adult femalesthanadult fact that the or rarity of fathers and nuclear familiesultimately depends on males. The frequency were dead,manumitted, partof another 'where'the missingadultmales are.Perhaps they an effort or non-existent.'86 If theordering of names does reflect to count slaves property, ofhusband-wife infamilial sequences. Following groups,it ispossible to testthefrequency as the would appear in the inscription of thefree thepattern marriage population,a slave adultwomen, rangingfrom sequence 'adultmale, adult female'.There are thirty-two woman was 'married' to sixty fifteen at the timeof yearsold, in the list. Ifan individual be preceded in thelistby an adultmale. Of thethirty thecensus,she should,presumably, name or age two adultwomen, tenare in a sequence that is indeterminate: thepreceding adultwomen for whom we have sufficient demographic is lost.That leaves twenty-two data (sex + age) about theprecedingname to know whetheror not itwas a possible
husband. Of those, fourteen do not have a husband listed before them; eight, possibly, do. WOMEN AND MARRIAGE ADULT SLAVE

married Possibly
Apparently not married Uncertain

8
14 IO

males. Presumably, a male who was married at the time of The testcould be repeatedfor Of thesixteen thecensus adultmen, ranging from would be followedby an adult female. Of the eighteento sixty-six years old, threeare in a sequence that is indeterminate. of them remaining thirteen, eightare followedby an adult female;five are not (twobeing our 'widowers'- thatis, followedby a child slavewith his name).
MEN AND MARRIAGE ADULT SLAVE Possibly married Apparently not married 8 5

Uncertain

183 some number of We should recall that urban slaves could be sent to the countryside for rearing. Likewise, exposed infants may have been imported into the population. 184 For Egyptian data suggesting slave children named after their Aiov?aiov-Aiov?aiov. Euxuxiav?v-Euxuxiav, father, see Bagnall and Frier, op. cit. (n. 142), 159. 185 Mortality frequently produced single-parent families, Bagnall and Frier, op. cit. (n. 142), 123. 186 cf. a slave population with relatively high fertility in Brazil: A. Metcalf, 'Searching for the slave family in colonial Brazil: a reconstruction from S?o Paulo', Journal of Family History 16 (1991), 283-97, at 290: 'Fathers of their children remained nebulous figures on the fringes of the family.'

THE

GREEK

CENSUS

INSCRIPTIONS

OF

LATE

ANTIQUITY

II5

- adultmale followedby The inscription thusincludes eight possible conjugal sequences adult female.It is important tonote that even these arenot certaintobemarriages. Of the six pairs, twoof themincludecoupleswho have a relatedname (Zosimos and Zosima; Ammianos andAmmias). These could be brother-sister pairsormarriages (or evenboth), and it is impossibletoknow forsure. The conjugal scenariosdisproportionately include theoldest slaves.The ages of the men are sixty-six, c. forty, sixty-five, sixty, forty, twenty-eight, and twenty 'married' fifty, most likely yearsold. The two nucleargroupsare thefamily ofEutychos,forty something yearsold, hiswifeTheodoula, twenty-five, Lampadion, x yearsold, and Eutychos,four, along with the familyof Ammianos, around forty, Ammias, twenty, and the baby is Eutychos.The numberof nuclearunits which can be identified with relative certainty are otherprobable cases.187 thusstrikingly small,althoughthere The most glaring pattern is thatallmales fifty or older are apparently married.This is a striking anomaly.It ispos was a rewardforsurvival or long-term sible that marriage service, butperhaps thesimplest we are able to seemarriagesprecisely men have not died or explanation is that when the been freed. The marriage rateamongmen ishigherthanamongwomen, especiallyifthe of thetenremaining three unmarried'fathers' are excludedaswidowers (eight adultmen women could be explainedeither weremarried),so thatthelarge number of unmarried by a real shortage of availablemen or by thefactthatadultmen have beenmanumitted.188 is sound, thereare two patterns that seem to conflict: If this interpretation many on a few marriedwomen. Reproductivesuccess is usuallycontingent children, relatively certain levelof stability and privacy in the slave's personal life. Though ultimatelyit remainsnecessary to entertaina varietyof interpretive solutions to the age structure on Thera (e.g. breeding,importation, thereare con exhibited pathogens,randomness), vergingreasons to accept a dual resolutioninwhich familylifeand male manumission were both important on theisland.The hypothesis of family lifeis supported prima facie by the sheernumberofwomen and childrenin thepopulation, a fact which had been predicted but neverempirically verified. The ordering principlebywhich theengravers clue in favour recordedtheslaves is in itself anothersignificant of thisview.The factthat as is thesmallbut with thechildrenis further adultwomen are associated corroboration, wheremen precede children,theyshare the elegant fact that in two of threesequences same name.The high of marriage ratesamong availablemen also signal the importance success is not incom within thispopulation.The hypothesis of reproductive familylife that sees child exposure as an additional source of new patiblewith an interpretation slaves. The principalsupport male manumissionare also compelling. The reasonstobelieve in When we mightexpect thefiscal distor is theshiftin thesex ratioaftertheage of thirty. tion in favourof femalesto abate, theproportionof women in thepopulation grows. sex specific, Mortality regimesin under-developed populations are usuallynot strongly Most we must admit thatdifferential though mortalityby genderremainsa possibility. a persuasiveargu successon this estate is itself theevidenceforreproductive importantly, as it impressionistically ment formale manumission: if thispopulationwas as fertile are needed,andmanumission would be an effective appears, thenfathers explanationfor are.A final is that husbandsand fathers where the 'missing' supportforthisreconstruction
For example, followed by three children including age unknown', 40', followed by 'Moax?* 'Euae?f]V, 12'. 'Euas?f]v, 188 It is also possible that the slave 'mothers' were 'married' to men who belonged to another estate, though there is evidence that masters tried to keep their slaves' relationships in-house. M. Flory, 'Family in familia: kinship and 3 (1978), 78-95, especially 78-82; Cato, De agr. 143 (152); Varro, De re rustica 1.17.5, community in slavery', A]AH SC 273 (1980), 145). Also, CT 2.10.6; Columella, De re rustica 1.8.5; Tertullian, Ad uxorem 2.8.1 (Ed. C. Munier, 187

2.25.1 (a.D. 325) assumes that if slave families were separated between owners, the family was broken apart. in Economic History 7 (1982), 239-86, at 270-1, shows that R. Steckel, 'The fertility of American slaves', Research in America was variable, and as the size of properties increased, the the incidence of cross-plantation marriage incidence of cross-plantation marriages declined. Itwas rare in Brazil: Metcalf, op. cit. (n. 186), 286.

iii6

KYLE

HARPER

manumission.190

theallowance of slave familiesisconsonant with thepracticeofmanumission,sinceboth The point to a styleof dominationmore subtle than unremitting pain incentives.189 ofLate Roman estate implications forthestudy managementaremanifold,since itshould have to account for the mobilization of femaleand child labour,theuse of familialrela on theestate through of domination,and thecreationof hierarchies tionsas a strategy

The argument for natural reproduction within the Roman slavepopulationhas always on logicalplausibility thanevidence. relied rather The inscription from Thera providesus at last with concrete data in supportof thehypothesis. This isnot to claim thattheslave fortheslave supplyclearlydrew on alternative populationwas entirely self-reproducing, estate was of theoveralldistribution this sources,andwe do notknow how representative - and Butwhatever thelimits of thisinformation and socially.19' of slaves,geographically are it is source a thattheonlydocumentary for large,rural they realenough important which has been deduced as neces slave estateexhibitspreciselythedemographic profile of time. saryto sustaina largeslavepopulationover a longstretch
VI CONCLUSIONS

are a unique but challengingset of documents thatyield The Greek census inscriptions These inscriptions precious insightsinto theLate Roman countryside. were probably middle of thefourth and they of theaggres engravedafterthe century, may be artefacts ofEgyptprovide thebasic com fiscal sively centralizing policyofValens. The documents but thecensus inscriptions allow us toexploreclassicquestionsabout parativeframework, wealth and labour in a regionof theEmpirewith an oldermunicipal eliteand ingreater proximity to centraltrade networks. Significantly, theinscriptions confirm theimageof a where theurban aristocracy rooted invillage structures, could influence but countryside in the inscriptions Most of thepropertiesreflected not utterly dominate the ruralsector. are curial-scale on theorderof several from modest to lavish. Properties holdings,ranging with otherprovincial, curialestates. The inscriptions hundrediugera compare suggest that but they also insistthata broad class of urbanowners landed wealthwas highly stratified, of agricultural resources. most striking participatedin theexploitation Perhaps the datum, essentially unremarked,is theextentof land held by the senatorialorder. If the later reflect the transformation of theEastern fourth-century dating is right,the inscriptions momentum. of particular eliteat an instant As documentary of labourusage in theLate Empire, the inscrip evidencefor patterns valuable.The inscriptions furnish tionsare especially data about thedensity of registered workers.The documentsgive the labour and theoverall deployment of slaves and free of a healthy,intensely-exploited TheMagnesia inscription impression countryside. reveals on larger was disproportionately This patternshouldnot thatlabour properties. registered even as contractland-leasing largerlandholdings, between smallerurban landlordsand on coloni.The ultimatestate independent villagerssurvivedtheriseof fiscallegislation is thatslave labour was unexpectedly in mentmade by thecensus inscriptions important thisperiod and region. Although slavery appears to have been a minoritycomponentof
189 cf. U. Roth, 'To have and to be: food, status, and the peculium of agricultural slaves', JRA 18 (2005), 278-92, for the earlier period. 190 Concern for female labour, see above, n. 117. Interest in child labour is visible in contemporary agricultural 2.2 (Ed. H. Beckh (1895), 34). Jerome was certainly aware that familial relations were an manuals: Geoponica 6 (Ed. C. C. Mierow in Classical instrument of domination: Jerome, Vita Malchi Essays Presented to J. A. Kleist in the estate accounts from Hermonthis in P. Lips. (1946), 33-60, at 44). Internal hierarchies: see the apxiy??)pyoi 97 at 8.23 and 14.27. 191 For dissimilar demographic dynamics within subregions of Brazil, Metcalf, op. cit. (n. 186), 284.

be read as a sign of crisis, but rather as a clue that fiscal dependency was correlated with

THE

GREEK

CENSUS

INSCRIPTIONS

OF

LATE

ANTIQUITY

II7

was nevertheless the labour force on elite-ownedland, it structurally vitalwithin a com plexmixed-labour system. The inscriptions implythatelitesexercisedcontrolover rural productionand thatslavery was instrumental in theorganization of estates. The new fragment from Thera permitsan unprecedentedlook at thedemographic of a slave-basedestate.The endurance profile of the Roman slave system is itself amatter is an important of controversy, and theinscription piece of evidencein favour of survival. If theslave system continuedintothefourth century, thedominanceof naturalreproduc tionin theslave supplyispractically certain, but this does not diminishthesignificance of Thera. This documentprovidesour first theevidence from documentary supportfora rural slave population with large numbers of women and children.Perhaps more males weremanumittedin their unexpectedis theconclusionthatsome adultyears,since toempiricalstudy. rural manumissionhabitshave neverbeen susceptible The supposition ofmale manumission and servilefamiliesrequiresa model of estatemanagement that mode of slave emphasizescentralcontrolwithout positing a radicallytransformative based production. The census inscriptions bringto life a broad landedaristocracy whose internal composi tion was influx. thedense and varied labourforce They reflect deployedtowork theland. This sampleof theagrarianeconomy demonstrates thatinpatterns ofwealth and labour, that social change inLate Antiquityisnot reducibletonarratives posit a linearaccumula tionof property, a dominantform of dependency, or a smoothtransition betweenlabour roleas an energetic, systems. elitestrove Slavery played an under-appreciated enterprising to controlproductionin a vibrantcountryside. The eastern Mediterranean in thefourth century was home to a dynamicand complex society.Itwas thesortof society where a merchantcould rise to become an imperialfunctionary and be rewarded with land and are thedebris of a society The census inscriptions slaves acrossGreece.192 which could prompta contemporary priest to complain thathis flock would pay no mind to serious how tobuy land,and how to buy slaves, matters,because 'we spendevery day scheming, - we are insatiable. '.193 and how to increase our property University ofOklahoma kyleharper~ou.edu

a man known to Libanius, Or. 62.46-8 e.g. Heliodorus, (Ed. R. Foerster, Opera (1903-8), vol. 4, 370). In acta apostolorum 8? jrn?? ovap G?Xsiv x? xoia?xa 32.2 (PG 60, col. 237): i^?? John Chrysostom, Kai otico? kX&?g) xf|v o?aiav ?kXi ottco? ji?v ?ypov cbvnacojxe?a, Kai ottco? ?v?paTto?a, 7toif|aco|nev, ?ia^?yeaGai, o? taxu?avoiuev Ka0' 8K?axr|v fjuipav ?ia?-eyonevoi K?pov. 193

192

ii8 I: THE

KYLE

HARPER OF THERA

APPENDIX

SLAVES

IN THE CENSUS

INSCRIPTION

NAME
STONE A 'Y'yciav Ktrquilptov lagpa3tiav
Zx6)(Tptv

NAME PARENT AGE SEX MARRIAGE

50 z8 20 F M F
M

PARENT AGE SEX MARRIAGE

Unmarried Married Married

'Ouy6Xikov K-riqcyiplov A~pobi6kov Zwouipiov FkaiiKlv 066oukov Xct6vqv 'Enup66trcov

E. .. va 35

M M 25 33 M M F M F M
M Unmarried Father

.0.o. C.40 Ic?4avov 'Enacy&aO1vC.40 ApaKovti6av 'EnUy60iv 24 Tcipjv Oco66Tflv Ei5-rv~iav ATc??X&v
... (Tqv .G1061V At6?copov 7

M F M F F F F M
F M M

Unknown Unknown Unmarried Unmarried Unmarried

Unmarried Unknown

56 36 46

'Icakiav Kicwva FUaiKilV Z6Gl6tgov

... 1Tov

8 5 30 IO 8 35
I5

F M F M
M

A~p66ettgov Fka6tKflv ... (5 lines)

. v

M F

Unmarried Mother

'EituKcTrav

56 z'Ov)X6kIov 'E7EKWT1TaV I8

F M F F M F F
M

Unknown Married Married Unknown Married Married

.rTov Xcolrtpav
Eu5 ?1vov

M F
M

Unmarried Mother

AkXcavbpov Eutuuxtav6v Ei)yevtov Xcet)cpav EXaivnqv EHtuXiav

(Dtkoug?vqv

6 3

M M M F F F
F

EUt'UXiav E6)yvIov Mou0oy7vtav E6TuXiav

co@p6vIov

55 50 40

Apocnviv EO3TFipUV EHy7vtov ZAoain1tov Married Married Unmarried Mother Zwaigtrv

E)... v ... ..XtoV

I5 7 4 65 6o
I8 I3

16trctpav ZMu6tov Eu',cuXiav Fkau6Kv KakTuIpi9v

... av ... v ... ov Apocyiviv ... ... ?Cgav

F 66 6o 40 4 M F F F
F(?) 20 3 25 9 F M

F F M M F
M

Unknown

Married Married

46

STONE B raFULKcv Zwir6pav 16)pova Hap6dktov 'EXku6a

Unmarried

34 Io 8

EiruXiav
... vav . ..ov

6
30

F
F

5 c. 40 -Z FlavvUxLov EHcvafOiav 50 Io Zcocipnv

..Kp..lt. 8 ...6qv

F F M M F M F F
F

Unknown

Mother

Unmarried Unknown Mother

. v

THE NAME AGE

GREEK SEX

CENSUS MARRIAGE

INSCRIPTIONS PARENT NAME

OF

LATE AGE

ANTIQUITY SEX MARRIAGE

119 PARENT

...v ... E...

C. 40 40

EIruiav OE660ukov 'Enarya6O~v 'Dtk6O&vov Kkavuluviv 'EiLiKrCr1V AptIuvOV Agptdlaba

Eb'tuxov

4 56

F M F M F F M F
M

MoGY6) EHcp3iv Xaphv 'Yyciav Unknown

Iz IO 7

zi c. 40 zo
I

Unmarried Married Father Married Mother Unmarried Mother Married Married Father Mother

Z6cntgov EilruXtav6v C.40 9 Ei6-cuXiav D... .vrv 51 'EXnt6tav6v 48 zz E.. i6a

.... Uv ... ...vov C.-40

F M F F M M F F M F
F(?)

Unmarried Unmarried Unknown Unmarried

Father

lIkdpav ZOxuigiv Ei)t)uOv EOco6o)XTlv Aapu66tov ED5-ruov AyOcova Tt5rllv ZM6aiov 'EniKt1rcnv E65y&'Viov
E fpa ... ... ... v lov

-4 9 40 25 4 6o 5z

F F M F M M M F M F M

Apocnv6v Eu'TuvXiav
... kuuiicpov

30 Z5 5
IO

M F
M

Unknown Unmarried Father

Married Married

Atov6ctov AIov6ucov 'Ekniria

M06av

M M F
F

Unmarried Mother

Iov 'O4uRXO 16,16?1pav X6retppav KX

Unknown

M F F

56 27 8

A~po6icaiv 'OuD6X6oV Ei~a?cpihv

4 30 50

F M M

Unmarried Unknown

N? B