The Making of a Roman Imperial Estate: Archaeology in the Vicus at Vagnari, Puglia 9781803272054, 9781803272061, 1803272058

The Making of a Roman Imperial Estate presents excavations and analysis of material remains at Vagnari, in southeast Ita

181 124 49MB

English Pages 392 Year 2022

Report DMCA / Copyright

DOWNLOAD FILE

Polecaj historie

Ronan McMahon's Big Book of Profitable Real Estate Investing: The Simple Secrets of Making a Fortune in Real Estate 9781911260

277 64 2MB Read more

The Archaeology of Peasantry in Roman Spain 9783110757415, 9783110757200

This volume aims to present an updated portrait of the Roman countryside in Roman Spain by the comparison of different t

184 49 16MB Read more

A Fortress in Brooklyn: Race, Real Estate, and the Making of Hasidic Williamsburg 9780300258370

The epic story of Hasidic Williamsburg, from the decline of New York to the gentrification of Brooklyn Hasidic William

599 44 4MB Read more

The Positioning of the Roman Imperial Legions 9781841718965, 9781407329055

The positioning of the legions of the Imperial Roman army provides a window into both the thinking and the course of eve

264 112 8MB Read more

A Fortress in Brooklyn: Race, Real Estate, and the Making of Hasidic Williamsburg 0300231091, 9780300231090

The epic story of Hasidic Williamsburg, from the decline of New York to the gentrification of Brooklyn "A rich chr

399 60 10MB Read more

The Social Dynamics of Roman Imperial Imagery 1108835120, 9781108835121

Images relating to imperial power were produced all over the Roman Empire at every social level, and even images created

952 162 11MB Read more

Imperial Ideals in the Roman West 9781107005082, 2011008366

144 81 120MB Read more

The Archaeology of Greek and Roman Slavery 9781472540928, 9780715631959

Slavery is a word heavy with emotional and political overtones – to be owned by another person and treated as a commodit

356 35 30MB Read more

Imperial Ideology and Provincial Loyalty in the Roman Empire 9780520923720

The Roman empire remains unique. Although Rome claimed to rule the world, it did not. Rather, its uniqueness stems from

295 33 2MB Read more

The Roman Imperial Mausoleum in Late Antiquity 9781107644410, 9780521513715, 2009005514

This book is the first comprehensive study of the mausolea of the later Roman emperors. Constructed between c.AD 244 and

682 247 487MB Read more

The Making of a Roman Imperial Estate: Archaeology in the Vicus at Vagnari, Puglia
9781803272054, 9781803272061, 1803272058

Author / Uploaded
Maureen Carroll (editor)

Table of contents :
Cover
Title Page
Copyright page
Contents
List of Figures
CHAPTER 1
Figure 1.1 Map of South Italy showing the location of Vagnari and other relevant settlements.
Figure 1.2 View from San Felice over the Vagnari plateau (yellowish-green vegetation, photo middle), the slope of Monte Marano (right), and the water of the Lago di Serra del Corvo/Diga del Basentello in the distance.
Figure 1.3 Drone photo of the Vagnari plateau. The vicus lies to the left of the ravine (in the centre of the photo), the cemetery to the right. In the background right is the hill of San Felice.
Figure 1.4 Original geophysics plot-out by John Hunt with excavated remains superimposed on it.
Figure 1.5 The photo (2012) shows the stone-built drains under the walls and floors of the portico building that showed up on the geophysics as lines.
Figure 1.6 Overall multi-phase plan of excavated remains 2012–2018.
Figure 1.7 The Iron Age settlement of Silvium (Botromagno), seen from Gravina in Puglia.
CHAPTER 2
Figure 2.1 Map showing settlements of the Mid-Imperial period and topographical features mentioned in the text.
CHAPTER 3
Figure 3.1 View of the upper Basentello River valley from Irsina.
Figure 3.2 Mesozoic limestone on the west side of the ravine at Gravina in Puglia.
Figure 3.3 Sauropod foot print in the Gravina Mesozoic limestone, which has been used by Conti et al. 2005 to determine the age and geographic setting of the limestone deposits.
Figure 3.4 Geological overlay of the deep and surface structure of southern Italy (after Sauer et al. 2010: fig. 1).
Figure 3.5 General stratigraphic column of the Apulian Platform unit succession and their ages (after Petrullo et al. 2017: fig. 2).
Figure 3.6a–b Marine marl exposures in the Bradano River valley west-southwest of Irsina.
Figure 3.7 The Apennine Blue Clay also outcrops in the erosional channels of streams around Vagnari, as it does here just north of the archaeological site (Google Earth Image).
Figure 3.8 Middle to late Pleistocene coastal terraces mapped and described along the Gulf of Taranto near Metaponto (after Sauer et al. 2010: fig. 2).
Figure 3.9 Comparison of coastal terrace ages in the Metaponto area (red dots in top diagramme) with the Marine O18 Isotope Stages or the record of O18 content of ocean sediment cores in the lower diagram.
Figure 3.10 Along a stream cut just north of the Vagnari site, the grey-coloured upper Pliocene-lower Pleistocene marine marl, which forms the bedrock of the area, is exposed.
Figure 3.11 Cross-bedded conglomerate formations such as these along the road leading south out of Irsina seem to be part of an upper Pliocene to lower Pleistocene sequence of fan-delta-front to shallow marine conglomerates and sandstones.
Figure 3.12a–d Beach deposits exposed at the Bosco locality south of Gravina in Puglia (a-b).
Figure 3.13 Fluvial sand deposits at the top of the slope above the San Felice villa site.
Figure 3.14a–b Fluvial sands on the slope above the San Felice villa site.
Figure 3.15a–b Stream and spring deposits northeast of Castello di Monteserico.
Figure 3.16 Here the fluvial unit consists of a chaotic slurry of sands, silts, and gravels.
Figure 3.18 The deposits overlying the fluvial deposits are comprised of sequences of thinly laminated clayey silts with accumulations of indurated calcium carbonate between the laminated units.
Figure 3.19 Detail of the deposits overlying the fluvial deposits comprised of sequences of thinly laminated clayey silts with accumulations of indurated calcium carbonate between the laminated units.
Figure 3.20 Approximate elevation of the Upper Basentello river near Vagnari above mean sea level during the last 800,000 years.
Figure 3.21a–b Just north of Vagnari vicus, recent erosion along the stream incision has exposed the cross-sections of two streams of different ages and with different fill.
Figure 3.22 The lower channel with oxidized Pliocene fill from the top of the plateau.
Figure 3.23 The contact between the two channel fills of the 130,000-year-old channel.
Figure 3.24 Younger stream cut-and-fill in the stream channel just north of the Vagnari vicus site.
Figure 3.25 The Irsina exposure east of the town (on the crest of the hill in the distance).
Figure 3.26 Stratigraphy of the Arroyo Italiano exposure southwest of the Vagnari vicus site.
Figure 3.27a–b Holocene record of fluvial deposits with superimposed soil horizons nested within the stream channel north of the Vagnari vicus site, with (a) a close up of the deposits, and (b) the well-developed angular-blocky structure of the soils.
Figure 3.28a–f Flood control canals constructed during the 1950s to move rainwater quickly out of fields on the valley floor accelerated stream velocities to the point that they began head-ward erosion in streams above the upper ends of the canals (a-b).
Figure 3.29 Comparison of the sum of spring and summer rainfall with the sum of the radiocarbon dates on erosion in southern Italy (red line).
Figure 3.30a–b Modern summer erosion currently typical of the tributaries of the upper Basentello River.
Figure 3.31 Principal component analysis of pollen from Lago Grande Monticchio core 90D from the crater of Monte Vulture.
Figure 3.32a–c Comparison of the pollen rations of the three dominant periods from Lago Grande di Monticchio.
Figure 3.33 Possible spring locality south of the Vagnari cemetery site.
CHAPTER 4
Figure 4.1 Simplified plan of excavated remains in the north-west part of the vicus, with attested and estimated room arrangements. Plan J. Moulton.
Figure 4.2 Levelled natural chalk and clay deposit with visible tool marks from Phase 1 (foreground). Photo M. Carroll.
Figure 4.3 Cylindrical pit 4021. Photo M. Carroll.
Figure 4.4 Grey-gloss oil lamp from the lowest deposit (4039) in cylindrical pit 4021. Photo M. Carroll.
Figure 4.5 Excavated remains belonging to Phase 2. Plan J. Moulton.
Figure 4.6 Robber trench (4057/6050) of the Phase 1 wall 4017, view from the north. Photo M. Carroll.
Figure 4.7 Wall 4006, faced on the north inner side (top) with a vertical tile facing. Photo M. Carroll.
Figure 4.8 Lower right: wall 4006; background: wall 4059; lower right, remnant of wall 4062. View from the south. Photo M. Carroll.
Figure 4.9 Drone photograph of trenches in 2018. Lower right: robber trenches of Phase 2 buildings; upper trench: wall 6022 (Phase 5). Photo V. Ferrari and G. Ceraudo.
Figure 4.10 Cobblestone floor (4015) of Room L, bounded by wall 4006. View from the west. Photo M. Carroll.
Figure 4.11 Wall 5003, view from the south. Photo M. Carroll.
Figure 4.12 View, from the east, of the stone drain 5045/4054, with parts of the side walls of the drain removed to show the cobblestone foundations. Photo M. Carroll.
Figure 4.13 View, from the west, of the stone-built drain 5013/2024. Photo M. Carroll.
Figure 4.14 View, from the west, of wall 5003 in which a gap for drain 5045 has been blocked up with masonry and mortar (middle of photo) when the drain was given up.
Figure 4.15 Grain storage pit 5008 with remains of its backfill containing pottery, loom weights, animal bone, and metal. Photo M. Carroll.
Figure 4.16 Grain storage pit 6061 with the remnants of the ceramic dolium still in situ. Photo M. Carroll.
Figure 4.17 Mortar-lined grain storage pit 4020 with the impression of a flat-bottomed dolium preserved at the bottom. Photo M. Carroll.
Figure 4.19 Excavated remains belonging to Phase 3. Plan J. Moulton.
Figure 4.20 The cella vinaria at Vagnari, showing intact or partially intact dolia in mortar basins (6, 7, 9), robbed-out dolium impressions (2, 5, 8, 10), and circular darker traces of dolium holes not excavated (1, 3, 4). Plan J. Moulton.
Figure 4.21 Mortar-lined dolium basins, with a large fragment of a dolium preserved in basin 3020 (left) and just the basin for 3012 (right). View from the north.
Figure 4.22 Mortar-lined basin 3012 for a dolium, view from the north. Photo M. Carroll.
Figure 4.23 Channel 3025 cut through the mortar floor of the cella vinaria, view from the east.
Figure 4.24 The western part of drain 4054/5013 cut by the placement of dolium 3020 in Room A. View from the east.
Figure 4.25 View from the southeast of stone-built drain 5045 (left), blocked up so the water and waste could be diverted into stone drain 5051 (right). Photo M. Carroll.
Figure 4.26 Excavated remains belonging to Phase 4. Plan J. Moulton.
Figure 4.27 View from the south at east-west wall 3008 (top) abutting onto the older north-south running wall 3010 of Phase 2 (right). Photo M. Carroll.
Figure 4.28 View from the east of wall 2018, the western wall of the cella vinaria in Phase 2 (right), cut by wall 2013, the western wall of the portico building, in Phase 4 (left). Photo M. Carroll.
Figure 4.29 View from the southeast of the northern wall 1005 of the portico building (Room D) with a doorway and threshold. Photo M. Carroll.
Figure 4.30 Wall 3008 of Room D with its internal plaster skim in situ. Photo M. Carroll.
Figure 4.31 Burnt daub 2010/3011 lying on the beaten earth floor in Room D in the portico building. Photo M. Carroll.
Figure 4.32 Chunks of burnt daub 2010/3011 with black patches and burned out organic material. Photo M. Carroll.
Figure 4.33 The various courses of wall 5028. Photo M. Carroll.
Figure 4.34 Sondage with walls 4012 and 4018. Photo M. Carroll.
Figure 4.35 Stone-built drain 3035, the northern end of drain 5045 in Phase 3. Photo M. Carroll.
Figure 4.36 View from the north of stone-built drain 1009, the Phase 4 of the Phase 3 drain 3035/5051 (left).
Figure 4.37 Northern Phase 5 extension of stone-built drain 1009 where it pierces and flows through the northern outside wall 1008 of the portico porch. View from the south. Photo M. Carroll.
Figure 4.38 Excavated remains belonging to Phase 5. Plan J. Moulton.
Figure 4.39 Excavating the porch (Room E) of the portico building, with the external wall 2005 on the right. View from the east.
Figure 4.40 Detail of wall 6047. Photo M. Carroll.
Figure 4.41 Tile hearth 5070. Photo M. Carroll.
Figure 4.42 Tile hearth 5044. Photo M. Carroll.
Figure 4.43 Lower third of the dolium 3038, with a student as scale. Photo M. Carroll.
Figure 4.44 Human skeletons in the dolium 3038. Photo M. Carroll.
Figure 4.45 Excavated remains belonging to Phase 6. Plan J. Moulton.
Figure 4.46 Large fragments of a glass windowpane in situ. Photo M. Carroll.
Figure 4.47 Deposits 4052 and 5014 in Room C, with traces of burning. View from the north. Photo M. Carroll.
Figure 4.48 Collapsed roof tiles 1004, 3027, 4007 in Rooms B, C, and L. Running through the deposits is wall 4006. View from the east. Photo M. Carroll.
Figure 4.49 Collapsed roof, with tiles on either side of wall 4013 which separated Room B from Room C, later robbed out leaving this void. View from the west. Photo M. Carroll.
Figure 4.50 Wall 6022 foundations and post holes 6031, 6042 and 6043 from scaffolding used to dismantle the wall. View from the north. Photo M. Carroll.
Figure 4.51 Post hole 6031 from scaffolding in Phase 6. Photo M. Carroll.
CHAPTER 5
Figure 5.1 Regional Red Slip plate from backfilled grain storage pit 5037, Phase 2. Photo M. Carroll.
Figure 5.2 Eastern Sigillata red slip bowl FV300. Photo M. Carroll.
Figure 5.3 Fragment of a large, collared jar FV1043 from Phase 2. Photo M. Carroll.
Figure 5.4 Plain ware bowl FV1151. Photo M. Carroll.
Figure 5.5 Maker’s stamp on the mortarium FV1157/1161. Photo M. Carroll.
Figure 5.6 Lamp fragments from Phases 2, 5, 6, and 7. Photo D.R. Griffiths and H. Russ.
Figure 5.7 Lamp fragments from Phase O. Photo D.R. Griffiths and H. Russ.
Figure 5.8 Misfired plain ware cup FV605. Photo M. Carroll.
CHAPTER 6
Figure 6.1 Roman funerary relief (from Rome?) depicting a man and woman (left) on the premises of their vineyard and winery full of dolia.
Figure 6.2 Detail of the relief in Figure 6.1, showing lidded dolia in the winery. Photo M. Carroll.
Figure 6.3 Remains of a tall, cylindrical dolium (Cat. 6.3). Photo M. Carroll.
Figure 6.4 Fragment of a ceramic lid from a dolium (Cat. 6.6). Photo M. Carroll.
Figure 6.5 Lower third of dolium 3038 (Cat. 6.1) in situ with numerous cracks.
Figure 6.6 Finger indentations on the top of the vertical wall of the dolium (Cat. 6.1), an indication of having been coil built. Photo M. Carroll.
Figure 6.7 Fragment of a dolium with black volcanic temper and reddish-brown grit. Photo M. Carroll.
Figure 6.8 Map of central and southern Italy showing the Roman Magmatic Province and the Ernici-Roccamonfina Magmatic Province, as well as various relevant sites. Volcanoes are marked as red triangles.
CHAPTER 7
Figure 7.1 Vessel glass fragments.
Figure 7.2 Caroline Jackson examining all the window glass fragments from context 3040 (Cat. 7.WF01), with students Kayla Thiessen and Thomas Watson. Photo M. Carroll.
Figure 7.3 Window glass fragments from various contexts.
CHAPTER 8
Figure 8.1 Silver victoriatus, Cat. 8.2, obverse. Photo M. Stirn.
Figure 8.2 Silver victoriatus, Cat. 8.2, reverse. Photo M. Stirn
CHAPTER 9
Figure 9.1 Bronze finger ring, Cat. 9.1. Photo M. Carroll.
Figure 9.2 Bronze finger ring, Cat. 9.2. Photo M. Carroll.
Figure 9.3 Bronze arm ring, Cat. 9.3. Photo M. Carroll.
Figure 9.4 Bronze chain with threaded links, Cat. 9.4. Photo M. Carroll.
Figure 9.5 Bronze Mid-La Tène fibula, Cat. 9.5. Photo M. Carroll.
Figure 9.6 Bronze Aucissa fibula, Cat. 9.6. Photo M. Carroll.
Figure 9.7 Fragmentary bronze belt buckle, Cat. 9.7. Photo M. Carroll.
Figure 9.8 Bronze belt buckle, Cat. 9.8. Photo M. Carroll.
Figure 9.9 Bronze belt(?) stud, Cat. 9.9. Photo M. Carroll.
Figure 9.10 Bronze belt buckle tongue, Cat. 9.10. Photo M. Carroll.
Figure 9.11 Bronze cosmetic or medical instrument shaft, Cat. 9.11. Photo M. Carroll.
Figure 9.12 Bronze mirror with silver wash, Cat. 9.12, cut to be recycled. Photo M. Carroll.
Figure 9.13 Bronze solid-cast boar’s head attachment, Cat. 9.13. Photo M. Carroll.
Figure 9.15 Bronze nail, Cat. 9.14. Photo M. Carroll.
Figure 9.16 Detail of the incised head of the bronze nail Cat. 9.14. Photo M. Carroll.
Figure 9.17 Bronze strip, possibly a mount for the lock of a wooden box, Cat. 9.15. Photo M. Carroll.
Figure 9.18 Bronze mount for a wooden box, Cat. 9.16. Photo M. Carroll.
Figure 9.19 Bronze net repair needle, Cat. 9.17. Photo M. Carroll.
Figure 9.20 Bronze casting mould, Cat. 9.18. Photo M. Carroll.
Figure 9.21 Bronze sheet with nail holes and iron corrosion ring, Cat. 9.19. Photo M. Carroll.
Figure 9.22 Fragment of a bronze vessel rim(?), Cat. 9.19. Photo M. Carroll.
Figure 9.23 Sheet metal strip of bronze, folded, Cat. 9.21. Photo M. Carroll.
Figure 9.24 Bronze sheet metal scrap, Cat. 9.22. Photo M. Carroll.
CHAPTER 10
Figure 10.1 Iron bit head, Cat. 10.1. Photo M. Carroll.
Figure 10.2 Iron bit head, Cat. 10.2. Photo M. Carroll.
Figure 10.3 Prong of an iron hoe, Cat. 10.3. Photo M. Carroll.
Figure 10.4 Fragmentary iron figure-of-eight links chain, Cat. 10.4. Photo M. Carroll.
Figure 10.5 Fragmentary large iron chain, Cat. 10.5. Photo M. Carroll.
Figure 10.6 Fragmentary large iron chain, Cat. 10.6. Photo M. Carroll.
Figure 10.7 Iron ring, Cat. 10.7. Photo M. Carroll.
Figure 10.8 Iron buckle, Cat. 10.8. Photo M. Carroll.
Figure 10.9 Iron steelyard arm, Cat. 10.9. Photo M. Carroll.
Figure 10.10 Various iron nails. Photo M. Carroll.
Figure 10.11 Fragmentary iron nail with discoidal head, Cat. 10.10. Photo M. Carroll.
Figure 10.12 Fragmentary iron nail with slightly domed head, Cat. 10.11. Photo M. Carroll.
Figure 10.13 Fragmentary iron nail with conical/pyramidal head, Cat. 10.12. Photo M. Carroll.
Figure 10.14 Fragmentary iron nail with circular head, Cat. 10.13. Photo M. Carroll.
Figure 10.15 Iron wall hook, Cat. 10.14. Photo M. Carroll.
Figure 10.16 Iron knife blade, Cat. 10.15. Photo M. Carroll.
Figure 10.17 Detail of metal ring on the tang of the iron knife blade, Cat. 10.15. Photo M. Carroll.
Figure 10.18 Iron knife blade, Cat. 10.16. Photo M. Carroll.
Figure 10.19 Iron knife blade, Cat. 10.17. Photo M. Carroll.
Figure 10.20 Fragmentary iron knife blade, Cat. 10.18. Photo M. Carroll.
Figure 10.21 Selection of iron hob nails. Photo M. Carroll.
CHAPTER 11
Figure 11.1 Iron smithing heart bottom. Photo M. Carroll.
Figure 11.3 Iron tap slag. Photo M. Carroll.
Figure 11.5 Piece of unworked iron, perhaps a partly worked bloom or a piece of iron bar to be reshaped.
Figure 11.7 Pie chart showing metal traces in a crucible from context 4027.
Figure 11.8 Pie chart showing metal traces from context 4023 on cobblestone floor 4015.
CHAPTER 12
Figure 12.1 Clipped piece of lead sheet, Cat. 12.19. Photo M. Carroll.
Figure 12.2 Clipped piece of lead sheet with striations from rolling, Cat. 12.13. Photo M. Carroll.
Figure 12.3 Selection of clipped lead sheet pieces with various striations. Drawing I. De Luis.
Figure 12.4 Round lead ingot, Cat. 12.40. Photo M. Carroll.
Figure 12.5 Solidified lead droplet. Photo M. Carroll.
Figure 12.6 Various pieces of lead scrap for recycling. Photo M. Carroll.
Figure 12.7 Lead sheet scrap folded in itself, ready for recycling. Photo M. Carroll.
CHAPTER 13
Figure 13.1 Misfired roof tile with vitrified surface, Cat. 13.15. Photo M. Carroll.
Figure 13.2 Misfired roof tile in Figure 13.2, bubbly fabric on the reverse side. Photo M. Carroll.
Figure 13.3 Excavating the collapsed roof over Rooms B and C. View from the south. Photo M. Carroll.
Figure 13.4 Tegula from the collapsed roof in Phase 6, Cat. 13.1. Photo M. Carroll.
Figure 13.5 Imbrex from the collapsed roof in Phase 6, Cat. 13.4. Photo M. Carroll.
Figure 13.6 Fragmentary tegula, Cat. 13.2. Photo M. Carroll.
Figure 13.7 Fragmentary tegula, Cat. 13.2. Photo M. Carroll.
Figure 13.8 Fragmentary imbrex with a wavy line, Cat. 13.5. Photo M. Carroll.
Figure 13.9 Fragmentary imbrex with wavy lines, Cat. 13.6. Photo M. Carroll.
Figure 13.10 Tegula fragment with incised anchor or arrow, Cat. 13.7. Photo M. Carroll.
Figure 13.11 Brick fragment with cat paw prints, Cat. 13.9. Photo M. Carroll.
Figure 13.12 Tegula fragment with impression of a hob-nailed shoe, Cat. 13.12. Photo M. Carroll.
Figure 13.13 Brick fragment with rows of incised dots, Cat. 13.8. Photo M. Carroll.
Figure 13.14 Round tile fragment from a hypocaust, Cat. 13.13. Photo M. Carroll.
Figure 13.15 Segmental tile, Cat. 13.10. Photo M. Carroll.
Figure 13.16 Detail of the impression of textile on the back of segmental tile, Cat. 13.10. Photo M. Carroll.
Figure 13.17 Segmental tile fragment, Cat. 13.11. Photo M. Carroll.
Figure 13.18 Figure 13.18 Column made of segmental tiles at Minturnae in southern Latium. Photo M. Carroll.
CHAPTER 14
Figure 14.1 Pieces of marble revetment and mosaic.
Figure 14.2 White marble revetment with greenish-grey veins, Cat. 14.4, reverse side showing reworking and chisel marks. Photo M. Carroll.
Figure 14.3 Detail of a piece of grey-veined marble revetment showing an incised line, Cat. 14.3. Photo M. Carroll.
Figure 14.4 (A) thin-section microphotograph of the white limestone composing the mosaic piece, Cat. 14.6, SF7-3-2001-13/8 (scale bar=0.5 mm, crossed polars); (B) thin section microphotograph of the bedding mortar surrounding the mosaic fragment (scale ba
Figure 14.5 XRPD patterns of the investigated white marble inlays: (A) 3040; (B) SF91-4-4030-16-4/1; (C) SF91-4-4030-16-5/2; (D) SF24-4-3006-15; (E) SF25-4-3001-15; (F) SF49-4-5058-17.
Figure 14.6 Thin-section microphotographs of the investigated white marble inlays (scale bar = 0.5mm, crossed polars): (A) 3040; (B) SF91-4-4030-16-4/1; (C) SF91-4-4030-16-5/2; (D) SF49-4-5058-17; (E) SF24-4-3006-15; (F) SF25-4-3001-15.
Figure 14.7 MGS (mm) variation ranges and median values of the most important Mediterranean white marbles used in antiquity (data after Gorgoni et al. 2002) and values measured in the samples presented in this work.
Figure 14.8 (A) thin-section microphotograph of the impure marble SF51-4-4030-17, Cat.14.4, (scale bar=0.5mm, crossed polars); (B) XRPD pattern of the same sample.
Figure 14.9 Mn content of the white marbles analyzed in this study compared with the variation ranges and median values of the most important Mediterranean white marbles used in antiquity (data after Moens et al. 1988).
Figure 14.10 Scatterplot of the carbon and oxygen isotope compositions of the white marbles considered in this study and the confidence ellipses (probability distribution 99%) corresponding to some of the most important white marbles quarries exploited in
CHAPTER 15
Figure 15.1 Wall plaster fragment with dark greenish curvilinear motif on a greyish-white background, Cat. 15.1. Photo M. Carroll.
Figure 15.2 Wall plaster fragments with red paint, Cat. 15.2. Photo M. Carroll.
Figure 15.3 Wall plaster fragments with red and yellow paint, Cat. 15.3-4. Photo M. Carroll.
CHAPTER 16
Figure 16.1 Basalt hand-mill of the biconcave type, Cat. 16.1. Photo M. Carroll.
Figure 16.2 Detail of basalt hand-mill, Cat. 16.1, showing the horizontal handle socket. Photo M. Carroll.
Figure 16.3 Basalt hand-mill of the top concave type, Cat. 16.2. Photo M. Carroll.
Figure 16.4 Tall stone mortar of fine-grained limestone, Cat. 16.4: a. rim fragment; b. rim fragment. Photos M. Carroll.
Figure 16.5 Inside polished and worn surface of the stone mortar, Cat. 16.4: a. inside rim and body fragment; b. inside body fragment. Photo M. Carroll.
Figure 16.6 Tenon hole in the body of the stone mortar, Cat. 16.4. Photo M. Carroll.
Figure 16.7 Detail of the tenon hole in the stone mortar, Cat. 16.4. Photo M. Carroll.
Figure 16.8 Ovoid cut stone weight, Cat. 16.5. Photo M. Carroll.
Figure 16.9 Ovoid cut stone weight, Cat. 16.6. Photo M. Carroll.
CHAPTER 17
Figure 17.1 Loom weight, Cat. 17.3, with markings on the front. Photo M. Carroll.
Figure 17.2 Detail of loom weight Cat. 17.3. Photo M. Carroll.
Figure 17.3 Loom weight, Cat. 17.4. Photo M. Stirn.
Figure 17.4 Possible coin impression of a chimaera on the top of loom weight Cat. 17.7. Photo M. Carroll.
CHAPTER 18
Figure 18.1 One side of a worked bone sheet, cut or sawn along the two long edges, Cat. 19.1. Photo M. Carroll.
Figure 18.2 Other side of the worked bone sheet Cat. 18.1. Photo M. Carroll.
Figure 18.3 Fragmentary shaft of a bone pin or awl, Cat. 18.7. Photo M. Carroll.
Figure 18.4 Fragmentary bone spatula head needle, Cat. 18.6. Photo M. Carroll.
Figure 18.5 Bone needle with a broken tip, Cat. 18.6. Photo M. Carroll.
Figure 18.6 Bone hair pin with a spherical head, Cat. 18.4.
Figure 18.7 Bone spoon with a round bowl, Cat. 18.2.
Figure 18.8 Small bone stopper from a bottle, Cat. 18.3.
CHAPTER 19
Figure 19. 1 Relative proportions of main domesticates for Phases 2–6. NISP = 295. Zones = 250.
Figure 19.2 Cattle astragalus measurements from Vagnari and comparative southern Italian sites.
Figure 19.3 Sheep/goat astragalus distal breadth (Bd) vs greatest lateral length (GLl).
Figure 19.4a–b Fossil shells from the local geology at Vagnari. Photo A. Trentacoste.
Figure 19.5 Map showing the location of Vagnari and comparative Roman sites in southern Italy referred to in this chapter.
Figure 19.6 Relative proportion of the main domesticates from Vagnari and comparative sites.
CHAPTER 20
Figure 20.1 Discriminant analysis results showing Vagnari samples compared with Amorgos crop processing stages (Jones 1987).
Figure 20.2 Vagnari cereal proportions by phase.
Figure 20.3 Correspondence analysis of Vagnari samples, categorised by phase. Text labels were left visible in this figure to highlight plant species that influenced sample clusters.
Figure 20.4 Weed seed flowering seasonality proportions from Vagnari samples.
CHAPTER 21
Figure 21.1 The dolium defossum with the top layer of the vicus skeletal material exposed.
Figure 21.2 Cranial elements associated with Individual A. Photo T.L. Prowse.
Figure 21.3 Lower limb bones associated with Individual A. Photo T.L. Prowse.
Figure 21.4 Lower limb bones associated with Individual B. Photo T.L. Prowse.
CHAPTER 22
Figure 22.1 Drain 5045 (E-W) and drain 5051 (N-S) with soil removed. Photo M. Carroll.
Figure 22.2 Roundworm (Ascaris sp.) eggs from drain 5045 at Vagnari vicus. Scale bars are 20μm.
CHAPTER 23
Figure 23.1 3-D reconstruction of the northwest sector of the vicus, view from the west. In the foreground is the cella vinaria. Drawing I. De Luis.
Figure 23.2 Map of the outline of the vicus, as investigated through geophysics and excavation, with the plotted scatter of roof tile in red.
Figure 23.3 Satellite image of the Vagnari plateau with the roof tile scatter plotted by C. Small and the vicus buildings superimposed. The sharp edges of the scatter indicate the limits of the survey zone. Image I. De Luis.
Figure 23.4 3-D reconstruction of the northwest sector of the vicus, view from the south. On the upper right is a courtyard building possibly serving as a market or a mansio. Drawing I. De Luis.
Figure 23.5 View of the Vagnari plateau, with the Casa Cantoniera along the road in the middle of the photo and the masseria Vagnari to the left. Photo M. Carroll.
Figure 23.6 The Fontana della Bonifica near the Casa Cantoniera on the road to Vagnari. Photo M. Carroll.
Figure 23.7 Excavated remains superimposed on the geophysics plot. The reservoir is the large rectangular feature at the bottom. Plan I. De Luis.
List of Tables
Table 3.1 Middle to Late Quaternary Marine Terrace ages before the present of coastal terraces in the Metaponto area and their Marine Isotope Stage (MIS) assignment (after Sauer et al. 2010; Brückner 1980).
Table 3.2 The Blytt-Sernander periods.
Table 5.1 Bulk data by phase by count/weight (grams).
Table 5.2 Relative proportions by ware class (by weight) and phase.
Table 5.3 Functional analysis by category and form using relative proportions (based on estimate vessel equivalents -EVEs- of featured vessels).
Table 5.4 Functional analysis using relative proportions (based on estimate vessel equivalents -EVEs- of featured vessels).
Table 6.1 Summary of analytical results, with lipids and selected sources identified.
Table 7.1 Frequency table for vessel glass fragments by phase and context.
Table 7.2 Frequency table for window glass fragments by phase and context.
Table 9.1 Number of finds per function group.
Table 9.2 Measurements of finger rings.
Table 14.1 Summary of principal mineralogical and textural characteristics of studied marbles derived by X-ray diffraction analysis (XRD) and observation of thin sections by the polarizing microscope (OM).
Table 14.2 Major (%wt) and trace elements (ppm) of the studied samples.
Table 14.3 C and O stable isotopic composition of the analysed samples.
Table 17.1 Phase of occupation and correspondent number of loom weights found in it.
Table 17.2 Calculation of the number of warp threads per loom weigh, per two loom weights, and of the warp threads per centimetres using a loom weight from context 4039.
Table 17.3 Calculation of the number of warp threads per loom weigh, per two loom weights, and of the warp threads per centimetres using a loom weight from context 5008.
Table 17.4 Calculation of the number of warp threads per loom weigh, per two loom weights, and of the warp threads per centimetres using a loom weight from context 4002.
Table 19.1 Bone preservation by phase. Only includes specimens with zones.
Table 19.2 Number of identified specimens (NISP) and zone counts by phase. Remains from flotation heavy fractions and hand collected specimens are grouped together.
Table 19.3 Body part distribution for Phases 2–6: minimum animal units (MAU) for each element.
Table 19.4 Bone fusion of the main domestic taxa, Phases 2–6.
Table 19.5 Mandible wear stages for the main domestic taxa, Phases 2–6. Wear stages follow Payne (1973) and O'Connor (1988).
Table 19.6 Land snails, zone counts for Roman phases.
Table 19.7 Livestock representation on comparative sites in southern Italy.
Table 20.1 Cultivated and wild taxa from Vagnari, Phase 1 samples.
Table 20.2 Cultivated and wild taxa from Vagnari, Phase 2 samples.
Table 20.3 Cultivated and wild taxa from Vagnari, Phase 3 samples.
Table 20.4 Cultivated and wild taxa from Vagnari, Phase 4 samples.
Table 20.5 Cultivated and wild taxa from Vagnari, Phase 5 samples.
Table 20.6 Cultivated and wild taxa from Vagnari, Phase 6 samples.
Table 22.1 Transmission routes and symptoms of common intestinal parasites found in the Roman Empire.
Table 22.2 Drain and context numbers of soil samples collected from Vagnari.
List of Plates
Plate 1 Colour-coded plan of the excavated remains in the northwest sector of the vicus, without the tile collapse. Plan J. Moulton.
Plate 2 Colour-coded plan of the excavated remains in the northwest sector of the vicus, with the tile collapse. Plan J. Moulton.
Plate 3 3-D reconstruction of the northwest sector of the vicus. Reconstruction I. De Luis.
Plate 4 Pottery Phase 1, scale 1:3. Drawing D.R. Griffiths, photo M. Carroll and H. Russ.
Plate 5 Pottery Phase 2, scale 1:3. Contexts 4060, 5042, 5046, 5050, 5062. Drawing D.R. Griffiths.
Plate 6 Pottery Phases 2 and 3, scale 1:3. 1–5 Phase 2, contexts 6046, 6053, 6062; 6 and 8 Phase 3, contexts 2012 and 3025. Drawing D.R. Griffiths.
Plate 7 Pottery Phases 4 and 5, scale 1:3. 1-3 Phases 4 and 4/5, contexts 1020 and 5008; 4-10 Phase 5, contexts 3013, 4023, 5047, 6022. Drawing D.R. Griffiths.
Plate 8 Pottery Phase 6, context 4030 below tile collapse, not including lids, scale 1:3. Drawing D.R. Griffiths.
Plate 9 Pottery Phase 6, context 4030 below tile collapse, lids only, scale 1:3. Drawing D.R. Griffiths.
Plate 10 Pottery Phase 6, below tile collapse, scale 1:3. Drawing D.R. Griffiths.
Plate 11 Pottery Phase 6, tile collapse contexts 1004 and 3024, scale 1:3. Drawing D.R. Griffiths, photo M. Carroll and H. Russ.
Plate 12 Pottery Phase 6, deposits 3005, 3006 and 3022 above tile collapse, scale 1:3. Drawing D.R. Griffiths.
Plate 13 Pottery Phase 6, contexts 2010, 4029 and 4052, scale 1:3. Drawing D.R. Griffiths.
Plate 14 Pottery Phase 6, context 5014, scale 1:3. Drawing D.R. Griffiths.
Plate 15 Pottery Phase 7, topsoil and plough soil deposits, scale 1:3. Drawing D.R. Griffiths.
Plate 16 Pottery Phase 7, topsoil and plough soil deposits, scale 1:3. Drawing D.R. Griffiths.
Plate 17 Pottery Phase O, contexts 1003, 3015, 1016, 2003 and 2006, scale 1:3. Drawing D.R. Griffiths.
Plate 18 Pottery Phase O, contexts 2015, 2016, 2019, 3019 and 4010, scale 1:3. Drawing D.R. Griffiths.
Plate 19 Pottery Phase O, contexts 5015, 6029, 6054 and 6059, scale 1:3. Drawing D.R. Griffiths.
Plate 20 Fragments of dolia and dolium lids, contexts 2019 and 2022 (Phase O), 4006/4009 (Cat. 6.6), 4028 (Cat. 6.4) and 4029 (Cat. 6.5), probably Phase 3, scale 1:5. Drawing D.R. Griffiths.
Plate 21 Vessel glass. 1. Cat. 7.VF11 - base and body of a patella cup; 2. Cat. 7.VF14 - base and body of a patella cup; 3. Cat. 7.VF12 - base of an unidentified vessel, most likely a cup; 4. Cat. 7.VF10 - base of an unidentified vessel; 5. Cat. 7.VF15 –
Plate 22 Vessel glass. 1. Cat. 7.VF16 and Cat. 7.VF17 - fragment of a bottle handle; 2. Cat. 7.VF21 -fragment of the body of a vessel glass decorated with vertical ribs; 3. Cat. 7.VF22 - fragment of the body of a vessel glass decorated with vertical ribs;
Plate 23 Iron, scale 1:2. 1. Cat. 10.16, knife blade; 2. Cat. 10.15, knife blade: 3. Cat. 10.5, large chain; 4. Cat. 10.4, small chain; 5. Cat. 10.8, buckle. Bronze, scale 1:1. 6. Cat. 9.13, boar head attachment; 7. Cat. 9.5, fibula; 8. Cat. 9.6, Aucissa
Plate 24 Bronze, scale 1:1. 1. Cat. 9.17, net repair needle; 2. Cat. 9.16, furniture mount; 3. Cat. 9.14, decorated nail; 4. Cat. 9.1, finger ring; 5. Cat. 9.8, buckle. Lead, scale 1:1. 6. Cat. 12.33, shell weight; 7. Cat. 12.32, weight with a stone; 8. C
Plate 25 Lead, scale 1:2. 1. Cat. 12.54, round base with bronze inlay; 2. Cat. 12.48, reinforcing strip with iron nails; 3. Cat. 12.52, pipe fragment; 4. Cat. 12.51, ceramic repair tenon. Marble, scale 1:3. 5. Cat. 14.2, fragment of Dokimeion marble; 6. C
Plate 26 Worked stone, scale 1:3. 1. Cat. 16.4, rim fragment of a limestone mortar; 2. Cat. 16.4, section through the body of the limestone mortar; 3. Cat. 16.4, detail of the limestone mortar with a tenon hole. Worked bone, scale 1:2. 4. Cat. 18.3, bottl
Plate 27 Ceramic loom weights, scale 1:2. 1. Cat. 17.2; 2. Cat. 17.3; 3. Cat. 17.4; 4. Cat. 17.5; 5. Cat. 17.6; 6. Cat. 17.7; 7. Cat. 17.19; 8. Cat. 17.11. Drawings S. Cann.
Plate 28 Ceramic loom weights, scale 1:2. 1. Cat. 17.12; 2. Cat. 17.13; 3. Cat. 17.4; 4. Cat. 17.15; 5. Cat. 17.17. Drawings S. Cann.
Preface and Acknowledgements
Maureen Carroll
1. Introduction. Exploring the Imperial Estate at Vagnari, 2012–2019
Maureen Carroll
2. The Topographical Context of Vagnari in the Roman Period: A Brief History of the Study Area
Alastair Small
3. The Landscape Context of Vagnari, Past and Present
Peter Wigand
4. Vagnari Vicus: Buildings and Chronology
Maureen Carroll, Kelsey Madden, and Jonathan Moulton
5. The Vessel Pottery and Lamps
David R. Griffiths
6. The Dolia Defossa and Viticulture at Vagnari
Maureen Carroll, Giuseppe Montana, Luciana Randazzo, Donatella Barca, and Benjamin Stern
7. Vessel and Window Glass
Camilla Bertini and Victoria Lucas
8. Roman Republican and Imperial Coins
David Wigg-Wolf
9. Copper-Alloy Artefacts
Stefanie Hoss
10. Iron Artefacts
Louis-Olivier Lortie, Maureen Carroll, and Stefanie Hoss
11. Ferrous Metallurgical Debris and its Chemical Analysis
Louis-Olivier Lortie
12. Lead, Lead-Working Debris, and Lead Sourcing
Maureen Carroll, Jane Evans, Vanessa Pashley, and Louis-Olivier Lortie
13. Building Ceramics: Brick, Tile, and Clay
Maureen Carroll, Giuseppe Montana, Luciana Randazzo, and Donatella Barca
14. Marble and Stone Revetment and Pavements: Context and Provenance
Maureen Carroll, Renato Giarrusso, Giuseppe Montana, Luciana Randazzo, and Giovanna Scopelliti
15. Painted Wall Plaster
Maureen Carroll
16. Worked Stone Utensils
Maureen Carroll
17. Loom Weights and Textile Weaving at Vagnari
Beatrice Triozzi
18. Worked Bone
Maureen Carroll
19. Animal Remains from Vagnari: Bones and Shells
Angela Trentacoste
20. The Botanical Remains
Matthew Stirn and Rebecca Sgouros
21. Human Remains in the Vicus
Tracy L. Prowse
22. Parasites and Human Health at Vagnari
Marissa L. Ledger and Piers D. Mitchell
23. The Making of an Imperial Estate
Maureen Carroll
Sintesi in Lingua Italiana della Ricerca presentata nel Volume
Beatrice Triozzi
APPENDIX 1. Catalogue of Excavated Features and DepositsAccording to Phases
APPENDIX 2. List of Catalogued Small Finds
APPENDIX 3. Pottery Fabric Descriptions
APPENDIX 4. Catalogue of Featured Ceramic Vessels
APPENDIX 5. Catalogue of Oil Lamps
APPENDIX 6. Inventory Recording Form for Human Skeletons
Bibliography
PLATES

Citation preview

The Making of aEstate: The Making of a Roman Imperial Archaeology in the Vicus at Vagnari, Puglia Roman Imperial Estate: Archaeology in the Vicus at Vagnari, Puglia

Maureen Carroll Archaeopress Roman Archaeology 88

The Making of a Roman Imperial Estate Archaeology in the Vicus at Vagnari, Puglia

Maureen Carroll (ed.)

Archaeopress Archaeology

Archaeopress Publishing Ltd Summertown Pavilion 18-24 Middle Way Summertown Oxford OX2 7LG www.archaeopress.com

ISBN 978-1-80327-205-4 ISBN 978-1-80327-206-1 (e-Pdf) © Archaeopress and the individual authors 2022

Cover: Views of excavations at Vagnari Vicus

All rights reserved. No part of this book may be reproduced, or transmitted, in any form or by any means, electronic, mechanical, photocopying or otherwise, without the prior written permission of the copyright owners. This book is available direct from Archaeopress or from our website www.archaeopress.com

Contents

List of Figures�� iv List of Tables��x List of Plates�� xi Preface and Acknowledgements�� xiii Maureen Carroll CHAPTER 1 Introduction. Exploring the Imperial Estate at Vagnari, 2012–2019��1 Maureen Carroll CHAPTER 2 The Topographical Context of Vagnari in the Roman Period: A Brief History of the Study Area��9 Alastair Small CHAPTER 3 The Landscape Context of Vagnari, Past and Present��16 Peter Wigand CHAPTER 4 Vagnari Vicus: Buildings and Chronology��39 Maureen Carroll, Kelsey Madden, and Jonathan Moulton CHAPTER 5 The Vessel Pottery and Lamps��65 David R. Griffiths CHAPTER 6 The Dolia Defossa and Viticulture at Vagnari��82 Maureen Carroll, Giuseppe Montana, Luciana Randazzo, Donatella Barca, and Benjamin Stern CHAPTER 7 Vessel and Window Glass��92 Camilla Bertini and Victoria Lucas CHAPTER 8 Roman Republican and Imperial Coins�� 103 David Wigg-Wolf CHAPTER 9 Copper-Alloy Artefacts�� 106 Stefanie Hoss

i

CHAPTER 10 Iron Artefacts�� 117 Louis-Olivier Lortie, Maureen Carroll, and Stefanie Hoss CHAPTER 11 Ferrous Metallurgical Debris and its Chemical Analysis��125 Louis-Olivier Lortie CHAPTER 12 Lead, Lead-Working Debris, and Lead Sourcing��129 Maureen Carroll, Jane Evans, Vanessa Pashley, and Louis-Olivier Lortie CHAPTER 13 Building Ceramics: Brick, Tile, and Clay�� 136 Maureen Carroll, Giuseppe Montana, Luciana Randazzo, and Donatella Barca CHAPTER 14 Marble and Stone Revetment and Pavements: Context and Provenance��143 Maureen Carroll, Renato Giarrusso, Giuseppe Montana, Luciana Randazzo, and Giovanna Scopelliti CHAPTER 15 Painted Wall Plaster�� 156 Maureen Carroll CHAPTER 16 Worked Stone Utensils�� 158 Maureen Carroll CHAPTER 17 Loom Weights and Textile Weaving at Vagnari��162 Beatrice Triozzi CHAPTER 18 Worked Bone�� 171 Maureen Carroll CHAPTER 19 Animal Remains from Vagnari: Bones and Shells��174 Angela Trentacoste CHAPTER 20 The Botanical Remains�� 188 Matthew Stirn and Rebecca Sgouros CHAPTER 21 Human Remains in the Vicus�� 206 Tracy L. Prowse

ii

CHAPTER 22 Parasites and Human Health at Vagnari�� 209 Marissa L. Ledger and Piers D. Mitchell CHAPTER 23 The Making of an Imperial Estate�� 215 Maureen Carroll Sintesi in Lingua Italiana della Ricerca presentata nel Volume��232 Beatrice Triozzi APPENDIX 1 Catalogue of Excavated Features and DepositsAccording to Phases��238 APPENDIX 2 List of Catalogued Small Finds�� 253 APPENDIX 3 Pottery Fabric Descriptions�� 257 APPENDIX 4 Catalogue of Featured Ceramic Vessels �� 265 APPENDIX 5 Catalogue of Oil Lamps�� 308 APPENDIX 6 Inventory Recording Form for Human Skeletons��315 Bibliography�� 322 PLATES�� 345

iii

List of Figures

CHAPTER 1

Figure 1.1 Map of South Italy showing the location of Vagnari and other relevant settlements.�� 1 Figure 1.2 View from San Felice over the Vagnari plateau (yellowish-green vegetation, photo middle), the slope of Monte Marano (right), and the water of the Lago di Serra del Corvo/Diga del Basentello in the distance.�� 2 Figure 1.3 Drone photo of the Vagnari plateau. The vicus lies to the left of the ravine (in the centre of the photo), the cemetery to the right. In the background right is the hill of San Felice.�� 2 Figure 1.4 Original geophysics plot-out by John Hunt with excavated remains superimposed on it. �� 3 Figure 1.5 The photo (2012) shows the stone-built drains under the walls and floors of the portico building that showed up on the geophysics as lines.�� 4 Figure 1.6 Overall multi-phase plan of excavated remains 2012–2018.�� 5 Figure 1.7 The Iron Age settlement of Silvium (Botromagno), seen from Gravina in Puglia.�� 6

CHAPTER 2

Figure 2.1 Map showing settlements of the Mid-Imperial period and topographical features mentioned in the text.�� 10

CHAPTER 3

Figure 3.1 View of the upper Basentello River valley from Irsina.�� 16 Figure 3.2 Mesozoic limestone on the west side of the ravine at Gravina in Puglia.�� 17 Figure 3.3 Sauropod foot print in the Gravina Mesozoic limestone, which has been used by Conti et al. 2005 to determine the age and geographic setting of the limestone deposits.�� 17 Figure 3.4 Geological overlay of the deep and surface structure of southern Italy (after Sauer et al. 2010: fig. 1).�� 18 Figure 3.5 General stratigraphic column of the Apulian Platform unit succession and their ages (after Petrullo et al. 2017: fig. 2).�� 18 Figure 3.6a–b Marine marl exposures in the Bradano River valley west-southwest of Irsina. �� 19 Figure 3.7 The Apennine Blue Clay also outcrops in the erosional channels of streams around Vagnari, as it does here just north of the archaeological site (Google Earth Image).�� 20 Figure 3.8 Middle to late Pleistocene coastal terraces mapped and described along the Gulf of Taranto near Metaponto (after Sauer et al. 2010: fig. 2).�� 21 Figure 3.9 Comparison of coastal terrace ages in the Metaponto area (red dots in top diagramme) with the Marine O18 Isotope Stages or the record of O18 content of ocean sediment cores in the lower diagram. �� 21 Figure 3.10 Along a stream cut just north of the Vagnari site, the grey-coloured upper Pliocene-lower Pleistocene marine marl, which forms the bedrock of the area, is exposed. �� 22 Figure 3.11 Cross-bedded conglomerate formations such as these along the road leading south out of Irsina seem to be part of an upper Pliocene to lower Pleistocene sequence of fan-delta-front to shallow marine conglomerates and sandstones. �� 22 Figure 3.12a–d Beach deposits exposed at the Bosco locality south of Gravina in Puglia (a–b). Shells (c-d) were recovered from these deposits as evidence of their coastal origin. Photos P. Wigand.�� 23 Figure 3.13 Fluvial sand deposits at the top of the slope above the San Felice villa site. �� 24 Figure 3.14a–b Fluvial sands on the slope above the San Felice villa site. �� 25 Figure 3.15a–b Stream and spring deposits northeast of Castello di Monteserico. �� 25 Figure 3.16 The fluvial unit consists of a chaotic slurry of sands, silts, and gravels. �� 26 Figure 3.17 The deposits overlying the fluvial deposits are comprised of sequences of thinly laminated clayey silts with accumulations of indurated calcium carbonate between the laminated units. �� 26 Figure 3.18 The deposits overlying the fluvial deposits are comprised of sequences of thinly laminated clayey silts with accumulations of indurated calcium carbonate between the laminated units. �� 26 Figure 3.19 Detail of the deposits overlying the fluvial deposits comprised of sequences of thinly laminated clayey silts with accumulations of indurated calcium carbonate between the laminated units. �� 26 Figure 3.20 Approximate elevation of the Upper Basentello river near Vagnari above mean sea level during the last 800,000 years. �� 27 Figure 3.21a–b Just north of Vagnari vicus, recent erosion along the stream incision has exposed the cross-sections of two streams of different ages and with different fill.�� 28 Figure 3.22 The lower channel with oxidized Pliocene fill from the top of the plateau.�� 28 Figure 3.23 The contact between the two channel fills of the 130,000-year-old channel.�� 29 Figure 3.24 Younger stream cut-and-fill in the stream channel just north of the Vagnari vicus site.�� 29 Figure 3.25 The Irsina exposure east of the town (on the crest of the hill in the distance).�� 30 Figure 3.26 Stratigraphy of the Arroyo Italiano exposure southwest of the Vagnari vicus site.�� 31 Figure 3.27a–b Holocene record of fluvial deposits with superimposed soil horizons nested within the stream channel north of the Vagnari vicus site, with (a) a close up of the deposits, and (b) the well-developed angular-blocky structure of the soils.�� 32 Figure 3.28a–f Flood control canals constructed during the 1950s to move rainwater quickly out of fields on the valley floor accelerated stream velocities to the point that they began head-ward erosion in streams above the upper ends of the canals (a-b). This has now progressed to the Vagnari vicus site and beyond. Most canals have filled with eroded sediment

iv

from the tributaries and are choked with vegetation. Vagnari North exposure upstream erosion 2013 and 2018 (c-d). In 2013 (e) the main channel was one meter higher than in 2014 (f), and the channel was filled in places with armoured mud balls. �� 33 Figure 3.29 Comparison of the sum of spring and summer rainfall with the sum of the radiocarbon dates on erosion in southern Italy (red line). �� 34 Figure 3.30a–b Modern summer erosion currently typical of the tributaries of the upper Basentello River. �� 35 Figure 3.31 Principal component analysis of pollen from Lago Grande Monticchio core 90D from the crater of Monte Vulture.�� 35 Figure 3.32a–c Comparison of the pollen rations of the three dominant periods from Lago Grande di Monticchio.�� 37 Figure 3.33 Possible spring locality south of the Vagnari cemetery site.�� 38

CHAPTER 4

Figure 4.1 Simplified plan of excavated remains in the north-west part of the vicus, with attested and estimated room arrangements. Plan J. Moulton.�� 40 Figure 4.2 Levelled natural chalk and clay deposit with visible tool marks from Phase 1 (foreground). Photo M. Carroll.�� 41 Figure 4.3 Cylindrical pit 4021. Photo M. Carroll.�� 41 Figure 4.4 Grey-gloss oil lamp from the lowest deposit (4039) in cylindrical pit 4021. Photo M. Carroll.�� 41 Figure 4.5 Excavated remains belonging to Phase 2. Plan J. Moulton.�� 42 Figure 4.6 Robber trench (4057/6050) of the Phase 1 wall 4017, view from the north. Photo M. Carroll.�� 43 Figure 4.7 Wall 4006, faced on the north inner side (top) with a vertical tile facing. Photo M. Carroll.�� 43 Figure 4.8 Lower right: wall 4006; background: wall 4059; lower right, remnant of wall 4062. View from the south. Photo M. Carroll.�� 44 Figure 4.9 Drone photograph of trenches in 2018. Lower right: robber trenches of Phase 2 buildings; upper trench: wall 6022 (Phase 5). Photo V. Ferrari and G. Ceraudo.�� 44 Figure 4.10 Cobblestone floor (4015) of Room L, bounded by wall 4006. View from the west. Photo M. Carroll.�� 45 Figure 4.11 Wall 5003, view from the south. Photo M. Carroll.�� 45 Figure 4.12 View, from the east, of the stone drain 5045/4054, with parts of the side walls of the drain removed to show the cobblestone foundations. Photo M. Carroll. �� 46 Figure 4.13 View, from the west, of the stone-built drain 5013/2024. Photo M. Carroll.�� 46 Figure 4.14 View, from the west, of wall 5003 in which a gap for drain 5045 has been blocked up with masonry and mortar (middle of photo) when the drain was given up. �� 46 Figure 4.15 Grain storage pit 5008 with remains of its backfill containing pottery, loom weights, animal bone, and metal. Photo M. Carroll.�� 46 Figure 4.16 Grain storage pit 6061 with the remnants of the ceramic dolium still in situ. Photo M. Carroll.�� 46 Figure 4.17 Mortar-lined grain storage pit 4020 with the impression of a flat-bottomed dolium preserved at the bottom. Photo M. Carroll.�� 48 Figure 4.18 Excavating a back-filled grain storage pit 4053, view from the east, with wall 4006 (left) and drain 4054 (right). Photo M. Carroll. �� 48 Figure 4.19 Excavated remains belonging to Phase 3. Plan J. Moulton.�� 48 Figure 4.20 The cella vinaria at Vagnari, showing intact or partially intact dolia in mortar basins (6, 7, 9), robbed-out dolium impressions (2, 5, 8, 10), and circular darker traces of dolium holes not excavated (1, 3, 4). Plan J. Moulton.�� 49 Figure 4.21 Mortar-lined dolium basins, with a large fragment of a dolium preserved in basin 3020 (left) and just the basin for 3012 (right). View from the north. �� 50 Figure 4.22 Mortar-lined basin 3012 for a dolium, view from the north. Photo M. Carroll.�� 50 Figure 4.23 Channel 3025 cut through the mortar floor of the cella vinaria, view from the east. �� 51 Figure 4.24 The western part of drain 4054/5013 cut by the placement of dolium 3020 in Room A. View from the east.�� 51 Figure 4.25 View from the southeast of stone-built drain 5045 (left), blocked up so the water and waste could be diverted into stone drain 5051 (right). Photo M. Carroll.�� 51 Figure 4.26 Excavated remains belonging to Phase 4. Plan J. Moulton.�� 52 Figure 4.27 View from the south at east-west wall 3008 (top) abutting onto the older north-south running wall 3010 of Phase 2 (right). Photo M. Carroll.�� 53 Figure 4.28 View from the east of wall 2018, the western wall of the cella vinaria in Phase 2 (right), cut by wall 2013, the western wall of the portico building, in Phase 4 (left). Photo M. Carroll.�� 53 Figure 4.29 View from the southeast of the northern wall 1005 of the portico building (Room D) with a doorway and threshold. Photo M. Carroll.�� 53 Figure 4.30 Wall 3008 of Room D with its internal plaster skim in situ. Photo M. Carroll.�� 54 Figure 4.31 Burnt daub 2010/3011 lying on the beaten earth floor in Room D in the portico building. Photo M. Carroll.�� 54 Figure 4.32 Chunks of burnt daub 2010/3011 with black patches and burned out organic material. Photo M. Carroll.�� 54 Figure 4.33 The various courses of wall 5028. Photo M. Carroll.�� 55 Figure 4.34 Sondage with walls 4012 and 4018. Photo M. Carroll.�� 55 Figure 4.35 Stone-built drain 3035, the northern end of drain 5045 in Phase 3. Photo M. Carroll.�� 56 Figure 4.36 View from the north of stone-built drain 1009, the Phase 4 of the Phase 3 drain 3035/5051 (left). �� 56 Figure 4.37 Northern Phase 5 extension of stone-built drain 1009 where it pierces and flows through the northern outside wall 1008 of the portico porch. View from the south. Photo M. Carroll.�� 57 Figure 4.38 Excavated remains belonging to Phase 5. Plan J. Moulton.�� 58 Figure 4.39 Excavating the porch (Room E) of the portico building, with the external wall 2005 on the right. View from the east.�� 59 Figure 4.40 Detail of wall 6047. Photo M. Carroll.�� 59 Figure 4.41 Tile hearth 5070. Photo M. Carroll.�� 60

v

Figure 4.42 Tile hearth 5044. Photo M. Carroll.�� 60 Figure 4.43 Lower third of the dolium 3038, with a student as scale. Photo M. Carroll.�� 60 Figure 4.44 Human skeletons in the dolium 3038. Photo M. Carroll.�� 60 Figure 4.45 Excavated remains belonging to Phase 6. Plan J. Moulton.�� 62 Figure 4.46 Large fragments of a glass windowpane in situ. Photo M. Carroll.�� 63 Figure 4.47 Deposits 4052 and 5014 in Room C, with traces of burning. View from the north. Photo M. Carroll.�� 63 Figure 4.48 Collapsed roof tiles 1004, 3027, 4007 in Rooms B, C, and L. Running through the deposits is wall 4006. View from the east. Photo M. Carroll.�� 63 Figure 4.49 Collapsed roof, with tiles on either side of wall 4013 which separated Room B from Room C, later robbed out leaving this void. View from the west. Photo M. Carroll.�� 63 Figure 4.50 Wall 6022 foundations and post holes 6031, 6042 and 6043 from scaffolding used to dismantle the wall. View from the north. Photo M. Carroll.�� 63 Figure 4.51 Post hole 6031 from scaffolding in Phase 6. Photo M. Carroll. �� 64

CHAPTER 5

Figure 5.1 Regional Red Slip plate from backfilled grain storage pit 5037, Phase 2. Photo M. Carroll.�� 68 Figure 5.2 Eastern Sigillata red slip bowl FV300. Photo M. Carroll.�� 68 Figure 5.3 Fragment of a large, collared jar FV1043 from Phase 2. Photo M. Carroll.�� 69 Figure 5.4 Plain ware bowl FV1151. Photo M. Carroll.�� 69 Figure 5.5 Maker’s stamp on the mortarium FV1157/1161. Photo M. Carroll.�� 71 Figure 5.6 Lamp fragments from Phases 2, 5, 6, and 7. Photo D.R. Griffiths and H. Russ.�� 73 Figure 5.7 Lamp fragments from Phase O. Photo D.R. Griffiths and H. Russ.�� 74 Figure 5.8 Misfired plain ware cup FV605. Photo M. Carroll.�� 76

CHAPTER 6

Figure 6.1 Roman funerary relief (from Rome?) depicting a man and woman (left) on the premises of their vineyard and winery full of dolia. �� 82 Figure 6.2 Detail of the relief in Figure 6.1, showing lidded dolia in the winery. Photo M. Carroll.�� 83 Figure 6.3 Remains of a tall, cylindrical dolium (Cat. 6.3). Photo M. Carroll.�� 83 Figure 6.4 Fragment of a ceramic lid from a dolium (Cat. 6.6). Photo M. Carroll.�� 84 Figure 6.5 Lower third of dolium 3038 (Cat. 6.1) in situ with numerous cracks. �� 85 Figure 6.6 Finger indentations on the top of the vertical wall of the dolium (Cat. 6.1), an indication of having been coil built. Photo M. Carroll.�� 85 Figure 6.7 Fragment of a dolium with black volcanic temper and reddish-brown grit. Photo M. Carroll.�� 86 Figure 6.8 Map of central and southern Italy showing the Roman Magmatic Province and the Ernici-Roccamonfina Magmatic Province, as well as various relevant sites. Volcanoes are marked as red triangles. �� 87

CHAPTER 7

Figure 7.1 Vessel glass fragments. �� 93 Figure 7.2 Caroline Jackson examining all the window glass fragments from context 3040 (Cat. 7.WF01), with students Kayla Thiessen and Thomas Watson. Photo M. Carroll.�� 96 Figure 7.3 Window glass fragments from various contexts. �� 98

CHAPTER 8

Figure 8.1 Silver victoriatus, Cat. 8.2, obverse. Photo M. Stirn.�� 103 Figure 8.2 Silver victoriatus, Cat. 8.2, reverse. Photo M. Stirn�� 103

CHAPTER 9

Figure 9.1 Bronze finger ring, Cat. 9.1. Photo M. Carroll.�� 106 Figure 9.2 Bronze finger ring, Cat. 9.2. Photo M. Carroll.�� 106 Figure 9.3 Bronze arm ring, Cat. 9.3. Photo M. Carroll.�� 107 Figure 9.4 Bronze chain with threaded links, Cat. 9.4. Photo M. Carroll.�� 107 Figure 9.5 Bronze Mid-La Tène fibula, Cat. 9.5. Photo M. Carroll.�� 108 Figure 9.6 Bronze Aucissa fibula, Cat. 9.6. Photo M. Carroll.�� 108 Figure 9.7 Fragmentary bronze belt buckle, Cat. 9.7. Photo M. Carroll.�� 108 Figure 9.8 Bronze belt buckle, Cat. 9.8. Photo M. Carroll.�� 109 Figure 9.9 Bronze belt(?) stud, Cat. 9.9. Photo M. Carroll.�� 109 Figure 9.10 Bronze belt buckle tongue, Cat. 9.10. Photo M. Carroll.�� 110 Figure 9.11 Bronze cosmetic or medical instrument shaft, Cat. 9.11. Photo M. Carroll.�� 110 Figure 9.12 Bronze mirror with silver wash, Cat. 9.12, cut to be recycled. Photo M. Carroll.�� 110 Figure 9.13 Bronze solid-cast boar’s head attachment, Cat. 9.13. Photo M. Carroll.�� 111 Figure 9.14 Bronze solid-cast boar’s head attachment, Cat. 9.13. Photo M. Carroll.�� 111 Figure 9.15 Bronze nail, Cat. 9.14. Photo M. Carroll.�� 111 Figure 9.16 Detail of the incised head of the bronze nail Cat. 9.14. Photo M. Carroll. �� 111 Figure 9.17 Bronze strip, possibly a mount for the lock of a wooden box, Cat. 9.15. Photo M. Carroll.�� 111 Figure 9.18 Bronze mount for a wooden box, Cat. 9.16. Photo M. Carroll.�� 112 Figure 9.19 Bronze net repair needle, Cat. 9.17. Photo M. Carroll.�� 112 Figure 9.20 Bronze casting mould, Cat. 9.18. Photo M. Carroll.�� 112 Figure 9.21 Bronze sheet with nail holes and iron corrosion ring, Cat. 9.19. Photo M. Carroll. �� 112 Figure 9.22 Fragment of a bronze vessel rim(?), Cat. 9.19. Photo M. Carroll. �� 113

vi

Figure 9.23 Sheet metal strip of bronze, folded, Cat. 9.21. Photo M. Carroll.��113 Figure 9.24 Bronze sheet metal scrap, Cat. 9.22. Photo M. Carroll.��113

CHAPTER 10

Figure 10.1 Iron bit head, Cat. 10.1. Photo M. Carroll.��117 Figure 10.2 Iron bit head, Cat. 10.2. Photo M. Carroll.��117 Figure 10.3 Prong of an iron hoe, Cat. 10.3. Photo M. Carroll.��118 Figure 10.4 Fragmentary iron figure-of-eight links chain, Cat. 10.4. Photo M. Carroll.��118 Figure 10.5 Fragmentary large iron chain, Cat. 10.5. Photo M. Carroll.��118 Figure 10.6 Fragmentary large iron chain, Cat. 10.6. Photo M. Carroll.��118 Figure 10.7 Iron ring, Cat. 10.7. Photo M. Carroll.��118 Figure 10.8 Iron buckle, Cat. 10.8. Photo M. Carroll.��119 Figure 10.9 Iron steelyard arm, Cat. 10.9. Photo M. Carroll.��119 Figure 10.10 Various iron nails. Photo M. Carroll.��120 Figure 10.11 Fragmentary iron nail with discoidal head, Cat. 10.10. Photo M. Carroll.��120 Figure 10.12 Fragmentary iron nail with slightly domed head, Cat. 10.11. Photo M. Carroll.��120 Figure 10.13 Fragmentary iron nail with conical/pyramidal head, Cat. 10.12. Photo M. Carroll.��120 Figure 10.14 Fragmentary iron nail with circular head, Cat. 10.13. Photo M. Carroll.��120 Figure 10.15 Iron wall hook, Cat. 10.14. Photo M. Carroll.��120 Figure 10.16 Iron knife blade, Cat. 10.15. Photo M. Carroll.��121 Figure 10.17 Detail of metal ring on the tang of the iron knife blade, Cat. 10.15. Photo M. Carroll.��121 Figure 10.18 Iron knife blade, Cat. 10.16. Photo M. Carroll.��122 Figure 10.19 Iron knife blade, Cat. 10.17. Photo M. Carroll.��122 Figure 10.20 Fragmentary iron knife blade, Cat. 10.18. Photo M. Carroll.��122 Figure 10.21 Selection of iron hob nails. Photo M. Carroll.��122

CHAPTER 11

Figure 11.1 Iron smithing heart bottom. Photo M. Carroll.��126 Figure 11.2 Iron smithing heart bottom. Photo M. Carroll.��126 Figure 11.3 Iron tap slag. Photo M. Carroll.��126 Figure 11.4 Pie chart showing the types of slags, including undiagnostic slag (non-diagnostic slag) and smithing heart bottoms (SHB).��127 Figure 11.5 Piece of unworked iron, perhaps a partly worked bloom or a piece of iron bar to be reshaped.��126 Figure 11.6 Pie chart showing the main contexts for metallurgical debris.��127 Figure 11.7 Pie chart showing metal traces in a crucible from context 4027.��127 Figure 11.8 Pie chart showing metal traces from context 4023 on cobblestone floor 4015. ��127

CHAPTER 12

Figure 12.1 Clipped piece of lead sheet, Cat. 12.19. Photo M. Carroll.��129 Figure 12.2 Clipped piece of lead sheet with striations from rolling, Cat. 12.13. Photo M. Carroll.��129 Figure 12.3 Selection of clipped lead sheet pieces with various striations. Drawing I. De Luis.��130 Figure 12.4 Round lead ingot, Cat. 12.40. Photo M. Carroll.��130 Figure 12.5 Solidified lead droplet. Photo M. Carroll.��131 Figure 12.6 Various pieces of lead scrap for recycling. Photo M. Carroll.��132 Figure 12.7 Lead sheet scrap folded in itself, ready for recycling. Photo M. Carroll.��132

CHAPTER 13

Figure 13.1 Misfired roof tile with vitrified surface, Cat. 13.15. Photo M. Carroll.��136 Figure 13.2 Misfired roof tile in Figure 13.2, bubbly fabric on the reverse side. Photo M. Carroll.��136 Figure 13.3 Excavating the collapsed roof over Rooms B and C. View from the south. Photo M. Carroll.��137 Figure 13.4 Tegula from the collapsed roof in Phase 6, Cat. 13.1. Photo M. Carroll.��137 Figure 13.5 Imbrex from the collapsed roof in Phase 6, Cat. 13.4. Photo M. Carroll.��138 Figure 13.6 Fragmentary tegula, Cat. 13.2. Photo M. Carroll.��138 Figure 13.7 Fragmentary tegula, Cat. 13.2. Photo M. Carroll.��138 Figure 13.8 Fragmentary imbrex with a wavy line, Cat. 13.5. Photo M. Carroll.��138 Figure 13.9 Fragmentary imbrex with wavy lines, Cat. 13.6. Photo M. Carroll.��138 Figure 13.10 Tegula fragment with incised anchor or arrow, Cat. 13.7. Photo M. Carroll.��139 Figure 13.11 Brick fragment with cat paw prints, Cat. 13.9. Photo M. Carroll.��140 Figure 13.12 Tegula fragment with impression of a hob-nailed shoe, Cat. 13.12. Photo M. Carroll.��140 Figure 13.13 Brick fragment with rows of incised dots, Cat. 13.8. Photo M. Carroll.��140 Figure 13.14 Round tile fragment from a hypocaust, Cat. 13.13. Photo M. Carroll.��140 Figure 13.15 Segmental tile, Cat. 13.10. Photo M. Carroll.��141 Figure 13.16 Detail of the impression of textile on the back of segmental tile, Cat. 13.10. Photo M. Carroll.��141 Figure 13.17 Segmental tile fragment, Cat. 13.11. Photo M. Carroll.��141 Figure 13.18 Column made of segmental tiles at Minturnae in southern Latium. Photo M. Carroll.��141

CHAPTER 14

Figure 14.1a–f Pieces of marble revetment and mosaic.��144 Figure 14.2 White marble revetment with greenish-grey veins, Cat. 14.4, reverse side showing reworking and chisel marks. Photo M. Carroll.��145

vii

Figure 14.3 Detail of a piece of grey-veined marble revetment showing an incised line, Cat. 14.3. Photo M. Carroll.�� 145 Figure 14.4 (A) thin-section microphotograph of the white limestone composing the mosaic piece, Cat. 14.6, SF7-3-2001-13/8 (scale bar=0.5 mm, crossed polars); (B) thin section microphotograph of the bedding mortar surrounding the mosaic fragment (scale bar=0.5 mm, crossed polars); (C) XRPD pattern of the mosaic limestone; (D) XRPD pattern of the mosaic bedding mortar. �� 147 Figure 14.5 XRPD patterns of the investigated white marble inlays: (A) 3040; (B) SF91-4-4030-16-4/1; (C) SF91-4-4030-16-5/2; (D) SF24-4-3006-15; (E) SF25-4-3001-15; (F) SF49-4-5058-17.�� 148 Figure 14.6 Thin-section microphotographs of the investigated white marble inlays (scale bar = 0.5mm, crossed polars): (A) 3040; (B) SF91-4-4030-16-4/1; (C) SF91-4-4030-16-5/2; (D) SF49-4-5058-17; (E) SF24-4-3006-15; (F) SF25-4-3001-15.�� 149 Figure 14.7 MGS (mm) variation ranges and median values of the most important Mediterranean white marbles used in antiquity (data after Gorgoni et al. 2002) and values measured in the samples presented in this work.�� 149 Figure 14.8 (A) thin-section microphotograph of the impure marble SF51-4-4030-17, Cat.14.4, (scale bar=0.5mm, crossed polars); (B) XRPD pattern of the same sample.�� 151 Figure 14.9 Mn content of the white marbles analyzed in this study compared with the variation ranges and median values of the most important Mediterranean white marbles used in antiquity (data after Moens et al. 1988).�� 152 Figure 14.10 Scatterplot of the carbon and oxygen isotope compositions of the white marbles considered in this study and the confidence ellipses (probability distribution 99%) corresponding to some of the most important white marbles quarries exploited in classical times (data after Attanasio et al. 2006).�� 153

CHAPTER 15

Figure 15.1 Wall plaster fragment with dark greenish curvilinear motif on a greyish-white background, Cat. 15.1. Photo M. Carroll.�� 156 Figure 15.2 Wall plaster fragments with red paint, Cat. 15.2. Photo M. Carroll.�� 156 Figure 15.3 Wall plaster fragments with red and yellow paint, Cat. 15.3-4. Photo M. Carroll.�� 157

CHAPTER 16

Figure 16.1 Basalt hand-mill of the biconcave type, Cat. 16.1. Photo M. Carroll.�� 158 Figure 16.2 Detail of basalt hand-mill, Cat. 16.1, showing the horizontal handle socket. Photo M. Carroll.�� 158 Figure 16.3 Basalt hand-mill of the top concave type, Cat. 16.2. Photo M. Carroll.�� 158 Figure 16.4 Tall stone mortar of fine-grained limestone, Cat. 16.4: a. rim fragment; b. rim fragment. Photos M. Carroll.�� 159 Figure 16.5 Inside polished and worn surface of the stone mortar, Cat. 16.4: a. inside rim and body fragment; b. inside body fragment. Photo M. Carroll.�� 159 Figure 16.6 Tenon hole in the body of the stone mortar, Cat. 16.4. Photo M. Carroll.�� 160 Figure 16.7 Detail of the tenon hole in the stone mortar, Cat. 16.4. Photo M. Carroll.�� 160 Figure 16.8 Ovoid cut stone weight, Cat. 16.5. Photo M. Carroll.�� 160 Figure 16.9 Ovoid cut stone weight, Cat. 16.6. Photo M. Carroll.�� 160

CHAPTER 17

Figure 17.1 Loom weight, Cat. 17.3, with markings on the front. Photo M. Carroll.�� 164 Figure 17.2 Detail of loom weight Cat. 17.3. Photo M. Carroll. �� 164 Figure 17.3 Loom weight, Cat. 17.4. Photo M. Stirn.�� 165 Figure 17.4 Possible coin impression of a chimaera on the top of loom weight Cat. 17.7. Photo M. Carroll.�� 165

CHAPTER 18

Figure 18.1 One side of a worked bone sheet, cut or sawn along the two long edges, Cat. 19.1. Photo M. Carroll.�� 171 Figure 18.2 Other side of the worked bone sheet Cat. 18.1. Photo M. Carroll.�� 171 Figure 18.3 Fragmentary shaft of a bone pin or awl, Cat. 18.7. Photo M. Carroll.�� 171 Figure 18.4 Fragmentary bone spatula head needle, Cat. 18.6. Photo M. Carroll.�� 171 Figure 18.5 Bone needle with a broken tip, Cat. 18.6. Photo M. Carroll.�� 172 Figure 18.6 Bone hair pin with a spherical head, Cat. 18.4.�� 172 Figure 18.7 Bone spoon with a round bowl, Cat. 18.2. �� 172 Figure 18.8 Small bone stopper from a bottle, Cat. 18.3. �� 172

CHAPTER 19

Figure 19.1 Relative proportions of main domesticates for Phases 2–6. NISP = 295. Zones = 250.�� 176 Figure 19.2 Cattle astragalus measurements from Vagnari and comparative southern Italian sites. �� 179 Figure 19.3 Sheep/goat astragalus distal breadth (Bd) vs greatest lateral length (GLl).�� 179 Figure 19.4a–b Fossil shells from the local geology at Vagnari. Photo A. Trentacoste.�� 181 Figure 19.5 Map showing the location of Vagnari and comparative Roman sites in southern Italy referred to in this chapter. �� 183 Figure 19.6 Relative proportion of the main domesticates from Vagnari and comparative sites. �� 184

CHAPTER 20

Figure 20.1 Discriminant analysis results showing Vagnari samples compared with Amorgos crop processing stages (Jones 1987).�� 192 Figure 20.2 Vagnari cereal proportions by phase.�� 193 Figure 20.3 Correspondence analysis of Vagnari samples, categorised by phase. Text labels were left visible in this figure to highlight plant species that influenced sample clusters.�� 194 Figure 20.4 Weed seed flowering seasonality proportions from Vagnari samples.�� 195

viii

CHAPTER 21

Figure 21.1 The dolium defossum with the top layer of the vicus skeletal material exposed. �� 206 Figure 21.2 Cranial elements associated with Individual A. Photo T.L. Prowse.�� 207 Figure 21.3 Lower limb bones associated with Individual A. Photo T.L. Prowse.�� 207 Figure 21.4 Lower limb bones associated with Individual B. Photo T.L. Prowse.�� 208

CHAPTER 22

Figure 22.1 Drain 5045 (E-W) and drain 5051 (N-S) with soil removed. Photo M. Carroll.�� 210 Figure 22.2 Roundworm (Ascaris sp.) eggs from drain 5045 at Vagnari vicus. Scale bars are 20μm.�� 212

CHAPTER 23

Figure 23.1 3-D reconstruction of the northwest sector of the vicus, view from the west. In the foreground is the cella vinaria. Drawing I. De Luis.�� 217 Figure 23.2 Map of the outline of the vicus, as investigated through geophysics and excavation, with the plotted scatter of roof tile in red. �� 218 Figure 23.3 Satellite image of the Vagnari plateau with the roof tile scatter plotted by C. Small and the vicus buildings superimposed. The sharp edges of the scatter indicate the limits of the survey zone. Image I. De Luis.�� 218 Figure 23.4 3-D reconstruction of the northwest sector of the vicus, view from the south. On the upper right is a courtyard building possibly serving as a market or a mansio. Drawing I. De Luis.�� 220 Figure 23.5 View of the Vagnari plateau, with the Casa Cantoniera along the road in the middle of the photo and the masseria Vagnari to the left. Photo M. Carroll.�� 223 Figure 23.6 The Fontana della Bonifica near the Casa Cantoniera on the road to Vagnari. Photo M. Carroll.�� 223 Figure 23.7 Excavated remains superimposed on the geophysics plot. The reservoir is the large rectangular feature at the bottom. Plan I. De Luis.�� 224

ix

List of Tables

Table 3.1 Middle to Late Quaternary Marine Terrace ages before the present of coastal terraces in the Metaponto area and their Marine Isotope Stage (MIS) assignment (after Sauer et al. 2010; Brückner 1980).�� 20 Table 3.2 The Blytt-Sernander periods.�� 36 Table 5.1 Bulk data by phase by count/weight (grams).�� 66 Table 5.2 Relative proportions by ware class (by weight) and phase.�� 66 Table 5.3 Functional analysis by category and form using relative proportions (based on estimate vessel equivalents -EVEs- of featured vessels).�� 77 Table 5.4 Functional analysis using relative proportions (based on estimate vessel equivalents -EVEs- of featured vessels).�� 77 Table 6.1 Summary of analytical results, with lipids and selected sources identified. �� 88 Table 7.1 Frequency table for vessel glass fragments by phase and context.�� 95 Table 7.2 Frequency table for window glass fragments by phase and context. �� 97 Table 9.1 Number of finds per function group.�� 106 Table 9.2 Measurements of finger rings. �� 107 Table 14.1 Summary of principal mineralogical and textural characteristics of studied marbles derived by X-ray diffraction analysis (XRD) and observation of thin sections by the polarizing microscope (OM).�� 146 Table 14.2 Major (%wt) and trace elements (ppm) of the studied samples.�� 150 Table 14.3 C and O stable isotopic composition of the analysed samples.�� 151 Table 17.1 Phase of occupation and correspondent number of loom weights found in it. �� 163 Table 17.2 Calculation of the number of warp threads per loom weigh, per two loom weights, and of the warp threads per centimetres using a loom weight from context 4039.�� 167 Table 17.3 Calculation of the number of warp threads per loom weigh, per two loom weights, and of the warp threads per centimetres using a loom weight from context 5008.�� 167 Table 17.4 Calculation of the number of warp threads per loom weigh, per two loom weights, and of the warp threads per centimetres using a loom weight from context 4002.�� 167 Table 19.1 Bone preservation by phase. Only includes specimens with zones.�� 176 Table 19.2 Number of identified specimens (NISP) and zone counts by phase. Remains from flotation heavy fractions and hand collected specimens are grouped together.�� 177 Table 19.3 Body part distribution for Phases 2–6: minimum animal units (MAU) for each element.�� 178 Table 19.4 Bone fusion of the main domestic taxa, Phases 2–6.�� 178 Table 19.5 Mandible wear stages for the main domestic taxa, Phases 2–6. Wear stages follow Payne (1973) and O'Connor (1988).�� 179 Table 19.6 Land snails, zone counts for Roman phases.�� 181 Table 19.7 Livestock representation on comparative sites in southern Italy.�� 185 Table 20.1 Cultivated and wild taxa from Vagnari, Phase 1 samples.�� 196 Table 20.2 Cultivated and wild taxa from Vagnari, Phase 2 samples.�� 197 Table 20.3 Cultivated and wild taxa from Vagnari, Phase 3 samples.�� 199 Table 20.4 Cultivated and wild taxa from Vagnari, Phase 4 samples.�� 201 Table 20.5 Cultivated and wild taxa from Vagnari, Phase 5 samples.�� 203 Table 20.6 Cultivated and wild taxa from Vagnari, Phase 6 samples.�� 204 Table 22.1 Transmission routes and symptoms of common intestinal parasites found in the Roman Empire.�� 210 Table 22.2 Drain and context numbers of soil samples collected from Vagnari.�� 211

x

List of Plates

Plate 1 Colour-coded plan of the excavated remains in the northwest sector of the vicus, without the tile collapse. Plan J. Moulton.�� 346 Plate 2 Colour-coded plan of the excavated remains in the northwest sector of the vicus, with the tile collapse. Plan J. Moulton.�� 347 Plate 3 3-D reconstruction of the northwest sector of the vicus. Reconstruction I. De Luis.�� 348 Plate 4 Pottery Phase 1, scale 1:3. Drawing D.R. Griffiths, photo M. Carroll and H. Russ.�� 349 Plate 5 Pottery Phase 2, scale 1:3. Contexts 4060, 5042, 5046, 5050, 5062. Drawing D.R. Griffiths.�� 350 Plate 6 Pottery Phases 2 and 3, scale 1:3. 1–5 Phase 2, contexts 6046, 6053, 6062; 6 and 7 Phase 3, contexts 2012 and 3025. Drawing D.R. Griffiths.�� 351 Plate 7 Pottery Phases 4 and 5, scale 1:3. 1-3 Phases 4 and 4/5, contexts 1020 and 5008; 4-10 Phase 5, contexts 3013, 4023, 5047, 6022. Drawing D.R. Griffiths.�� 352 Plate 8 Pottery Phase 6, context 4030 below tile collapse, not including lids, scale 1:3. Drawing D.R. Griffiths.�� 353 Plate 9 Pottery Phase 6, context 4030 below tile collapse, lids only, scale 1:3. Drawing D.R. Griffiths.�� 354 Plate 10 Pottery Phase 6, below tile collapse, scale 1:3. Drawing D.R. Griffiths.�� 355 Plate 11 Pottery Phase 6, tile collapse contexts 1004 and 3024, scale 1:3. Drawing D.R. Griffiths, photo M. Carroll and H. Russ.�� 356 Plate 12 Pottery Phase 6, deposits 3005, 3006 and 3022 above tile collapse, scale 1:3. Drawing D.R. Griffiths.�� 357 Plate 13 Pottery Phase 6, contexts 2010, 4029 and 4052, scale 1:3. Drawing D.R. Griffiths.�� 358 Plate 14 Pottery Phase 6, context 5014, scale 1:3. Drawing D.R. Griffiths.�� 359 Plate 15 Pottery Phase 7, topsoil and plough soil deposits, scale 1:3. Drawing D.R. Griffiths.�� 360 Plate 16 Pottery Phase 7, topsoil and plough soil deposits, scale 1:3. Drawing D.R. Griffiths.�� 361 Plate 17 Pottery Phase O, contexts 1003, 3015, 1016, 2003 and 2006, scale 1:3. Drawing D.R. Griffiths.�� 362 Plate 18 Pottery Phase O, contexts 2015, 2016, 2019, 3019 and 4010, scale 1:3. Drawing D.R. Griffiths.�� 363 Plate 19 Pottery Phase O, contexts 5015, 6029, 6054 and 6059, scale 1:3. Drawing D.R. Griffiths.�� 364 Plate 20 Fragments of dolia and dolium lids, contexts 2019 and 2022 (Phase O), 4006/4009 (Cat. 6.6), 4028 (Cat. 6.4) and 4029 (Cat. 6.5), probably Phase 3, scale 1:5. Drawing D.R. Griffiths.�� 365 Plate 21 Vessel glass. 1. Cat. 7.VF11 - base and body of a patella cup; 2. Cat. 7.VF14 - base and body of a patella cup; 3. Cat. 7.VF12 - base of an unidentified vessel, most likely a cup; 4. Cat. 7.VF10 - base of an unidentified vessel; 5. Cat. 7.VF15 – base of a bowl on foot; 6. Cat. 7.VF13 - base of an unidentified vessel, most likely a cup or a bowl; 7. Cat. 7.VF5 - rim of an unidentified vessel, possibly a cup. Scale 1:1. Drawings S. Cann, montage C. Bertini.�� 366 Plate 22 Vessel glass. 1. Cat. 7.VF16 and Cat. 7.VF17 - fragment of a bottle handle; 2. Cat. 7.VF21 -fragment of the body of a vessel glass decorated with vertical ribs; 3. Cat. 7.VF22 - fragment of the body of a vessel glass decorated with vertical ribs; 4. Cat. 7.VF23 - fragment of the body of a vessel glass decorated with vertical ribs; 5. Cat. 7.VF18 - fragment of the body of a vessel glass decorated with ribs; 6. Cat. 7.VF8 - fragments of neck, probably from a small bottle; 7. Cat. 7.VF1 fragment of neck, probably from a small bottle. Scale 1:1. Drawings S. Cann, montage C. Bertini.�� 367 Plate 23 Iron, scale 1:2. 1. Cat. 10.16, knife blade; 2. Cat. 10.15, knife blade: 3. Cat. 10.5, large chain; 4. Cat. 10.4, small chain; 5. Cat. 10.8, buckle. Bronze, scale 1:1. 6. Cat. 9.13, boar head attachment; 7. Cat. 9.5, fibula; 8. Cat. 9.6, Aucissa fibula; 9. Cat. 9.9, stud; Cat. 9.11, medical instrument. Drawings S. Cann and I. De Luis.�� 368 Plate 24 Bronze, scale 1:1. 1. Cat. 9.17, net repair needle; 2. Cat. 9.16, furniture mount; 3. Cat. 9.14, decorated nail; 4. Cat. 9.1, finger ring; 5. Cat. 9.8, buckle. Lead, scale 1:1. 6. Cat. 12.33, shell weight; 7. Cat. 12.32, weight with a stone; 8. Cat. 12.44, net weight; 9. Cat. 12.41, net weight; 10. Cat. 12.53, cap or lid. Drawings S. Cann and I. De Luis.�� 369 Plate 25 Lead, scale 1:2. 1. Cat. 12.54, round base with bronze inlay; 2. Cat. 12.48, reinforcing strip with iron nails; 3. Cat. 12.52, pipe fragment; 4. Cat. 12.51, ceramic repair tenon. Marble, scale 1:3. 5. Cat. 14.2, fragment of Dokimeion marble; 6. Cat. 14.1, fragment of Pentelic marble; 7. Cat. 14.7, fragment of Pentelic marble; 8. Cat. 14.3, fragment of Dokimeion marble; 9. Cat. 14.4, white marble with greenish veins. Drawings S. Cann.�� 370 Plate 26 Worked stone, scale 1:3. 1. Cat. 16.4, rim fragment of a limestone mortar; 2. Cat. 16.4, section through the body of the limestone mortar; 3. Cat. 16.4, detail of the limestone mortar with a tenon hole. Worked bone, scale 1:2. 4. Cat. 18.3, bottle stopper. Worked bone, scale 1:1. 5. Cat. 18.2, spoon; 6. Cat. 18.5, needle; 7. Cat. 18.4, hair pin. Ceramic loom weights, scale 1:2. 8. Cat. 17.8, fragmentary loom weight; 9. Cat. 17.10, fragmentary loom weight; 10. Cat. 17.16, fragmentary loom weight; 11. Cat. 17.1, fragmentary loom weight. Drawings S. Cann and I. De Luis.�� 371 Plate 27 Ceramic loom weights, scale 1:2. 1. Cat. 17.2; 2. Cat. 17.3; 3. Cat. 17.4; 4. Cat. 17.5; 5. Cat. 17.6; 6. Cat. 17.7; 7. Cat. 17.19; 8. Cat. 17.11. Drawings S. Cann.�� 372 Plate 28 Ceramic loom weights, scale 1:2. 1. Cat. 17.12; 2. Cat. 17.13; 3. Cat. 17.4; 4. Cat. 17.15; 5. Cat. 17.17. Drawings S. Cann.�� 373

xi

xii

Preface and Acknowledgements Maureen Carroll

In this monograph, the historical and physical context of the imperial estate and its central vicus is explored. The results of the 2012–2019 excavations and study seasons are presented, and questions are addressed that pertain to the role of the imperial estate in the Roman economy and the importance of the emperor as landowner in the context of elite status and competition.

contributions to the project. He was ably helped in the last years by Paolo Moramarco. The colleagues in the Centro Operativo of the Soprintendenza in Gravina, —Cira Digiesi, Francesca Ariani Picciallo, and Michele Colonna—, provided excellent support (and coffee) over the years in the old monastery of San Sebastiano, and we forged a warm working and personal relationship with them. Vito Nicefalo in Gravina kindly oversaw the security of the excavation in 2018. I was very pleased to accept the generous offer of Veronica Ferrari and Giuseppe Ceraudo from the Laboratorio di Topografia antica e fotogrammetria dell’Università del Salento to carry out the excellent drone photography of the site and its landscape in 2018. The Fondazione Ettore Santomasi Pomarici in Gravina enabled our students to delve into the history of the area by providing guided tours of their collection of archaeological artefacts from Silvium, and their library was a peaceful place to study and consult publications.

An endeavour such as an excavation requires considerable input from a variety of people, especially if it runs for several years, and I have the pleasurable task here of acknowledging a range of individuals and institutions. First and foremost, I would like to express my gratitude to Alastair and Carola Small who encouraged me to continue their research in the vicus and who so graciously offered their help in many ways and shared information with me over the years. They have been a joy to work with. Tracy Prowse, whose excavations take place in the field opposite ours, has shared insights and information on the burials she investigates, shouldered responsibilities for permit applications, and been an excellent partner in organising all aspects of the infrastructure of an excavation. Myles McCallum and Hans vanderLeest, who directed the excavations on top of the hill at San Felice, also were generous with their expertise and time, and helped out with many matters of fieldwork organisation in the early years of the vicus excavation.

Our students benefited every year from a tour of underground Gravina and I should like to thank Michele Parisi, president of the Associazione Gravina Sotteranea, for his unflagging support and interest in the local archaeology. Other good friends in Gravina and Altamura include Tonio Creanza and Jennifer Bell who run the Fornello Cave Project and Messors workshops on cultural heritage. Their hospitality and friendship were always important and much appreciated. We have enjoyed the hospitality of several restaurants and pizzerias in Gravina, and some of their owners have gone beyond the call of duty in providing for our teams, especially Filippo Garibaldi, Tommaso Lobaccaro, and Raffaele Mercede. Franco Taccogna’s famous lasagna lunch also was a highlight every summer.

Mario de Gemmis Pelliciari, who very sadly passed away in November 2020, was an ideal landowner, always being helpful and interested in what we found at Vagnari. He was generous with his time, and his willingness to allow us to dig holes in his wheat fields every year and to use the 18th-century Masseria Vagnari as an equipment store was much appreciated. I am also grateful to the Soprintendenza for the annual granting of permits for fieldwork and scientific analysis of artefacts, in particular Luigi LaRocca, Francesca Radina, Maria Rosaria De Palo, and Marisa Corrente. Colleagues at the British School at Rome, primarily Stefania Peterlini and Stephen Kay, provided pivotal institutional support for the permit application process. Preparatory work on site plans and permit documentation would not have been possible without the valuable expertise of Franco Taccogna in Gravina. Measuring in our trenches and digitally drawing site plans every year were his important

I am deeply indebted to the people who acted as site supervisors in the vicus fieldwork: Jonathan Boffey, Chris Griffiths, Giuseppe Garofolo, and Christina Westhoff, and, especially, Jonathan Moulton, who was with me from the very first season and who can fix anything, including a car, and Kelsey Madden, whose enthusiasm for the site was infectious. Various colleagues have discussed aspects of the site and its finds with me, from soils and rocks, and dolia and burnt daub, to cisterns and granaries, and I would xiii

like to thank them warmly here. These include Caroline Cheung, Suzanne Frey-Kupper, Dennis Kehoe, Camilla Norman, Andrew Morcom, Silvia Pallecchi, Tim Parkin, Chrissy Partheni, Paul Pettitt, Ben Russell, Sandro Sebastiani, Marleen Termeer, Astrid van Oyen, Peter Wigand, and Andrew Wilson.

Masterson, Jamila Moraes, Mackenzie Priest, Holly Rosevear, Evan Wynne, Yue Zhang.

I am very fortunate to have worked with excellent colleagues and specialists on the compilation of this volume. It has been wonderful to collaborate with Donatella Barca, Camilla Bertini, Coralie Clover, Jane Evans, Stefanie Hoss, Renato Giarrusso, David Griffiths, Caroline Jackson, Marissa Ledger, Louis-Olivier Lortie, Victoria Lucas, Kelsey Madden, Piers Mitchell, Giuseppe Montana, Jonathan Moulton, Vanessa Pashley, Tracy Prowse, Luciana Randazzo, Giovanna Scopelliti, Rebecca Sgouros, Alastair Small, Ben Stern, Matt Stirn, Angela Trentacoste, Beatrice Triozzi, Peter Wigand, and David Wigg-Wolf. The beautiful artefact drawings by Sally Cann are an asset to the publication. I am grateful to Erin McGowan from Archaeopress who was instrumental in producing plates of these drawings. Irene de Luis, with great patience and skill, has produced plans, drawings, and a 3-D reconstruction of the vicus, for which I am very grateful. Hannah Russ kindly produced the figures of Roman lamp fragments. My colleague at the University of York, Helen Goodchild, provided the excellent map of volcanic zones in western Italy. And Jonathan Moulton, who has drawn up the site plan and the phase plans, was always a valuable collaborator in figuring out where features in the trenches go and what they mean, never complaining when I requested yet another alteration to a plan. I am grateful to Beatrice Triozzi for her translation of the synopsis of our research results into Italian. I thank David Davison warmly for accepting this volume for publication with Archaeopress.

2017: Amy Brogan, Colby Lorenz, Madeleine Nelson, Wayne Oldfield, Nicole Phillips, Tom Rose, Briana Sands, Molly Stevens, Nick Travaglini.

2016: Caroline Cervera, Sophia Draznin-Nagy, Janae Lunsford, Kelsey Madden, Ben Radford, Molly Stevens, Kayla Thiessen, Nick Travaglini, Thomas Watson.

2018: Casey Boettinger, Greg Bowen, Sara Chick, Sarah Eaton, Kathryn Faragher, Sarah Hayes, Katie Hullock, Georgina Goodison, Andrew Morcom, Caitlyn Pallas. None of this research would have been possible without financial support from many generous funding bodies. The British Academy-Leverhulme Trust has been a valuable contributor three times for various aspects of the project, the Roman Society has helped financially from the beginning, and the Rust Family Foundation was a supporter in 2017, 2018, and 2019. A grant from the Social Sciences and Humanities Research Council of Canada (SSHRC) to Tracy Prowse (PI) and me (CoPI) enabled us to study the lead artefacts and leadworking debris and to explore possible lead poisoning of the vicus inhabitants as it is detectable in the dental remains of the skeletons. We thank the Lincoln College Zilkha Trust for the C14 funds and the Oxford Radiocarbon Accelerator Unit for expertise. Finally, the University of Sheffield provided the infrastructure and support for many phases of the research, and the alumni office, the Erasmus exchange programme, and Your Global Sheffield made funds available to some students to enable them to experience archaeology and fieldwork in a small Italian town. The University of York also provided financial assistance with aspects of the publication. I am grateful to all of them.

I should like to thank all the student volunteers from Britain, Canada, Australia, the U.S.A., the Czech Republic, Portugal, China, and Singapore who participated in the excavations and who withstood very high temperatures, occasional locust plagues, and, at least once a season, torrential rainfall and mudslides:

Finally, this volume was written over the course of 2020 and 2021, the most unimaginably difficult year(s), due to the corona virus taking its toll on so many people and affecting so many lives. There have been some delays in finishing the book, originally planned for December 2020. Library access has been limited, and access to museum collections has been non-existent. Had the situation been different, there might have been some references we could have chased up or some artefact comparanda we might have located, but in the current circumstances that would have necessitated further delays in completing our work. We have done remarkably well to get the job done and disseminate the valuable knowledge gained in this project, and I am grateful to all colleagues who have gone the extra mile to fit this research into their disrupted schedules.

2012: Joao Freire de Andrade, Aaron Kulakiewicz, Gabrielle Lawrence, Jonathan Moulton, Lucie Paštiková, Mara Polansky, John Quarrell, Victoria Rose, Martin Trenz. 2013: Sam Bromage, Jak Martin, Otis Gilbert, Mark Mason, Sharnvir Dhillon, Lindsay Mitchell, Courtenay Chrichton-Turley, Catherine Kendall, James Platt, Florence Douglas. 2015: Abigail Birks, Esmie Carter, Amy Derrick, Robin Downing, Kathryn Goulding, Marcus Losty, Molly

xiv

CHAPTER 1

Introduction. Exploring the Imperial Estate at Vagnari, 2012–2019 Maureen Carroll

Figure 1.1 Map of South Italy showing the location of Vagnari and other relevant settlements. Drawing I. De Luis, after a map by C. Small.

Via Appia, the ancient north-south road originating in Rome, was located, and this major Roman artery, as well as a nearby east-west drove way (tratturo) from upland Lucania to Gravina used for transhumance in antiquity, must have contributed considerably to the prosperity of Vagnari and even to the selection of this location for the settlement (Gabba 1990; Small 2011a: 11–12; Small and Small 2011: 383–86; Piepoli 2014; Adamo 2016: 94).

The site of the Roman settlement at Vagnari lies about 15km northwest of Gravina in Puglia (Provincia Bari) in agricultural land that today is heavily ploughed and planted annually with wheat (Figure 1.1). It lies on a gently undulating plateau overlooking a valley to the north, on the other side of which the land rises up to Monte Marano (Figure 1.2). South of Vagnari, the land slopes upwards to the San Felice plateau (Figure 1.3). To the west lies the valley of the Basentello River, a tributary of the Bradano, the main watercourse in the west separating the pre-Apennine hills from the plateau of the Murge in the east. The Bradano, emptying into the Ionian Sea off the south coast of Italy, marks the boundary between Apulia (modern Puglia) and Lucania (modern Basilicata). Somewhere very near Vagnari, the

The discovery of the Roman village or vicus and the extensive land holdings that belonged to it is the achievement of Alastair and Carola Small who mapped this large agricultural estate (potentially at least 25km2 in size) through field-walking and surface collection in 1999 and 2000. The collected material, including 1

The Making of a Roman Imperial Estate

Figure 1.2 View from San Felice over the Vagnari plateau (yellowish-green vegetation, photo middle), the slope of Monte Marano (right), and the water of the Lago di Serra del Corvo/Diga del Basentello in the distance. Photo M. Carroll.

Figure 1.3 Drone photo of the Vagnari plateau. The vicus lies to the left of the ravine (in the centre of the photo), the cemetery to the right. In the background right is the hill of San Felice. Photo Veronica Ferrari and Giuseppe Ceraudo, Laboratorio di Topografia antica e fotogrammetria dell’Università del Salento.

2

Carroll: 1. Introduction. Exploring the Imperial Estate at Vagnari, 2012–2019

Figure 1.4 Original geophysics plot-out by John Hunt with excavated remains superimposed on it. Drawing J. Moulton.

the roof tiles made and stamped by imperial slaves, confirmed that the settlement on the plateau at Vagnari was Roman in date and that the estate to which it had belonged was an imperial property established in the 1st century AD. It was the largest and richest site detected in the entire survey zone. Three distinct, but related, focal points of the imperial estate subsequently were investigated archaeologically (Figure 1.3). The vicus on the Vagnari plateau north of the ravine was explored in a series of exploratory trenches in 2001– 2010 by teams led by Alastair Small from the University of Edinburgh and Giuliano Volpe from the University of Foggia (Favia et al. 2011; Favia et al. 2011a). The Roman cemetery, lying south of the vicus on the other side of the ravine, has been the focus of ongoing investigations by Tracy Prowse from McMaster University since 2003 (Small et al. 2007; Prowse and Small 2009; Prowse et al. 2014; Brent and Prowse 2014). Survey work by the Smalls also led to the discovery of another Roman site

on the hill above Vagnari at San Felice. This modest villa site, built in the second half of the 1st century BC, and in imperial possession from the 1st to the mid-2nd century AD, was excavated by Myles McCallum from St. Mary’s University and Hans vanderLeest from Mount Allison University from 2005 to 2013 (McCallum et al. 2011; McCallum and vanderLeest 2014). Two years after Alastair Small completed his fieldwork in the vicus in 2010, the University of Sheffield began to excavate and explore the buildings, the agricultural facilities, and the manufacturing provisions in the village. Our aim was to gain insight into the socioeconomic complexities of the estate, the role of slave and free labour, and the working and living conditions of the inhabitants. Crucial in deciding where to begin excavating was the resistivity plot-out produced by John Hunt in 2006–2007. Figure 1.4 shows the original geophysics plan with the excavated remains 3

The Making of a Roman Imperial Estate

Figure 1.5 The photo (2012) shows the stone-built drains under the walls and floors of the portico building that showed up on the geophysics as lines. Photo M. Carroll.

of all excavation campaigns superimposed on it. In the northern sector of the plot-out, a structure was discernible that was almost 30m long and with a southwest-northeast orientation, and it appeared that this building was divided internally by walls into small rooms of identical or very similar size. These invited comparison with rows of cells known at some Roman villas and often interpreted as slave quarters, although not without deserved scepticism (Carandini and Ricci 1985: 157–63, 175–80, figs. 157–58, 167–68; Marzano 2007: 129–48; Marzano and Métraux 2018: 17–18). It is clear that imperial slaves worked at Vagnari, and they will have worked at other imperial estates in the region. The imperial slave Gratus, who oversaw a tile-making operation for the emperor and stamped his tiles with Grati Caesaris, for example, not only supplied the vicus at Vagnari with roof tiles, but also the buildings at nearby San Felice and San Gerolamo as part of the same estate (Small et al. 2003). And the imperial properties in Apulia were not only numerous, but also large in size, with an extensive workforce and population (Maiuro 2012: 203). The emperor would have owned many slaves in his workforce, especially on a rural estate, and so it was entirely plausible for us when we began work at Vagnari that slave quarters might have formed part of the vicus, although we wanted to test whether the structures or the associated material culture would enable us to identify them (Small et al. 2003; Chelotti

1994 and 2007). With the resistivity plot in mind, the site of the building potentially subdivided into small cells on the northern edge of the vicus was the one chosen to launch the Sheffield project, but cells of identical size are not what we uncovered, because the ‘walls’ dividing the building into what looked like rooms turned out simply to be stone drains under the floor of the building (Carroll 2014) (Figure 1.5). As it became clear in subsequent excavation seasons, the rooms in this building, however they originally were divided from each other, were probably multipurpose, being used for storage, crop processing, habitation, and other things we cannot identify. Whether slaves were involved in anything happening in this building cannot be proven. Far from having a disappointing first season, however, we were encouraged by the condition and extent of the archaeological remains to continue working here, opening large trenches annually until 2018 (Figure 1.6; see also Plates 1–2). Two study seasons were conducted in 2014 and 2019. The investigations have far exceeded our expectations. Despite the lack of a luxury villa or imperial residence on it, the vicus cannot be categorised simply as a ‘Roman industrial village’ (Gualtieri 2018: 168), because the excavations have shown that this imperial estate had a broad economic basis, ranging from cereal crop cultivation to tile production and 4

Carroll: 1. Introduction. Exploring the Imperial Estate at Vagnari, 2012–2019

Figure 1.6 Overall multi-phase plan of excavated remains 2012–2018. Plan J. Moulton.

metal-working, with further diversification to include viticulture in the 2nd century AD, all of which have left traces in the vicus (Carroll 2016; Carroll 2019; Montana et al. 2021). A rich array of material culture, including ceramics, glass, metals, and worked stone and marble, has been retrieved, and extensive assemblages of archeobotanical and faunal remains have been added to the physical evidence for life at Vagnari. The excavations also have revealed a range of stone-built structures in this sector of the vicus that were first erected in the early 1st century AD, possibly under Augustus, and subsequently altered, expanded, and refurbished until shortly before the middle of the 3rd century. All together, these strands of evidence shape our understanding of the diversity of the economy of the estate and the role of the vicus and its inhabitants in organising and managing work and income for the emperor in Italy and beyond. This is important also because this imperial estate is situated in a landscape that is a considerable distance from Roman urban

centres in eastern and coastal Apulia, all of which have been the subject of far more intense archaeological exploration and whose economic, cultural, and social integration in the Roman state is better understood (Grelle and Silvestrini 2013; Fioriello and Magiatordi 2013; Grelle et al. 2017). Our research at Vagnari also highlights the great importance of this site for an understanding of earlier periods following the Roman conquest of the region, long before the imperial estate was established. Although we have considerable information from Roman historical texts about key events, military campaigns, and named leaders involved in expansionist activities and conflicts in the 3rd century BC, the ancient literary sources on southeast Italy in the 2nd century are scarce (Yntema 2013: 237–43). It is on archaeology that we must rely for an understanding of the profound impact of the Roman conquest on the culture, society, and economy of the populations inhabiting the settlements and landscapes 5

The Making of a Roman Imperial Estate Field survey by Carola and Alastair Small suggested that, on the basis of diagnostic pottery, the Vagnari plateau had been inhabited in the 4th century BC, with occupation ceasing in the 3rd century BC (Small 2011a: 16; C. Small 2011: 61–62, fig. 2.13). This is now supported and confirmed by our fieldwork. There was a lack of 3rd-century material in our excavation, although we did retrieve a Neapolitan bronze coin of the third century BC and a silver victoriatus minted in Rome in 211 BC. Very occasionally fragments of residual pottery predating the 3rd century were encountered. This settlement, and others in the Peucetian countryside in western Apulia, would have been connected to and perhaps dependent on Silvium. Small (2014c: 22) suggests that an area of c. 100 km2 must have fallen within its territory. They all must have suffered with the conquest of Silvium in 306, the Roman conflicts between Rome and Carthage in the 3rd century, and the Roman reprisals aimed at the supporters of Hannibal, the Peuceti included, at the end of the second Punic war in 202 BC (Grelle 2010; Fronda 2010: 253–79, 307–24). Disruption and abandonment in the 3rd century not only are attested archaeologically at Vagnari, but also at other smaller settlements in the vicinity, as recent fieldwork at Jazzo Fornasiello near Gravina and at the Iron Age settlement on the San Felice plateau indicates (Lambrugo and Pace 2017: 36–37; Depalo 2017: 35; Small

in the relevant regions. It is in this context that the exploration of the site of Vagnari has an immense value. Pre-Roman Apulia was inhabited on the coast by various Italic peoples, including the Dauni in northern Apulia and the Messapi in the south, but the central coast and western inland area of Apulia belonged to the Peuceti, an independent and culturally important political entity in the Iron Age with a major defended settlement on an extensive plateau, today’s Botromagno, opposite Gravina (Small 2011a: 15; Small 2014a; Lombardo 2014) (Figure 1.7). Rome had embarked on campaigns to annex and control the independent territories in Italy from the 4th century BC, and Apulia was a key region for such activities (Fronda 2010: 13–34). The large Peucetian settlement on Botromagno, known to Greek and Roman writers as Silvium, was sacked by the Romans and prisoners were taken in 306 BC (Diodorus Siculus, Library of History 20.80). After this, the settlement collapsed and went into sharp decline, as excavations in the 1960s, 1970s, and 1990s indicated (Small 1992: 13–15; Small 2001: 44–45). Recovery in the 3rd century was not possible because of continuing conflicts on southern Italian soil, above all between Rome and Carthage and their respective allies (Fronda 2010: 53–99).

Figure 1.7 The Iron Age settlement of Silvium (Botromagno), seen from Gravina in Puglia. Photo F. Taccogna.

6

Carroll: 1. Introduction. Exploring the Imperial Estate at Vagnari, 2012–2019

2001: 40, 44; Small and Small 2017: 14; on Lucania, see Isayev 2007: 169–74).

coastal Apulia redistributed through centuriation and assigned for agricultural and commercial exploitation to individuals, such as demobilised soldiers and new settlers, the region around Silvium and the surrounding rural settlements in more remote western Apulia was left undivided and used in part as grazing land (de Ligt 2004; Roselaar 2009; Small 2014a: 54–56). Chelotti (2014: 253–54) highlights the ‘political-institutional void’ which characterised some areas of Apulia in the second half of the 2nd century BC, including the territory in which Peucetian settlements such as Silvium, Altamura, and Gioia del Colle-Monte Sannace had been located.

Land conquered by the Romans became common property of the Roman state, ager publicus, however the impact of this annexation in Apulia can be measured best through archaeological exploration, again underscoring the value of excavating at Vagnari (Grelle 2010; Roselaar 2010). Roman historians such as Appian (Civil Wars 1.7) and Horace (Epode 3) gave the impression that Apulia long after the Roman conquest was not a desirable or liveable place, and they referred to it as ‘made desolate by war’ and ‘parched and dry’. But this is political propaganda, and archaeology tells a different story.

Whether the late Republican settlement and its lands at Vagnari were managed by a member of the Roman or the Apulian elite, we do not know. Neither do we know the mechanism for the acquisition of the land by the emperor in the early 1st century AD, but because Silvium and its hinterland possibly belonged to unassigned ager publicus, a realistic scenario is that a piece of this simply was carved out for the imperial property at Vagnari (Chelotti 2010; Chelotti 2014: 252– 54, 257–58). This is related to the settlement history of this unurbanised region which did not see the transfer of Roman state land assets to municipal ownership in the 2nd and 1st centuries BC (Mangiatordi 2011: 429, n. 61; Fiorielli and Mangiatordi 2013). And there was a link between imperial domains and public land holdings, the former often deriving from the latter (Purcell 2014: 270–71). After an interruption of late Republican occupation at Vagnari of perhaps a few decades, the curtain opens on the first phase of Vagnari as a property newly acquired by the princeps, and the vicus flourished; the villa on the hill at San Felice, until then run privately, also is now incorporated into the imperial estate and refurbished with tiles supplied by the same Gratus Caesaris, as at Vagnari (Small et al. 2003). Given that Vagnari was established in an area without an urban character, a large village such as this that acted as an administrative and redistributive centre of the emperor’s extensive domain not only will have attracted a diverse population to live and work there, but also functioned as a nucleated settlement providing some of the daily amenities, social connections, and economic networks offered by bigger towns and cities elsewhere in the southeast of Italy. If the survey data in the Ager Venusinus has been interpreted correctly, the territory surrounding the Roman colony of Venusia/ Venosa (founded 291 BC, reinforced 200 and 43 BC) in the late Republican and early Imperial period is marked by a plethora of small and medium-sized farms and villas, but villages are lacking (Launaro 2011: 137–38, 302–16, Table A.24). Because the colony of Venusia was the urban focal point with political and social roles in this region, villages, as nucleated rural settlements, may not have been needed to fill settlement gaps or provide otherwise missing services. Further west in the

Crucial structural and artefactual evidence emerged in 2016 that demonstrates the inaccuracy, or at least the politically biased nature, of Roman historical sources such as these. After a hiatus of perhaps a century, people were living again on the Vagnari plateau in the 2nd century BC. This resuscitation of occupation fits the picture of renewed occupation driven, perhaps, by Roman economic interests, with local involvement, in the late Republican period elsewhere in the region (Roselaar 2014). A new Roman villa and vicus were built on the ruins of Silvium in the mid-2nd century, for example, probably with a focus on agriculture and textile production, and the contemporaneous establishment of a Roman villa built over the remains of Iron Age structures is attested at Monte Irsi on the border between Apulia and Lucania (Small 1977: 99– 101; Small 1992: 15–18; Small 2001: 39, 44–49; Goffredo 2017: 307–09). Furthermore, on the San Felice plateau above Vagnari, a private villa was established in the second half of the 1st century BC (McCallum et al. 2011: 36; McCallum and vanderLeest 2017; Goffredo 2017: 311–13). The late Republican settlement at Vagnari, thriving possibly because of its advantageous location on the Via Appia and the tratturo connecting it to Silvium and sites in Lucania and beyond to eastern Apulia, continued to be occupied until the mid- or late 1st century BC. Alastair Small has argued convincingly that the Via Appia was not extended into Apulia until the second half of the 2nd century BC, and this must have been a significant factor in reviving settlement at Vagnari (Small 2011a: 18–19). Economic opportunities offered by the Roman dominance of southern Italy and the resulting connectivity with areas beyond southern Italy and the Adriatic also will have fostered this renewed occupation of land still classified as Roman ager publicus, but worked and exploited by Italians and Romans alike (Roselaar 2014; Roselaar 2019: 103–08). Whereas the Gracchan agrarian reform law of 133 BC saw arable land in the fertile territories in eastern and 7

The Making of a Roman Imperial Estate explored. The results of the 2012–2019 excavations are discussed, with an overview of buildings and other features according to the phases of settlement to which they belong. Specialist reports and discussions on all categories of the material culture and environmental material retrieved at Vagnari take up the next sections of the volume. A concluding chapter contextualises the late Republican and Imperial settlements and their population, the evidence for agricultural and industrial production, and the water supply of the plateau. It also addresses the question of ownership, continuity, and the role of the imperial estate in the Roman economy and the social, political, and ideological importance to the emperor of this form of land ownership in the framework of elite status and competition.

territory of Silvium, on the other hand, the lack of towns of Roman municipal status (municipia) any closer than 50–60km (Genusia/Ginosa, Rubi/Ruvo di Puglia) must have been a factor that contributed to the significance of the large, nucleated settlement at Vagnari and the assumption of its function as a political, social, and economic focus. The longevity of the central settlement of the imperial estate is testament to its importance to the region (Maiuro 2012: 295). Its position on the Via Appia, linking Vagnari to Venusia in the northwest and Tarentum to the southeast on the Ionian Sea, since 122 BC a Roman colony, must have gone a considerable way to alleviate some of the remoteness of the whole district (Marchi 2014; Piepoli 2014). In the next chapters, the historical and physical context of the imperial estate and its central vicus is

8

CHAPTER 2

The Topographical Context of Vagnari in the Roman Period: A Brief History of the Study Area Alastair Small In 1965 when I first became involved in the archaeology of the territory of Gravina in Puglia, scholarly interest (such as it was) was focussed on the Peucetian city of Botromagno and next to nothing was known of the archaeology of the area in the Roman period. The main topics that were debated then were the course of the Via Appia, the identification of the Iron Age site on Botromagno with the settlement known to the Romans as Silvium, the duration of the site, and its relationship to the medieval ravine city of Gravina. The debate was carried out in a vacuum of archaeological evidence, and it involved a good deal of imaginative theorising. The literary texts were scanty and there was no authentic epigraphic evidence. When Theodor Mommsen came to write up the area in volume IX of the Corpus Inscriptionum Latinarum (abbreviated as CIL), published in 1883, he referred to Silvium only briefly in discussing the Antonine Itinerary (AI) as a 3rd-century AD source for the topography of Apulia. He provided no entry under the toponym because he found no inscriptions that he could allocate to it. This whole tract of land was epigraphically a void: Titulis universus hic tractus destituitur.

There was a long history of identifying Silvium with Garagnone. The theory was first put into print by Lucas Holstenius (Lukas Holste, 1596–1661) in his Annotationes on Italia Antiqua written by Philippus Cluverius (Philipp Clüver, 1580–1622), and published posthumously in 1624 (for Holstenius, see especially Vian 2014). This was the first serious geographical work on ancient Italy, in which Cluverius analysed the ancient toponyms region by region and summarised the Greek and Latin sources which refer to them. In his section on places in the southern part of Apulia (pp. 700–07: Lib IV. Cap. XII. De Mediterraneis Apuliae locis) he took account (inter alia) of Venusia, Bantia (Banzi) and Mons Vultur (Monte Vulture), and in his section on the Messapian peninsula (pp. 707–20: Cap. XIII. De Messapia peninsula) he discussed Tarentum at some length, but, Banzi apart, he had nothing to say of the region between Venosa and Taranto. Holstenius, whom Cardinal Francesco Barberini had appointed as his librarian, filled this gap with information on the iter from Venusia to Tarentum contained in a set of manuscript maps held in the cardinal’s library in which the courses of the itinera throughout the kingdom of Naples were (in Holstenius’ view) accurately represented.1 The word iter (plural itinera) can include a route of any kind — a foot-path or track as well as a road maintained by royal authority —, so in all likelihood the iter which Holstenius saw on the map was the drove road for transhumant sheep (tratturo) registered by the Dogana della Mena delle Pecore which skirted the territory of Venosa and passed below the scarp of the Murge by way of Garagnone to Gravina before leading on across the Murge to Taranto and the Ionian Gulf (Figure 2.1). The Annotationes, which Cardinal Barberini published five years after the death of his protégé, were mere notes, jotted down without discussion. In the section which interests us, Holstenius listed a series of toponyms shown on the map which he identified with stations on the Via Appia mentioned in the AI. Since the Itinerary located Silvium XX Roman miles from Venusia, he identified it with Gorgolione, i.e. Garagnone; the next station in the Itinerary, Blera (XIII miles from Silvium) he located at Gravina, and the next after that, Sub Lupatia (XIIII miles from Blera), at

Apart from the lack of epigraphic evidence, there was a problem in reconstructing the ancient topography of the area in that the figure given for the stretch of the Via Appia between Venusia and Silvium in the AI and that shown on the map known as the Tabula Peutingeriana, or Peutinger Table (from now on, TP), a medieval copy of a Roman map of the known world between Venusie and Silutum (presumably a corruption for Silvium) varied. In the AI, the distance is XX Roman miles (Cuntz 1929: section 121.3); in the TP it is XXXV (Levi and Levi 1967: section VI.5). Mommsen believed that the AI was more likely to be correct — in spite of the fact that the total figure which it gives for the distance between Venusia (Venosa) and Tarentum (Taranto) is too short by about 10 Roman miles — and on this assumption he located Silvium at Garagnone where there was a medieval castle on the edge of the Murge, about three quarters of the way between Gravina and Spinazzola (Figure 2.1). At least that is what he wrote on p. 27 of CIL IX. But the great man must have nodded, because in the map Tab. II, which he commissioned from Heinrich Kiepert and published at the end of the volume, Silvium is located at Gravina.

Inter [erroneously for Iter] Venusia Tarentum sic explicandum existimo ex optimis Tabulis manuscriptis Emin. Card. Barberini, ubi itinerum ductus per Regnum Neapolitanum accurate exprimuntur.

1

9

The Making of a Roman Imperial Estate

Figure 2.1 Map showing settlements of the Mid-Imperial period and topographical features mentioned in the text. The green line outlines the Basentello survey area. B=Botromagno; G=Garagnone; LS=La Santissima; MI=Monte Irsi; MM=Masseria Macchitella; MS=Monte Serico; SF=San Felice; SS=Santo Staso; ST=Santa Teresa; V=Vagnari. Base map by Carola Small.

Altamura. In no case does he cite any ancient remains that might support these identifications.

The idea that Silvium was located at Garagnone persisted for most of the 19th century. Even Giovanni Jatta, scholarly local antiquary and founder of the Museo Jatta in his family’s palazzo at Ruvo, who had grave doubts about Pratilli’s reliability, followed him in locating Silvium at Garagnone (Jatta 1844: 46–47), although he did not believe that it was situated on the low ground where Pratilli’s paving slabs were supposedly located. Pratilli, he said, must have been asleep when he passed the place since he made no mention of the castle at Garganone which Jatta believed to occupy the site of the ancient city.

This gap in evidence was supplied, after a fashion, by the Jesuit priest and antiquary, Francesco Maria Pratilli, in his monumental work on the Via Appia published in 1745. Pratilli accepted Holstenius’ identification of Silvium with Garagnone, and to prove that this was an important ancient site he invented an inscription recording the holding of triennial games and a distribution of food to the people in honour of Liber Pater which had been found, he said, in the vicinity of Garagnone. He also claimed to have seen disturbed paving slabs of the road in the same area. Mommsen despised Pratilli as an antiquary who laced his work with forgeries, and he rejected his text of the dedication to Liber Pater as a fraud, but he nevertheless accepted the identification of Silvium with Garagnone, and, with the support of his authority, the identification of the Via Appia with the line of the drove road that passed below that site came to be generally accepted. No later scholar, however, mentions seeing the paving slabs recorded by Pratilli, and Calderoni Martini (1920: 38) said that he could find no evidence of them.

Kiepert who supplied the map for CIL IX had placed Silvium at Garagnone in the editions of his Atlas Antiquus up to and including that of 1876. But as we have seen, he relocated it to Gravina in CIL IX published in 1883. Evidently, he had read or heard something which led him to transfer the toponym from Garagnone to Gravina. I have not been able to discover precisely what it was with the limited library resources available to me during the Covid19 lockdown in which this chapter is being written, but it is very likely to have been connected with the revived interest in the TP generated by Desjardins’ publication of the western 10

Small: 2. The Topographical Context of Vagnari in the Roman Period

sections of this late medieval copy of an ancient map in 1869–1874. Desjardins did not himself question the identification of Silvium/Silutum with Garagnone. In fact, he was so convinced of its reliability that he thought that the xxxv miles (53km) clearly shown on the map for the distance of Silutum from Venusie must have been a scribal error for xviii (Kiepert 1876: 208). But he had made the evidence more easily available for other scholars to assess.

la città antica era posta su di una collina prossima alla città moderna) — i.e. on Botromagno. The idea that Silvium should be identified with Botromagno was definitively reinforced when Pasquale Calderoni Martini, a distinguished figure in the political life of Gravina and a knowledgeable coincollector, published a small monograph on Gravina e l’antica Silvium in 1920, followed by an article with the title ΣΙΔΙΝΩΝ published in the Bollettino del Circolo Numismatico Napoletano for 1921. The Greek term (of the Sidini) was the legend on an extremely rare issue of bronze coins of late 4th-century BC type which showed the head of Zeus on the obverse, and on the reverse, Hercules resting on his club imitating Tarentine types inspired by the famous bronze statue of Lysippus.

Some more recent scholars (notably Radke 1971: col. 114) have attempted to resolve the conflict between the evidence of the AI and the TP by assuming that Silutum of the TP was an altogether different statio from Silvium of the AI, but that is unlikely, since in the AI it is listed after Blera as an intermediate statio on the Via Appia between Venusia and Sub Lupatia, and in the TP it is shown on a road from Venusie to Sublubatia. The more straightforward view must be that Silutum is a scribal error for Silvium, and that the figure of xxxv miles given in the TP for the distance between Venusia and Silvium must be assessed on its own merits. The reading would locate Silvium somewhere a little short of Gravina, which is distant between ca. 38 and 42 Roman miles from Venosa, depending on the route taken.

The coin-type was already known from two examples in the British Museum which R.S. Poole published without illustration in the Catalogue of the Greek Coins in the British Museum (1873: 395, nos. 1 and 2). He recorded them as coming from Side or Sidis, an unknown place in Apulia or (Roman) Calabria. Raffaele Garrucci, however, was bolder: he republished one of the British Museum pieces with an illustration taken from a plaster cast (1885: part II, 119, plate XCV, no. 42), and suggested that the Sidini, whose location was unknown, might be the inhabitants of Silvium. This suggestion was ignored by B.V. Head when he published the pieces in the British Museum more fully in the Historia Numorum in 1911 (p. 49). He demonstrated that the coin was one of a small group minted by Apulian cities in the late 4th century BC, but he declined to identify its place of origin.

It is likely that it was the reading in the TP that led Heinrich Kiepert to locate Silvium at the site of medieval and modern Gravina in his map in the CIL IX, and it is an interesting indication of his dedication to topographical scholarship that he continually modified his atlas to take account of new ideas. This tradition was upheld by his son Richard, who inherited his father’s mapping business after his death in 1899. In the 1903 edition of the Atlas he moved Silvium yet again — across the ravine to the west bank of the Torrente di Gravina where Botromagno is situated, showing that he continued keep the Atlas in the forefront of topographical scholarship.

Calderoni Martini, however, had one of these coins in his own collection2 which had been found on Botromagno, and he was therefore able to claim that the city of the Sidini (which might have been called Sidion) could be identified with that settlement; and he argued that the Greek term Sidion passed by a simple phonetic change from Sidion to Sivion, and was then changed by the Romans, for reasons of assonance to Silvium. He might have added that the name Silvium evoked the forest environment — the silvae — which probably surrounded the city. The most likely time for the name-change to have occurred (as we now know, but Calderoni Martini did not) was when a new settlement was created on the hill in the second half of the 2nd century BC, overlying the remains of the Peucetian city.

It took some time, however, for the idea that Silvium was situated at Botromagno to be generally accepted, and as late as 1895 Christian Hülsen, another great name in ancient topography, wrote an article on the Appia via in the Realencyclopädie der classischen Altertumswissenschaft (Band II.1: 238–42), in which he confused the issue further by giving the distance between Venusia and Silvium as 35 Roman miles, but locating Silvium nevertheless at ‘Garagnone bei Spinazzola’. Heinrich Nissen in the second volume of Italische Landeskunde (1902: 861) left the question open, merely noting that the 20 miles of the AI would locate Silvium at Garagnone, and the 35 of the TP at Gravina. Nunzio Jacobone in his book on Venusia published in 1909 (p. 73) was more decisive. He accepted the reading in the TP and located Silvium firmly at Gravina — except that the ancient city was set on a hill very close to the modern city (solo che

Calderoni Martini’s argument prevailed, and it is now generally accepted that the settlement on Botromagno was Silvium, the city of the Peuceti which, according to He left his collection to the Fondazione Santomasi in Gravina. The numismatic collection of the Fondazione has been published in 1995 by G. Libero Mangieri. Calderoni Martini’s coin is no. 24 in the catalogue. His argument is further confirmed by discovery by A. Florido of another on Botromagno (no. 25).

2

11

The Making of a Roman Imperial Estate Strabo (Geography 6.3.8), was situated furthest into the interior. Diodorus (Historical Library 20.80) tells us that it was captured by the Romans in 306 BC, and Plutarch (Sulla 27.6) records that a slave of Pontius, evidently a south Italian nobleman with an Oscan name, met Sulla there on his march from Tarentum to Rome in 83 BC. Its inhabitants, the Silvini, were mentioned by Pliny (Natural History 3.105) in his list of the communities in the interior of the Augustan district Regio II. Nothing else is known of Silvium from literary sources, other than that there was the statio with that name on the Via Appia recorded, as we have seen, in the AI and repeated in other late Roman Itineraries (the Anonymous of Ravenna 4.35 and Guido 48, but without distance measurements).

with the Serra Trono degli Avuzzi, ca. 1.5km north of Botromagno and separated from it by the watercourse of the Canale Capo d’Acqua, but the village is more likely to have been near the Fontana Sant’Angelo which is situated immediately below Botromagno and on the same side of the Canale. Given the vagueness of these toponyms it is impossible to locate the findspot precisely; but it was evidently not on the hill-top. Nevertheless, Nardone took these coins to indicate that the settlement of Silvium continued well into the 5th century AD. In his view, it was damaged in the great earthquake of AD 455 and destroyed in the following year by Genseric and his Vandals as they returned to Africa after sacking Rome. The surviving inhabitants then founded a new settlement in the ravine that developed into the city of Gravina. This is a surprising conclusion, since Nardone had already noted the two coins of Zeno which post-dated the Vandal invasion. He also muddied the water on the Via Appia by republishing the fraudulent dedication to Liber Pater which Pratilli claimed to have seen near Garagnone, apparently unaware that it had been debunked by Mommsen; and he cited the map in CIL IX as evidence that Mommsen located Silvium at Gravina, equally unaware of the fact that in the text Mommsen had located the settlement at Garagnone.

Calderoni Martini, however, also claimed that he had found numerous coins of the imperial period in (unspecified) excavations which proved that Silvium must have been flourishing in the Early Empire, although the finds became rarer from the Antonine period in the mid-2nd century onwards, and silver coinage disappeared after Alexander Severus in the first decades of the 3rd century AD. Even bronze coinage came to an end in the time of Julian the Apostate (AD 361–363). Nevertheless, he argued, Silvium continued to be inhabited down to AD 411 when it was destroyed by Alaric and his Goths in their invasion of South Italy. The surviving inhabitants took refuge in the caves of the ravine creating a nucleus of settlement which became the city of Gravina (Calderoni Martini 1920: 58–60).

These bold assumptions about the later development of Silvium did not detract from Calderoni Martini’s primary argument that the discovery of coins of the Sidini on Botromagno proved that Silvium was to be identified with the archaeological site on the hill-top, and the question of the location of Silvium was thus more or less settled by the early 1920s. But the route taken by the Via Appia between Venosa and Silvium/ Botromagno was another matter. Jacobone (1909), Ashby (1916), and Calderoni Martini (1920), writing in the first half of the 20th century, all accepted the theory that between Venosa and Gravina the Via Appia followed the route below the scarp of the Murge road by way of Garagnone. In 1952, however, Giuseppe Lugli, professor of ancient Roman topography at the University of Rome, published an article in which he combined the evidence of aerial photographs with the lists in the AI and other Itineraries to trace the line of the Via Appia between Rome and Otranto (Lugli 1952: 288–89). When he came to the stretch between Venusia and Silvium he ignored the theory that the road passed by way of Garagnone, and he proposed instead that it took a more westerly course from Venosa to below Palazzo San Gervasio where he claimed that there was a stetch of Roman paving to be seen. From there the road followed the valley of the Basentello river to its confluence with the Roviniero. The ancient topography of this area has since been drowned by the reservoir created when the dam was built in 1974 at the confluence of the two rivers to collect water for irrigation; but Lugli argued that

Calderoni Martini’s ideas were accepted, by and large, by his fellow Gravinese Domenico Nardone, another outstanding representative of the long Italian tradition of local historians and antiquaries, who took account of them in pp. 5–8 of the second edition of his book Notizie storiche sulla città di Gravina, published in 1922 (and republished by F. Raguso and M. D’Agostino with additional notes in 1990). Nardone was primarily interested in the medieval and later periods, and he was at his best in dealing with documentary evidence, but he had only limited understanding of the archaeology of the Roman period. In an introductory chapter he summarised what little was known of Silvium which, like Calderoni Martini, he saw as the immediate precursor of the medieval ravine city of Gravina. But whereas Calderoni Martini thought that the Roman city came to an end in the time of Alaric, Nardone argued for a rather later date. In the Museum of the Fondazione Santomasi at Gravina there were coins of post-Alaric emperors, including a gold solidus of Theodosius II (AD 408–450); and in his own collection Nardone had a solidus of Leo I (457–474) and two of Zeno (474–491) which, he said, had been found in the contrada Serralavozze, where there was the village of Sant’Angelo, a short distance from Botromagno. The contrada can probably identified 12

Small: 2. The Topographical Context of Vagnari in the Roman Period

the road might have taken one of two routes: either it crossed the ridge of Monte Marano (which would have involved a steep ascent) to the valley of the Pentecchia river which it followed to Silvium, or it passed by way of the Masseria Boldrini, Masseria Monsignore, Ponte Spinalva, and Casale Nardone to the same destination. The first two toponyms have changed since Lugli’s maps were made, but they appear on the 1865 edition of the IGM map for the area which makes it clear that the route which Lugli envisaged passed close to Vagnari and followed the most direct route to the ridge of San Felice. It then ran along the crest of the ridge before descending to join the Strada Statale no. 96 at the Ponte Spinalva. For most of its distance it corresponds to the line of the Via Appia shown on the map in Figure 2.1.

pattern. He published his preliminary results in 1972 in an article in which he described the sites he had encountered along both the suggested routes: the more northerly route below the scarp of the Murge and the more southerly route by the Basentello valley. He showed that there was no Roman site of importance in the vicinity of Garagnone on the more northerly route where Holstenius, Pratilli, and Mommsen had placed Silvium, and he decided that the more southerly route was more probable. Not only was it shorter, it also served a number of important sites, most obviously one at Monte Serico (Site V14 on the map in Figure 2.1) where a Roman settlement on the line of the road had replaced the Iron Age site near the hill-top. It was not, however, until after the publication of the article that Vinson came across the site of Vagnari which he described as ‘one of the largest and richest Roman sites that we found in Apulia’ (Small and Small, forthcoming). It was situated close to the road which Lugli had envisaged as leading from the confluence of the Basentello and Roviniero rivers to the ridge of San Felice (as in the first part of Lugli’s second option), though from there Vinson believed that it descended to the valley of the Torrente Pentecchia which it followed to Gravina (the latter part of Lugli’s first option). That road, in his view, corresponded to the Via Appia, and I agree with him for the most part, although I have argued elsewhere that it probably ran along the ridge of San Felice and only descended to cross the Pentecchia river once it had reached the end of the ridge at Santa Teresa (Small 2019).

That is essentially how matters stood at the time the British School at Rome became involved in a program of excavation and surface research on and around Botromagno. It began with a four-week field survey on the hill site carried out by Campbell Macknight and myself in 1965 in which we collected a wide range of material covering the whole of the 1st millennium BC, and the first two decades of the 1st millennium AD, but nothing later (summarised in Small 1992a: 25–27). The idea that the Roman city of Silvium lasted on the hilltop into the 5th century AD was, therefore, completely wrong. A subsequent field survey carried out by a team from the University of Lancaster in 1985 confirmed the point (Whitehouse 1986; Whitehouse et al. 2000: 289–95), as did yet another field survey done by the Consorzio Sidin between 1996 and 1998 (Naso 2000; Terrenato and Taylor 2000).

In 1971, at about the same time as Vinson came upon Vagnari, a small excavation was carried out by the Superintendency at the site of Santo Staso immediately below the hill of Botromagno, on its south side looking across the Pentecchia to Santa Teresa, and situated close to where the Via Appia could be expected to have passed. Only a brief notice of the excavation was published at the time in a summary of conference proceedings (Carletti 1972), and the real importance of the site could not be appreciated until the finds were published rather more fully by Sardone in 1984. The excavation exposed two walls of the 4th century BC and a third of the Late Antique period, loosely associated with a number of fragments of terracotta plaques decorated in relief with Christian symbols (Lattanzi 1984). They have been dated by G. Bertelli (2002) on the basis of their iconographic devices and style to the 6th century AD. Elena Lattanzi, who directed the excavation, thought that they could not have been used in the building partly exposed by the excavation, and must have been a deposit of plaques intended for use elsewhere. But much remains uncertain and the whole site warrants further excavation. A surface collection carried out on the site in 2003 and 2004 by

The excavations carried out between 1966 and 1970, mostly under the direction of Joan du Plat Taylor, also produced unexpected evidence. Some remains of the Peucetian city were uncovered, especially tombs and the walls of the later 4th century BC, but the great majority of the structures belonged to a Late Hellenistic settlement founded in the second half of the 2nd century BC which lasted into the early 1st century BC (Small 1992). Subsequent excavations have revealed more of the settlement of this period (Ciancio 1997: 241–65; Santoriello 2000; Whitehouse et al. 2000: 249– 58). Recently published evidence has shown that it was besieged and largely destroyed in the 70s BC, probably in the War of Spartacus, after which only a small part of the site continued to be occupied down to the Early Principate (Schinco and Small 2020; Small 2020). While the first phase of the excavations on Botromagno was in progress, and for several years afterwards, Sterling Peter Vinson carried out an extensive survey of the whole of the area between Gravina and Venosa with the aim of resolving the question of the line of the Via Appia in the light of the Roman settlement 13

The Making of a Roman Imperial Estate Annalisa Di Zanni and others showed that it had a long history extending from the early Iron Age to the Hellenistic period, and then again, perhaps after a short gap, from the early imperial period to the end of the 6th or early 7th century AD (brief preliminary report in A. Small and C. Small forthcoming; full publication pending). But the main significance of the site, as I have argued elsewhere (Small 2019: 235), is that it must be the Silvium of the Itineraries, and that the name of the abandoned hill-top settlement was preserved in the community established along the nearby road below the scarp of the hill.

1960s and early 1970s in conjunction with the British School’s excavations at Gravina.

Once it is accepted that the site at Santo Staso is the Silvium of the Itineraries, it becomes possible to calculate the distances along the Via Appia more precisely. Assuming that the southern solution is correct, the actual distance along the road between Venusia and Silvium would have been ca. 56km, or 38 Roman miles, only a little more than the XXXV miles of the TP, and significantly closer to it that the distance along the northern route by Garagnone which would have measured 62.5km, just over 42 Roman miles. But, as we have seen, the AI gives the distance as only XX miles, so it is very probable that an intermediate road station situated ca. XV miles from either Venusia or Silvium has dropped out of the list in this section of the road in the text as transmitted. The most probable candidate is Vinson’s site V16, located c. 14.8 Roman miles from Venusia, which he described as comparable in size and importance to Vagnari, although his site V14 (Monte Serico), situated 15 miles from Silvium/ Santo Staso is also possible.

The boundary of our own survey is marked in Figure 2.1 with a strong green line; those of the older surveys are less clear, but the oblique lines run in parallel across areas which probably lay outside the area covered. I have included in the map sites of this period recorded by Maria Luisa Marchi to the west of Spinazzola (Marchi 2010) and those found by Myles McCallum and his team in a field survey in the vicinity of Monte Serico (McCallum and Hyatt 2014; McCallum et al. 2013; McCallum et al. 2015). Vagnari is situated near the northwest corner of our own survey area on the fertile land where the plateau has collapsed in an immense land-slip into the basin where the site is situated. To the south, there is a large empty zone crossed by the main east-west drove road that led into the Lucanian mountains. This must have been an area of rough grazing or perhaps forest.

Our own field survey of 1996 to 2008 helps to fill the gaps left by the older surveys between the scarp of the Murge and the Basentello river, and allows us to produce a distribution map of all the known sites occupied in the middle imperial period (i.e. in the 2nd and 3rd centuries AD) when the settlement at Vagnari on the north side of the ravine reached it maximum development, and before it was redeveloped in the Later Empire on the south side (Figure 2.1).

Elsewhere, the map shows ribbons of sites spaced out at irregular intervals on the arable land above the flood plains of the rivers and on the tops of the plateaus, as at San Felice above Vagnari, where the villa excavated by the Canadian team directed by Hans vanderLeest and Myles McCallum was situated (McCallum et al. 2011). It is likely to have been the administrative centre of the vast imperial property of which the vicus at Vagnari was the main habitation and production centre. The estate filled the triangle between the Basentello river, the east-west drove road, and the Via Appia, and perhaps extended beyond it to the Pentecchia river. In the Early Empire it had been thinly populated, but in the 2nd century AD the process of subdividing the imperial property into smaller landholdings rented out, probably, to coloni had already begun.

Vinson did not discuss the possibility that Vagnari was the missing road station in his 1972 article because he had not yet discovered the site at the time the volume went to press. But in any case, it is an improbable candidate since it is only 10.5km or 7.1 Roman miles from Santo Staso where we have located Silvium. But it may, nevertheless, have been a minor station on the road. The stations on the Via Appia recorded in the AI were probably all mansiones, equipped with inns where travellers could stay overnight. Between them would have been lesser stations, mutationes, where there would have been changes of horses for those on official business (Gelsomino 1966: esp. 172), and it is possible that Vagnari was one of these.

To the northwest of Vagnari there was a string of sites spaced out along the line of the Via Appia, comfortably above the flood plain of the Basentello. A few may have been villas, but most were probably smaller farms, and one or two may have been vici. Beyond them to the north, settlement was thinner, even along the line of the drove road that connected the coastal plain of the Ionian Gulf with the North Apulian plain, until it reached the valley of the Torrente di Gravina and its

Vinson found other sites too after the publication of his 1972 article, but this part of his work has remained unpublished until now. His list of sites and the material he found on them will be published shortly (in Small and Small, forthcoming), along with the discoveries made by Hugh Chapman and Dennis Aldridge in other surveys carried out in the vicinity of Gravina in the late

14

Small: 2. The Topographical Context of Vagnari in the Roman Period

tributary, the Canale Capo d’Aqua. The terrain here was well watered and there was a string of small settlements where the inhabitants were able to exploit the arable land and at the same time had access to rough pasture for their flocks on the plateau of the Murge. Further to the east beyond Gravina, the level of the plateau dropped, allowing the Via Appia to ascend it with little difficulty, and here there were other rows of sites clustered along the river valleys, which included some of the richest settlements in the whole area of the map. Three of them could be counted as villas in an area where there were few obviously rich sites. To the south of our survey area there was a string of sites on the Lucanian side of the Basentello below Monte Irsi. Two of them were probably villas, and there was another villa of this period on the hill-top, partially excavated by Edith Wightman in 1971–2 (Small 1977: 57–95).

which as we now know was redeveloped by the emperor as a productive centre for a much wider area. It is a topic fully explored in this book. Because there was no municipality anywhere within easy reach, there are no inscriptions recording municipal honours and benefactions and there was no urban culture to stimulate reading and writing. It might be thought that the rural population had no need of the epigraphic habit. There were, therefore, in Mommsen’s day, no inscriptions to throw light on the nature of Roman occupation of the territory. But it was wrong, as we have seen, to draw the conclusion that the area was largely depopulated. Moreover, the epigraphic vacuum is no longer as total as it was when CIL IX was published in 1883. Marina Silvestrini (2002: 121–24) has published a funerary inscription found in the Masseria Macchitella (MM on the map Figure 2.1) which was dedicated by one Antonius Fortunatus to the Di Manes of his wife (coniunx), Savonia Nevia, in the late 2nd or 3rd century AD. The inscription was not found in situ, but on the assumption that it was not brought from afar to decorate the farmyard, we can assume that the couple lived in, and probably owned, one of the properties somewhere in the vicinity.

But even these villas may have been no more luxurious than the villa on San Felice, mentioned above, which was a relatively modest structure. To judge by the quality of the artefacts found in the surface surveys, there were no really rich villas anywhere in the area short of La Santissima, situated a little to the northwest of Spinazzola near the watershed of the Basentello river, which probably fell within the territory of Venusia (Figure 2.1). There are the remains there of a sumptuous villa and vicus (Marchi 2005: 188–99; Marchi 2010: 232–38). That apart, there is nothing in the area of our map to compare with the much grander villas which have been excavated at the Masseria Ciccotti in the upper Bradano valley (Gualtieri 2008), at Mola di Bari on the Adriatic (Ciancio 2002: xiii–xiv, 1–26) and at Faragola on the edge of the Tavoliere (Volpe and Turchiano 2009). We know from brick-stamps that some of the landowners in our area were among the most powerful people in the land (Small 2014b), not to mention the emperor himself, but they did not live here. If they visited their properties at all, they are unlikely to have stayed for long.

There are, moreover, two fragments of Roman inscriptions which were found in an excavation context in the territory, and both were discovered at Vagnari: two small fragments of marble inscriptions which had been thrown into the cistern in the southwest quadrant of the imperial vicus when it was filled in at the end of the 4th century AD (Small 2011c: 429–430). Both are likely to have been funerary, one of the late 1st century BC or early 1st century AD, the other of the 2nd century AD or perhaps a little later. They give us a tantalising glimpse of a Vagnari very different from that of the humble population who were buried in the graves being excavated by Tracy Prowse and her team. These were individuals who could afford tombs with inscriptions cut in marble. If I had ample years and funds and could continue to enjoy the kind support of the landlord and the Superintendency, I would empty the whole of the cistern in the hope of finding more fragments of inscriptions. Who knows? They might even give us the Roman name of the settlement.

There were no cities anywhere in the vicinity to detain them. The nearest urban centre of importance was the colonia of Venusia situated c. 45km from Vagnari along the line of the Via Appia, too far to be an economic or social centre for the area. Those functions must have been taken over by the vici, and especially by Vagnari,

15

CHAPTER 3

The Landscape Context of Vagnari, Past and Present Peter Wigand

Figure 3.1 View of the upper Basentello River valley from Irsina. The river flows from north to south (left to right) along the base of the flat-topped plateau in the distance, which was originally the early Pleistocene shallow sea bottom. The valleys have been cut through the Pliocene marine marls, and then filled with erosional sediments from the surrounding hillsides. In the far distance on the left, the face of the 200-m high fault scarp that lies to the east behind the community of Spinazzola can be seen. Photo P. Wigand.

only fields, but hedgerow vegetation, riparian habitats, and steep slopes. Unfortunately, landscape conditions during the summer of 2018 were exactly those needed for both surface erosion and massive slope failures. Indeed, after the wildfires that raged over southern Italy and other regions around the Mediterranean during the summer of 2021, the stage is set in 2022 for a summer of not only additional fires, but a cycle of perhaps unparalleled, large-scale erosion. In the past, similar cycles of climate also affected the landscape and peoples of this region, and as human land use intensified the effects of climate change were magnified.

Introduction Since the early 2010s, Italy has been ravaged by episodic torrential rains causing major flooding that has both destroyed property and taken life. Large areas of farmland have experienced devastating topsoil erosion and destruction of infrastructure including roads, drainage canals, and bridges. In the Mezzogiorno, — southern Italy—, this engenders a heavy economic cost for a region already suffering decades of fiscal decline. In 2017 Italy suffered the driest spring and summer for sixty years with up to 80 percent less rainfall than the historic average (Heggie 2020, 2020a). This water shortage was compounded by a summer heatwave with temperatures soaring above 104 degrees Fahrenheit (40 degrees Celsius). The summer of 2019 was equally hot and dry (Heggie 2020a), despite the fact that the period was characterised by almost daily thunderstorms. In 2017, rampant wildfires, some triggered by untended field burning, and others intentionally set, burned not

The landscape has been subjected to such amazing changes recently, that the question can be posed: Is this unique, or has this region experienced such radical changes in the past as well, and what have those been? Beyond the question regarding the role of climate change and other natural phenomena (earthquakes and landslides, for example), what has been the role 16

Wigand: 3. The Landscape Context of Vagnari, Past and Present

of people and their activities upon the landscape as well? Fortuitously, the region around Vagnari has a wealth of information regarding landscape dynamics spanning millions of years. During the late Pleistocene and early Holocene, ‘environmental processes’ shaped the landscapes of the Mezzogiorno; climate was the primary factor in landscape dynamics. However, as human populations increased after the appearance of Neolithic peoples 8,000 years ago, and their technology became more pervasive, their shaping of the landscape became increasingly important. When did human culture became a driving factor in landscape change? Has that impact varied? What past relationships between climate, landscape and people can be used in understanding and dealing with current and future landscape dynamics in southern Italy? Geological Setting Pre-Quaternary Today the region around San Felice is characterised by flat-topped plateaus, some rolling hills, and deep valleys dissected by streams that have cut several hundred metres into what was once a large plateau (Figure 3.1). In order to understand the current topography, one needs to delve deeply into the geological past and trace the record of its formation. It is a record that stretches back into the Mesozoic, when the region was still a shallow sea. Geologically, the deposits from this period are part of what is known as the Adriatic-Bradanic Foredeep (Vezzani et al. 2010). The coastal areas and shallow seas of the Late Jurassic through the Cretaceous of this period are dated, in part, by diagnostic sauropod footprints that are being found in increasing numbers throughout the region. The limestones upon which the towns of Gravina and Matera lie are Mesozoic (Cretaceous) in age and contain sauropod footprints as well (Figures 3.2–3).

Figure 3.2 Mesozoic limestone on the west side of the ravine at Gravina in Puglia. Photo P. Wigand.

Cenozoic limestones outcrop in some areas, but they are discontinuous, with both Pliocene and Oligocene units being absent. This absence reflects the initiation of, and the continuing massive tectonics linked to, the Apennine orogeny (Figure 3.4). A massive fold and thrust belt related to this event lies sandwiched between the eastern foothills of the Apennines to the west and the Gargano and the Apulian Carbonate Platform on the east. It trends from northwest to southeast and has been heavily faulted by the ongoing activity in the region. It was during this period that the Bradanic Foredeep formed and began filling with limestones and then from the middle Miocene onward with marine marls (Figure 3.5). The Basentello River valley and the Vagnari site lie close to the western edge of the Foredeep. The tectonic activity at this location

Figure 3.3 Sauropod foot print in the Gravina Mesozoic limestone, which has been used by Conti et al. 2005 to determine the age and geographic setting of the limestone deposits. Photo P. Wigand.

17

The Making of a Roman Imperial Estate

Figure 3.4 Geological overlay of the deep and surface structure of southern Italy (after Sauer et al. 2010: fig. 1).

has affected the local topography and created a complex mosaic of landforms and deposits around Vagnari. The deposits within the Bradanic Foredeep have been heavily faulted and then overlain by upper Cenozoic marine deposits, primarily marls of upper Pliocene age. These marine marls form the ‘bedrock’ that underlies most of the upper Basentello River valley (Figures 3.5– 6). It is the semi-indurated ‘bedrock’ that underlies the Vagnari site itself (Figure 3.7). Quaternary Early through middle Quaternary geology and stream formation — regional The marine marls were in turn overlain by coastal deposits, comprised primarily of weakly cemented conglomerates and beach sands of lower Pleistocene age. These reflect the ongoing erosion of the Apennine Mountain range as it was being uplifted during its eastnortheastward movement across southern Italy during the late Cenozoic and early Pleistocene. This process can be visualised as if the Apennines were a large bulldozer blade that was both shedding alluvial deposits from its growing mountain mass, and at the same time pushing and compressing these and older deposits before it as it advanced eastward. The distortion front of the southern Italian landscape can best be seen in the difference in the shape and orientation of river drainage patterns from west to east. On the west, in the region being most affected by Apennine disturbance, the river drainages are dendritic or tree-shaped with a main trunk downstream and many twisting tributaries upstream. East of the area of disturbance the stream drainage pattern is linear with relatively straight main channels running to the

Figure 3.5 General stratigraphic column of the Apulian Platform unit succession and their ages (after Petrullo et al. 2017: fig. 2).

18

Wigand: 3. The Landscape Context of Vagnari, Past and Present

southeast and relatively straight tributary channels aligned in the same direction. Near the coast, Pliocene deposits have been buried by a succession of upper Pleistocene and Holocene terrestrial and marine deposits including recent river deposits, coastal plain alluvium, and beach and coastal dune deposits. In the interior, these deposits have been uplifted and incised by streams and small rivers. The valleys, in turn, have been filled with Quaternary terrestrial, and occasionally marine, deposits laid down during episodes of interglacial marine coast transgression events. As a result, there is a complex mosaic of bits and pieces of ancient stream channels exposed in the river valleys of the region at various locations and elevations. The problem is that one cannot use elevation as the only criterion to determine age. One must also use the geological context of the streams to determine their sequence. For example, exposures of early Quaternary stream channels do occur higher in the river valleys, but exposures of later Quaternary stream channels may occur at middle to lower elevations, and not necessarily at progressively lower elevations. The first driver of stream channel development has been the tectonic uplift of the region above sea level since the end of the Pliocene. Tectonic uplift led to a succession of longterm stream incision events that resulted in relics of previous channels being left at successively lower elevations during the Quaternary. Therefore, evidence of the earliest Quaternary stream channels is found in the upper slopes of the modern river valleys at higher elevations, whereas remnants of later Quaternary stream channels are found exposed at lower elevations in the river valleys. However, there is an additional

a

b Figure 3.6a–b Marine marl exposures in the Bradano River valley west-southwest of Irsina. Photos P. Wigand.

19

The Making of a Roman Imperial Estate

Figure 3.7 The Apennine Blue Clay also outcrops in the erosional channels of streams around Vagnari, as it does here just north of the archaeological site (Google Earth Image).

Terrace No.

T-0

T-1

T-2

T-3

T-4

T-5

T-6

T-7

T-8

T-9

T-10

MIS

1

5.1

5.3

5.5

7

9

11

13

15

17

19

Age BP

7

80

100

120

194

330

405

500

575

670

730

(years ka) Table 3.1 Middle to Late Quaternary Marine Terrace ages before the present of coastal terraces in the Metaponto area and their Marine Isotope Stage (MIS) assignment (after Sauer et al. 2010; Brückner 1980).

factor that complicates unravelling the stream history of the region, the cycle of glacial vs. interglacial sea level change, especially during the last 800,000 years.

the formation of regional drainage patterns. The gradual rise of the region on the north-northwestern edge and the growing downward tilt of the landscape in a north-northwest to south-southeast direction not only resulted in an orientation of streams in that direction, but also in a rectangular system of faults and joints which the streams followed. Although initially these streams followed shallow channels, they began forming deeper channels as the uplift of the region continues. West of the Basentello River valley, dendritic stream drainages developed in response to the deformed landscape created by the eastward expansion of the region affected by the Apennine orogeny.

Stream incision (cutting) occurred during glacial episodes when sea levels dropped (regression) significantly as global ocean water was tied up in glacial ice. Stream alluviation (filling) occurred during Interglacial periods when global ice sheets melted and sea levels rose to flood coastal areas. This has made unravelling the south Italian landscape a complex undertaking. The formation of at least 11 marine terraces spanning the last 800,000 years mark former interglacial coastlines on the southern Italian coastline (Table 3.1; Figures 3.8 and 3.9). Many, if not most, of the glacial coastlines have been flooded and now lie under the waters of the Gulf of Taranto.

Therefore, three evolutionary stages of stream development can be recognised in the central and eastern sections of southern Italy (Beneduce et al. 2004). Today, the first stage of stream formation is evidenced by relics of the flat erosional landscape at the top of the plateaus that emerged from the shallow sea that characterised the region 2 million years ago. The

Thus, it has been regional tectonic uplift in combination with the subsequent downcutting of streams during glacial events when sea levels dropped that have driven 20

Wigand: 3. The Landscape Context of Vagnari, Past and Present

Figure 3.8 Middle to late Pleistocene coastal terraces mapped and described along the Gulf of Taranto near Metaponto (after Sauer et al. 2010: fig. 2).

Figure 3.9 Comparison of coastal terrace ages in the Metaponto area (red dots in top diagramme) with the Marine O18 Isotope Stages or the record of O18 content of ocean sediment cores in the lower diagram. The lower diagramme is derived from Railsback et al. 2015: fig. 3. Working backwards from the present, which is MIS 1 in the scale, stages with even numbers have high levels of oxygen-18 isotope and represent cold glacial periods, while the odd-numbered stages with low levels of oxygen-18 represent the warm interglacial periods. High points mark decreased O18 isotope levels in ocean sediments (interglacials) when global temperatures were high enough for both light and heavy isotope water to evaporate from the oceans. Troughs between the peaks denote periods when O18 levels were higher in the ocean reflecting periods when primarily lighter isotopic water was evaporating from the oceans because global temperatures were only high enough to evaporate lighter isotopic water. Movement of O18 levels toward the orange line indicates increasingly warmer interglacial conditions, whereas movement of O18 levels downward toward the blue line indicates increasingly glacial conditions. The red dots in the upper diagram correspond with the more extreme interglacials. The trend line in the upper diagram follows the long-term tectonic rise of the southern Italian coastline during the last 800,000 years. When the Altamura Neanderthal lived ~135,000 years ago during MIS 5e times, southern Italy was roughly between 60 and 80m lower than it is now.

21

The Making of a Roman Imperial Estate level. This major fault orientation is mirrored in the course directions of many of the streams of the region. In addition, the courses of some minor streams are occasionally controlled by open synclines and gentle curves around the fracture zones. This description evidences the complex interplay between tectonics and the impact of glacial versus interglacial sea levels during the Quaternary.

Figure 3.10 Along a stream cut just north of the Vagnari site, the grey-coloured upper Pliocene-lower Pleistocene marine marl, which forms the bedrock of the area, is exposed. It is characterized by a blocky structure and is clearly thinly laminated. In this area, the top of the deposit has been eroded so it is uncertain how much of the original marl is missing. This formation is the Sub-Apennine Blue Clay. Photo P. Wigand.

second stage is represented by wide, gently inclined slopes adjacent to the foothills, often covered by coarse sediments. During the third stage, a series of marine terraces formed and the current drainage system developed creating both broad, flat valleys and welldefined channels and gorges. The more recent stage three streams are superimposed over and through relics of the previous stages, and are deeply incised into the Cretaceous limestone. These drainages are clearly responding to the tectonic uplift of the region. On the other hand, low-order streams and segments of some of the major rivers appear to be structurally controlled, for example by the effects of actual faulting. The transection of the region by many northwest to southeast trending faults has affected stream drainage orientation. About 21.5km northwest of the Vagnari site, a major thrust fault lies just northeast of the town of Spinazzola (Figure 3.1). Along that fault line, the local Mesozoic limestone, which underlies the AdriaticBradanic Foredeep deposits, is uplifted over 500m above sea level. South-southeastward from Spinazzola, this fault dips toward the Gulf of Taranto, until at Metaponto (ancient Metapontum) it plunges below sea

Figure 3.11 Cross-bedded conglomerate formations such as these along the road leading south out of Irsina seem to be part of an upper Pliocene to lower Pleistocene sequence of fan-delta-front to shallow marine conglomerates and sandstones. In places they grade laterally into lagoonal mudstones and diatomitic clays, or in other places they grade laterally into shell containing sandstones and siliciclastic calcarenites. The gravels are well-rounded and often imbricated (dipping downward) upstream into the direction of the water flow. These conglomerates lie at 525m above sea level today. Photo P. Wigand.

22

Wigand: 3. The Landscape Context of Vagnari, Past and Present

a

b

c

d Figure 3.12a–d Beach deposits exposed at the Bosco locality south of Gravina in Puglia (a-b). The indurated units can be seen as strong linear shelves in the deposits, lying at 432 m above sea level. Shells (c-d) were recovered from these deposits as evidence of their coastal origin. Photos P. Wigand.

Early through late Quaternary geology and stream formation — local

sides of the valleys, is a blue-grey, sandy or silty clay with chalk and other carbonate inclusions (Figure 3.10). These are what are called the Sub-Apennine clays (Figure 3.3). The Sub-Apennine blue clay formation is located at different sites within the Foredeep where clays were deposited in the early Pleistocene. All of these clays have been subject to diagenesis (including all the chemical, physical, and biological changes undergone by a sediment after its initial deposition, and during and after its lithification, exclusive of surficial alteration or weathering and metamorphism), unloading due to erosion, and weathering of the top strata. However, the clays have undergone different levels of unloading (deformation due to the release of the weight of overlying sediments) and have developed different levels of structure, due to their different locations in the basin. They have inclusions because they are the result of the erosion of nearby terrestrial or coastal deposits. This formation is exposed up to about three quarters of the way up the sides of the incised valleys (Campbell et al. 2011).

The entire upper Pliocene and lower Pleistocene geological sequence in the Basentello River valley covers a spectrum of marine and coastal depositional environments from deep water marine through shallow water, to shoreline (beach), and coastal dune deposits. In addition, there is a complex array of terrestrial, fluvial, and colluvial deposits. The sequence has been subjected to successive large-scale tectonic uplifts from the early to late Quaternary. In the lower 30km of the Bradano valley, Brückner (1982) identified eleven uplifted surfaces at elevations up to 370.25m above present sea level, and dating between 636,000 to 80,000 years ago, and coastal terraces resulting from marine transgressions (at least ten in the last 800,000 years) have also been identified; furthermore, a study at Alimini Piccolo south of Bari identifies two sea level rises (Primavera et al. 2011). In the immediate area of the Vagnari site, the local bedrock geology (Pieri et al. 1968; Boenzi et al. 2008) consists of three Pleistocene-age formations dating c. 2 million years before present. The two lower formations are marine (Calabrian), while the upper formation is a continental (Villafrancian) deposit. The lowest Calabrian formation, which outcrops on the floor and

Above the Sub-Apennine blue clay formation lies a formation of yellowish, calcareous, and quartzose sands intercalated with lenses of fine gravels and calcarenites. These deposits, which form most of the upper, steeper slopes of the valleys, are capped by a stratified, polygenetic conglomerate formation of carbonate and 23

The Making of a Roman Imperial Estate Overlying the conglomerates, and in some areas contemporaneous with them, are coastal sand deposits, such as those exposed at the top of the plateau in the Bosco Difesa Grande di Gravina, the woodland about five kilometres south of Gravina in Puglia. Some are beach deposits often containing various species of shells and tube worm cases, among other things, indicating that they are beaches (Figure 3.12a–d). However, there are also coastal dune deposits that are characterised by the cross-bedding typical of Aeolian dunes. Because these units have been subaerially exposed or have lain near the surface for much of the Quaternary, they have been strongly oxidised and are characterised by a bright reddish colour. They also contain cemented layers that have been deposited later. The cementing materials include iron and carbonate. The cementing material was carried in underground water and probably accumulated at boundaries between sand strata where the grain size changed, but they could also have been deposited at times when the top of the water table lay at that level. These sand units were deposited primarily during the terminal Pliocene, but some of the coastal dunes may be of early Pleistocene or Villafrancian age. There are extensive sand units exposed near the top of the plateaus just above the San Felice villa site and along the road as it drops down to that site (Figure 3.13). However, these sands are clearly stream deposits. They are characterised by sequences of fluvial cross bedding. They contain filled channel incisions and in some places strong carbonate units that may have been deposited in ancient soils or post-depositionally after their burial by later fluvial deposts (Figure 3.14a–b). In addition, they are not oxidised as the Bosco Difesa Grande di Gravina deposits are. Although these sands probably originated from the upper Pliocene coastal sands, they have lost their oxided coating while being reworked in streams, and they contain no marine shells. They are of a younger age as well and lie at an elevation of c. 483m, about 52m below the elevation of the conglomerate deposits across the valley at Castello di Monteserico and 42m below the conglomerates slightly further south at Irsina. However, they do not contain gravels from the conglomerates. This suggests one of two possibilites: either there were no nearby conglomerates, or the energy of the stream was not strong enough to carry the large conglomerate gravels. The medium to coarse sized sands indicate a stream velocity of only about ~ 1 to 50cm/sec. The low velocity stream was condititoned by the low gradient, i.e. the elevation of the entire stream drainage was close to sea level so stream velocities would have been slow. Streams such as these may be assigned to a period when low velocity, probably meandering stream systems, may have characterised the newly emerging landscape of

Figure 3.13 Fluvial sand deposits at the top of the slope above the San Felice villa site. In the lower third of the exposure, the cross-bedding of the fluvial sands is clear. It is overlain by horizontally bedded sands which appear to have been truncated and with a carbonate accumulation along the truncation surface. Above these are two units of cross-bedded fluvial sands, and a mixed assemblage of sands, carbonate units topped by a much later soil. The elevation of these units is around 484m above sea level. Photo P. Wigand.

siliceous gravels. It is frequently cross-bedded with lenses of brownish to reddish sand. The conglomeritic layer covers most of the plateau surfaces and, where it outcrops along the plateau rim, it often forms near vertical slopes such as can be seen on the plateau upon which the town of Irsina sits across the Basentello River valley south-southwest of Vagnari (Figure 3.11). Landslides have moved large slabs of this formation during periods of wetter climate. A major period was about 35,000 years ago, during a glacial interstade, but there is evidence near the Vagnari village site that such a landslide occurred during the Holocene as well.

24

Wigand: 3. The Landscape Context of Vagnari, Past and Present

with accumulations of indurated calcium carbonate between the laminated units. Identification of these as possible mud volcano deposits is based upon the author’s visit to several active mud volcanoes in northeastern Iran, west of the city of Gorgan, Golestan Province, in 2018. Investigations of the geomorphology of the largest of these, the GarniarigTapeh, indicate that these mud volcanoes have been formed during Quaternary as a shield-shaped hill or pool on the flat coastal plain southeast of the Caspian Sea (Arian and Sistanipour 2015). Mud volcanoes result from the rising of lighter materials that they have been triggered by buoyant driving mechanisms. Fluid for mud volcanism is supplied from various sources, including meteoric and volcanic waters, pore water expulsion, hot springs, mineral dehydration reactions, and gas hydrate destabilisation. Most mud volcanoes occur along convergent plate margins where fluidrich sediment has accumulated in deep-sea trenches at high rates. Such deposits then enter the subduction

a

b Figure 3.14a–b Fluvial sands on the slope above the San Felice villa site. In these two photos (a-b), relatively horizontally bedded fluvial sands have been cut by a channel that has been filled by cross-bedded fluvial sands. This has been truncated and is overlain by a soil that has been the locus for postdepositional carbonate accumulation (b). Photos P. Wigand.

southern Italy during the early Quaternary. They most likely date to the Stage 1 period of stream formation described above. Their importance for the Vagnari area is that these deposits supplied most of the sands that comprised the soils around the Vagnari village site.

a

Along the road that leads from the Basentello River and the Castello at Monte Serico lies an extensive exposure of fluvial and spring deposits (Figure 3.15a–b). It lies at an elevation of 363m within sight of and c. 172m below the conglomerates on top of Castello di Monteserico hill. This exposure reveals a complex pattern of fluvial deposits comprised of typical cut-and-fill cross-bedded sands with occassional sandy gravel units sandwiched between them. The gravel units consist of the rounded cobbles common in the upper Pliocene conglomerates that cap the plateaus in this area (Figure 3.15a–b). Although the cross-bedded sand units predominate, the presence of sandy gravel units with a sandy matrix suggests flood events, such as those that appear in the Holocene deposits from near the Vagnari village site that will be described below. The presence of both large gravels and sands in the same units suggests turbulent flow. Typical gravel bars would have little or no sand in them (Figure 3.16). The deposits overlying the fluvial deposits are complex (Figures 3.17–19). They are comprised of sequences of thinly laminated clayey silts

b

Figure 3.15a–b Stream and spring deposits northeast of Castello di Monteserico. These lie across the Basentello River valley from the Vagnari village site. Photo (a) displays the diversity of the fluvial deposits which were dominated by fluvial cut-and-fill channels. The fluvial deposits are overlain by what appear to be spring deposits or mud volcano deposits (b) with layers of indurated carbonate between the various depositional episodes. The whole sequence is blanketed by a large carbonate cap, and, above an unconformity, this is covered by the modern soil. Photos P. Wigand.

25

The Making of a Roman Imperial Estate

Figure 3.16 Here the fluvial unit consists of a chaotic slurry of sands, silts, and gravels. These are suggestive of flood deposits and are very similar to flood units that are found exposed in the sides of recent streams. Photo P. Wigand.

Figure 3.17 The deposits overlying the fluvial deposits are comprised of sequences of thinly laminated clayey silts with accumulations of indurated calcium carbonate between the laminated units. Photo P. Wigand.

Figure 3.18 The deposits overlying the fluvial deposits are comprised of sequences of thinly laminated clayey silts with accumulations of indurated calcium carbonate between the laminated units. Photo P. Wigand.

Figure 3.19 Detail of the deposits overlying the fluvial deposits comprised of sequences of thinly laminated clayey silts with accumulations of indurated calcium carbonate between the laminated units. Photo P. Wigand.

zone, where liquids and volatiles are released due to increasing compaction stress and temperature. The high sediment accumulation rate at passive continental margins and the lateral tectonic compression at active continental margins are probably the main reasons for spatial distribution of mud volcanoes. In addition, it is obvious that plastic clay layers must be presented for mud volcanoes to form. The Basentello Valley lies on the margin between the African and European plates. At times in the past this area has been very active when periods of uplift were occurring. The mud volcanoes are a result of this process.

the upper Basentello River valley. Using this diagramme, the ages of the fluvial deposits lying just upslope of the San Felice villa site, as well as the Monteserico mud volcano site, can be roughly extrapolated with respect to the general rise of the southern Italian landscape during the last 800,000 years (Figure 3.20). Based on Figure 3.20, the current elevation of the fluvial sand exposure above the San Felice villa site indicates that it is about 500,000 years old, and the Monteserico mud volcano site at 363 m is just over 200,000 years ago. This would place it in the Penultimate or Riss Glacial episode, and coincidently with the appearance of Neanderthal peoples in Italy (Marra et al. 2015).

The approximate age of the fluvial deposits overlying the San Felice villa site and the Monteserico mud volcano site can be roughly extrapolated from the general rise of the southern Italian landscape during the last 800,000 years (Figure 3.20). The approximate age of the various fluvial deposits that are observed in the upper Basentello River basin can be extrapolated by correlating their elevation with the corresponding age in Figure 3.20. These ages are approximate estimates, but they are based upon the rate of rise of southern Italy during the last million years and the current gradient of the Basentello River near fluvial exposures studied in

It is clear that the sediments that comprise the geology of the Vagnari village site area are a complex mixture of coastal sands, marine clays, and coastal and deltaic conglomerates that have been eroded from the surrounding plateaus. These all contribute to the formation of an intricate mosaic of surface and subsurface geology that effect both hydrology and the edaphic characteristics of the soil. Tectonically, each succeeding uplift resulted in a lowering of the base level of the drainage systems of 26

Wigand: 3. The Landscape Context of Vagnari, Past and Present

Figure 3.20 Approximate elevation of the Upper Basentello river near Vagnari above mean sea level during the last 800,000 years. This provides a rough estimate of the age of fluvial exposures found in the area. The lower line indicates the regional rise of southern Italy at the coast with each of the major coastal interglacial terraces marked (T-?). The upper line plots the relative elevation of the upper Basentello River in the Vagnari area for the same period.

upstream at 2 to 4 metres per year now. It reveals no rotation of the sediments as would be expected in a massive landslide. That is, if a significant landslide had occurred one would expect to see the downslope section of the slump rotated upward, and the upslope section of the slump rotated downwards.

the region. These triggered stream incision events that were initiated at the margin of the sea and progressively moved upstream. The rate of movement of incision upward and downward was determined in part by the geology of the sediments that the streams were cutting through. However, it was also conditioned by the nature of the uplift. The faster and higher the uplift, the faster the rate of both down cutting and head-ward movement of stream incision.

It is clear that the Vagnari terrace was not formed by a single land slide, but is an erosional terrace that matches others that can be traced across and up and down the valley at about the same elevation. In addition, the conglomerates that Campbell et al. identified as the capstone of the landslide, occur in only scattered areas south of and above the Vagnari site and towards San Felice. There is clear evidence of landslides around the San Felice site, but not at Vagnari. There are sections of conglomerates in areas south of the Vagnari terrace, but they are related to a series of other landslides that have been occurring during the last 35,000 years. Many of the conglomerate slabs that are seen in the Basentello and adjacent river valleys were detached from the cap rock as they were undermined by erosion beginning about 35,000 years ago. Boenzi et al. (2008) and Piccarreta et al. (2011) record many such slippages south of the Vagnari area in the record, especially after 35,000 years ago. This includes the one that may have destroyed the San Felice site.

Campbell et al. (2011) suggest that the mixed (nearly intact units of sands, gravels and clays) lithology that characterises much of the surface and near surface deposits of the area reflects numerous landslides that have whittled away at the plateau rim deposits and carried them well down-slope. Campbell et al. (2011) describe an extremely large landslide at the Vagnari site, that may have run over 1.5km from its detachment point and had a minimum depth in excess of 15m. They suggest that this large event was responsible for the creation of the entire Vagnari plateau. Since Campbell studied the Vagnari terrace 20 years ago, there has been a tremendous amount of head-ward erosion of the streams incising the terrace, as well as significant surface erosion, which has exposed underlying last interglacial soils. This has provided sediments and topography that lay hidden during the investigations of Campbell. In particular, the sediment record exposed in the stream-cut next to the primary east-west road north of the Vagnari vicus site. This channel is cutting

It is clear that colluvial slope wash and cycles of stream incision and deposition probably have been 27

The Making of a Roman Imperial Estate

a

b Figure 3.21a–b Just north of Vagnari vicus, recent erosion along the stream incision has exposed the cross-sections of two streams of different ages and with different fill. (a) An older channel cut-and-fill cross-section, (b) on the lower right with the highly oxidized sediment, has been buried by later buff-coloured colluvium, which has then had a more recent stream incision on the upper left of the exposure cut into it. Photos P. Wigand.

units, indicating a significant time duration. There are some units with more chaotic mix of gravels, sands, and silts which suggests a debris flow deposit that had eroded from the conglomerate caprock at the top of the plateau (Figure 3.22). Beneath the oxidised fill is a buff-coloured unit also comprised of debris flow material of sandy-silts and gravels. There is a unit of carbonate encrusted gravels between the two units which was deposited by carbonate containing ground water as it encountered the change in sediment grain size between the two units. The buff-coloured fluvial unit lies unconformably over Apennine blue clay into which the stream channel was cut (Figure 3.23). It is clear that the debris flow followed an original stream channel running perpendicular to the current channel.

Figure 3.22 The lower channel with oxidized Pliocene fill from the top of the plateau. The channel fill appears to lack the more uniform mixing that is seen in the younger deposit to the left of this channel in the exposure. Photo P. Wigand.

The stream cut-and-fill sequence just above and to the left of this channel is also filled with both stratified and debris flow deposits, but their source is not from the Pliocene deposits at the top of the plateau (Figure 3.24). Because the channel has been cut into sediments that overlie the adjacent channel filled with oxidised Pliocene sediments, it must be significantly younger, because it lies much closer to the modern surface of the terrace. It is older than the current stream channel, so must be Upper Pleistocene in age between 130,000 and 12,000 years old. This debris flow is also following an older stream channel which is orientated at right angles to the current stream cut.

the dominant processes in sediment movement in the area. This is borne out by examination of a stream cutand-fill section, the Vagnari north exposure, along the road just 371 m north of the Vagnari vicus site (Figure 3.21a–b). The stream channels which are exposed in the large channel cut were not visible to Campbell, but they lie in intact sediments. Both channels are cut into horizontally deposited older units. There are no blocks of rotated conglomerates in this area of the Vagnari terrace. The oldest channel, which is minimally of last interglacial age estimated from its elevation, contains oxidised stream gravels, and silts and sands derived from the conglomerate cap rock, probably the one lying along the rims of the plateau surrounding the Vagnari terrace. A younger channel which is filled with buffcoloured gravels, sands, and silts is of interstadial age. On the lower right, the older channel contains multiple units of stratified reddish gravels, and silt channel fill

The Monteserico and Vagnari exposures are at about the same elevation and fix the elevation of the floor of the Basentello River valley to about 130,000 years ago. It reflects the results of stream down-cutting resulting from both regional uplift and lower sea levels during the Riss (Penultimate) Glaciation. Rising sea levels during the Eemian Interglacial would have resulted in deposition of sediments in the valley of the Basentello 28

Wigand: 3. The Landscape Context of Vagnari, Past and Present

again. Lowered sea levels during the succeeding Würm Glaciation, combined with ongoing tectonic uplift, would have initiated renewed stream incision. This resulted in the establishment of the Basentello River valley at roughly its current elevation, leaving the areas where the Monteserico and Vagnari north exposures lie on terraces about 100m above the new valley floor. The Vagnari village site and the associated cemetery also lie on the surface of this Riss/Eemian terrace or what was once the floor of the Basentello River valley. Last glacial and Holocene climate and human impact upon landscape erosion and stream incision and fill cycles The longest most continuous sedimentary record that has been found thus far in the region is the one in a tributary of the Basentello River that originates from the north slope of the hill upon which Irsina is located (Figure 3.25). The exposure is over 1.5km long and in places over 9m high. Based upon correlations with three other exposures in the area that have been radiocarbon-dated, over 20,000 years of sedimentary depositional record are exposed. The sediments are comprised of laminated fluvial sands, silts, and clays. In addition, at least three major soils have developed in this sequence and provide excellent correlation with other fluvial exposures in the upper Basentello River valley.

Figure 3.23 The contact between the two channel fills of the 130,000-year-old channel. The indurated carbonates accumulated after deposition of the oxidized debris flow-flood deposit. The fill of this channel appears more chaotic than the fill of the later channel to the left. The channel lies unconformably over the Apennine blue clay. Photo P. Wigand.

The lowest soil has formed on the surface of a channel that was cut when sea level was lowered during a glacial stade and the Basentello River was incising its channel. Soil formation occurs when vegetation cover has stabilised the surface of the ground. It suggests that there was sufficient rainfall to accomplish this. As sea levels rose again during the late glacial, the channel was filled with sands and silts eroding off the surrounding slopes. Just below the 8400-year-old soil lie pond sediments containing snails, suggesting that the stream was sluggish and that it had pools with aquatic vegetation. This is found in other locals across the valley near the Vagnari village site. The 8400-year-old soil formed during the warmest portion of the Holocene. It indicates that the landscape was well vegetated and stable allowing the formation of another soil. Between the 8400-year soil and the 2100-year soil, sediment accumulation suggests that vegetation cover may have been less dense, allowing slope erosion to increase. The 2100-year-old soil is not only found in several stream exposures near the Vagnari village site, but also occurs in the Fossa Bradanica as well (Boenzi et al. 2008). Boenzi et al (2008) also record the beginning of wetter climate at 3700 cal. BP. A radiocarbon date taken from the base of a long episode of spring discharge from Baron Spring

Figure 3.24 Younger stream cut-and-fill in the stream channel just north of the Vagnari vicus site. It is also filled with debris flow sediments comprised of silts, sands, and small to medium sized gravels which are more uniformly mixed. Photo P. Wigand.

29

The Making of a Roman Imperial Estate

Figure 3.25 The Irsina exposure east of the town (on the crest of the hill in the distance). The radiocarbon ages of the three major soils are marked, as is the profile of the lowest soil which follows the surface of an ancient stream valley. These deposits overlie Apennine Blue Clay, which is exposed in the stream seen at the base of this exposure. This exposure lies at an elevation 20m above that of the Basentello river near Vagnari, but 70m below both the Vagnari North and Monteserico exposures. Photo P. Wigand.

east of the Basentello River just south of the Vagnari village site confirms this date, which is just after the appearance of Bronze Age culture in the area.

geomorphology along a 7km stretch of the Basento River just southwest of the town of Pomarico. Their study of the Fosso La Capriola sediments revealed Holocene cycles of erosion and deposition (Boenzi et al. 2008: fig. 7). They record a phase of deposition occurring in the Late Neolithic that was followed by a down-cutting phase. A second phase of sediment accumulation between 2800 and 2500 cal. BP coincides with the period of early Greek settlement. They suggest that this occurred in response to a cool-wet period coinciding with a glacial advance in the Alps. After a period of stability at ~2500 cal. BP, a new episode of deposition began, continuing until ~1600 BP. This, they suggest, was probably due to Roman land-use practices. Renewed erosion between 1620 to 1435 cal. BP was again followed by sedimentation between about 1435 and 1000 cal. BP. Down-cutting after 1000 cal. BP created a new fluvial terrace (the first since the late Pleistocene), and also resulted in stream channel deepening. They correlate this episode with the Medieval warm period. Finally, they see the Little Ice Age as the cause for recurrent flooding beginning ~750 to 700 cal. BP and continuing until ~400–300 cal. BP.

Campbell et al. (2011) examined four stratigraphic pits in the ravine dividing the Vagnari archaeological site into two halves, the vicus on the north and the cemetery on the south. The c. 3m sequence of sediment accumulation in the drainage appeared to be comprised of poorly stratified colluvium derived from adjacent fields. Optical stimulated dates (OSD) taken from one of the pits provide a framework for the stratigraphic record (Campbell et al. 2011). Three stratigraphic units have been identified, Unit 1 extends from the present to about 480 years ago; Unit 2 from 480 to ~1900 years ago; and Unit 3 is older than 1900 years. Unit 2, however, may be missing as much as 500 years (Campbell et al. 2011). Therefore, the record at the Vagnari village site has a roughly 1700-year sediment record extending from the present to the end or middle of the Roman period. The break at about 480 years ago corresponds to the date at the top of the youngest soil from other exposures in the area, and the date of about 1900 years ago corresponds roughly to the date at the top of the second or 2100 cal. BP soil in the other fluvial exposures in the area.

The most rigorous study of the Holocene alluvial deposits of the region is that of Piccarreta et al. (2011). They assessed 34 of the 42 published and unpublished radiocarbon dates available for the Holocene alluvial deposits from Basilicata province. They identified eight

Another study conducted about 35km south of San Felice provides a better framework for geomorphic studies in the region. Boenzi et al. (2008) studied the 30

Wigand: 3. The Landscape Context of Vagnari, Past and Present

to cooler, moister phases actually are related to cycles of torrential rains corresponding to warmer, drier episodes when a more intensely Mediterranean climate was present in the region. We, as well as Boenzi et al. (2008) and Piccarreta et al. (2011), see that the alluvial record from the beginning of Greek settlement about 2,500 years ago through the Roman, Medieval, and up to the modern period, is significantly different from the earlier Holocene record. We all suggest that although it does reflect climatic effects, it is coupled with much greater human impact upon the landscape as well. The Holocene portion of the depositional record, which was recorded in the Irsina exposure, is repeated in the depositional records exposed in the fluvial records of both Arroyo Italiano I and Arroyo Italiano II on the east side of the Basentello River valley just south of the Vagnari village site. The Arroyo Italiano I exposure lies 1km southwest of the Vagnari village site at an elevation of 258m above sea level or 74m below the elevation of the Vagnari village site (Figure 3.26). It is in the next drainage to the south. Its elevation is also almost the same as that of the Irsina exposure, which lies 8.5km to the south-southeast. The Vagnari north stream exposure 371m north of the Vagnari village site is also filled with Holocene deposits that lie unconformably upon scoured Apennine blue clays (Figure 3.27a–b). It has the same soils as those found in the Arroyo Italiano I locality, but they are much thinner, and the flood units have much smaller gravels that those found at Arroyo Italiano I. There are only thin lenses of gravels at the bottom of each flood deposit. These deposits in the middle sections of the streams lead to the valley floor. The deposits at the Arroyo Italiano I locality reflect the increase in water velocity following the convergence of floods coming from several branches of the stream. The carrying capacity of the stream increased significantly so that larger cobbles could be carried within a chaotic mixture of sands, silts, and clays. Floods also eroded the sides of the stream channel and added local sediments as well. This explains the fact that the flood deposits in the lower elevation sites are filled with larger gravels, as well as a poorly sorted mixture of sands and silts.

Figure 3.26 Stratigraphy of the Arroyo Italiano exposure southwest of the Vagnari vicus site. The two lower soils appear in the Irsina exposure, but the top one has been disturbed there by ploughing. The soils are not as thick here as at the Irsina exposure, but there is a well-developed, carbonate rich B horizon (the buff-coloured unit) here. Beneath the 8400-year-old soil there are pond sediments with snail shells, just as at the Irsina exposure. There are also units of gravels at the base of the 2100 cal. BP A horizon and at the base of the 500 cal. BP A horizon. These indicate that there were massive floods that scoured these channels out and then soils were formed in the sandy silts at the top of each of the two flood sediment units.

episodes of increased alluvial activity for the region. Their findings suggest that many of these phases also coincided with flooding events in other regions of the Mediterranean. Boenzi et al. (2008) correlated these events to the Alpine and Apennine Holocene glacial record. They indicate that erosional episodes correlated with drier interstadial events. Both Boenzi et al. (2008) and Piccarreta et al. (2011) suggest that the cycle of deposition and erosion has been climatically driven, and that flooding has been associated with cooler, moister phases. Erosional episodes, on the other hand, they propose, occurred during warmer, drier phases when high magnitude, but lower frequency flooding occurred. However, our investigation of the relationship between the radiocarbon dated erosion events and a climate model for the region suggests that the relationship is more complex. We see that the episodes of flooding that they correlate

During the last six decades there has been a significant increase in stream incision in the upper Basentello River valley It has two causes: 1) the construction of a network of concrete-lined drainage canals meant to alleviate flooding (Figure 3.28a–b); and 2) a shift to increased torrential rains. The canals increased the velocity of the water, which allowed it to carry more sediment. This initiated a major cycle of head-ward stream erosion that is accelerating due to the increased intensity of summer rainfall. Storms during the last six years have resulted in head-ward cutting of 5–6m 31

The Making of a Roman Imperial Estate

a

b

Figure 3.27a–b Holocene record of fluvial deposits with superimposed soil horizons nested within the stream channel north of the Vagnari vicus site, with (a) a close up of the deposits, and (b) the well-developed angular-blocky structure of the soils. The flood units have a thin lens of gravels at the base, but they are much thinner than those seen at Arroyo Italiano I. Photos P. Wigand.

in the Vagnari north exposure (Figures 3.28c–d) and downward incision in the Arroyo Italiano site of over 2m per year (Figures 3.28e–f). Unlike most of the last 2,000 years, most of the sediment is now carried into the Gulf of Taranto, instead of being deposited in the valley floor. Sediments have clogged the modern concrete canals during lower volume rainfall episodes. In effect, the streams are attempting to heal themselves. Therefore, although the current stream channel between the Vagnari village site and the cemetery is relatively recent, there have been several cycles of cutting and filling during the late Holocene. During the last 2,000 years, incision of the channel began about 1,700 BP, and refilled after 1,400 BP. The current incision began after 900 BP, but it was accelerated by the activities of the last 70 years. This explains the dating problem that occurred when the channel next to the Vagnari village site was investigated previously.

Holocene landscape changes have been dominated by short-term climate impacts and human activity, rather than long-term glacial-interglacial sea level cycles or tectonic uplift. De Musso et al. (2020) noted both climate and human caused changes in the Basento River valley. To explore the relationship between late Quaternary climate, and erosion and deposition processes in the region, a physical climate model that generates estimates of precipitation and temperature for thousands of years is used (Bryson 1992; Bryson and DeWall 2007). The climate model for the last 12,500 years can be used to calculate the effective precipitation, and then compare it with the Langbein and Schumm (1958) model to predict a climate-based erosion rate for the Puglia and Basilicata region (Wigand and McCallum 2017: fig. 15). It is clear that the effective rainfall record (a combination of actual rainfall and temperature) 32

Wigand: 3. The Landscape Context of Vagnari, Past and Present

a

b

c

d

f

e

Figure 3.28a–f Flood control canals constructed during the 1950s to move rainwater quickly out of fields on the valley floor accelerated stream velocities to the point that they began head-ward erosion in streams above the upper ends of the canals (a-b). This has now progressed to the Vagnari vicus site and beyond. Most canals have filled with eroded sediment from the tributaries and are choked with vegetation. Vagnari North exposure upstream erosion 2013 and 2018 (c-d). In 2013 (e) the main channel was one meter higher than in 2014 (f), and the channel was filled in places with armoured mud balls. Photos P. Wigand.

33

The Making of a Roman Imperial Estate

Figure 3.29 Comparison of the sum of spring and summer rainfall with the sum of the radiocarbon dates on erosion in southern Italy (red line). There is clearly a correspondence during the last 5,500 years.

shows a close correspondence with the dated erosion record of Piccarreta et al. 2011. In addition, it is the decline of spring and summer precipitation after 5,500 BP in the region that seems to be responsible for the onset of the Holocene cycles of erosion (Figure 3.29). And it is the increase in spring and summer rainfall during the last 5,500 years that causes cycles of erosion.

a major period of stream incision and associated slope erosion, which would have progressed upstream from the coast, and eventually reached the Vagnari area and beyond. The scale of the incision recorded in the streams of the region at 7,000 years ago is as great as any of the succeeding climate caused erosion/incision events after 4,000 BP. Similar events of stream incision occurring elsewhere around the Mediterranean tend to confirm the hypothesis that a lowering of sea level was the primary cause. However, forest clearance may have played a role as well in this erosion cycle. And tectonic uplift has been ongoing as well, contributing to stream incision.

There are only two major exceptions to the correspondence between climate and erosion, the earliest begins just after 7,500 cal. BP, but declines significantly after about 6,900 cal. BP though it continues to about 6,300 cal. BP. The latter cycle begins about 3,500 cal. BP and continues to just after 3,000 cal. BP. It also coincides with the arrival of a wave of agriculturalists who had entered the area and may have been clearing forests. However, the scale of the erosion event seems much larger than would be expected for small scale or primitive forest clearance. However, it does coincide with the age of the youngest marine terrace (described by Brückner 1980) on the coast at Metaponto, which is the same age as Marine O18 isotope stage 1. This is the age of the highest Holocene sea level globally. That is, as the Pleistocene glaciers melted ocean levels rose in response to increased global temperatures until they began to drop again after the Holocene thermal maximum. Thereafter, sea level fell over five metres as global temperatures cooled and global water began to form glaciers again. This drop of sea level would have caused the base level of streams in the region to fall by five metres as well and would have resulted in

This episode has no corresponding cycle of summer rain. It does correspond to a decrease in spring discharge just after an episode increased discharge at Baron Spring southwest of the Vagnari village site, which is dated to 3,702 cal. BP (Wigand and McCallum 2017: tab. 1). This erosion event may also be related to the late-Bronze Age cultural activity in southern Italy. It corresponds to a series of intense droughts over a 150-year period from 1250 to about 1100 BC that led to severe cultural disruption in the eastern Mediterranean. The interplay between spring and summer rainfall amounts may also have affected the scale of summer erosion events. Cycles of erosion seem to require a preceding drought and then a subsequent summer wet condition. Lower spring precipitation would reduce vegetation density increasing the exposure 34

Wigand: 3. The Landscape Context of Vagnari, Past and Present

b

a

Figure 3.30a–b Modern summer erosion currently typical of the tributaries of the upper Basentello River. Upstream incision (a), here at at Arroyo Italiano II, is typical and progresses at rates of over 2m per year or more depending upon the substrate. Water fills these channels to the brim during torrential rainfall events. Armoured mud balls (b) are typically found after the flood season. These are over 35cm in diametre and indicate a flood water velocity of over 4m per second. Photo P. Wigand.

Figure 3.31 Principal component analysis of pollen from Lago Grande Monticchio core 90D from the crater of Monte Vulture. Both the primary plants for each cluster as well as the dates associated with each cluster are plotted above. In this Biplot, the further from the centre of the diagramme the species or date the stronger the correlation of that group.

35

The Making of a Roman Imperial Estate

Preboreal

cold-dry

13,000-11,500 BP

(climate impact dominant)

Boreal

cool-dry-warming

11,500–8,900 BP

(climate impact dominant)

Atlantic

warm-moist-max. temp.

8900–5700 BP

(climate impact dominant)

Subboreal

warm-moist-cooling

5700–2600 BP

(climate impact predominant)

Subatlantic

cool-moist-cooling

2600–100 BP

(human impact predominant)

Table 3.2 The Blytt-Sernander periods.

of the ground surface to the erosive force of summer torrential rains. If spring rains were higher increased vegetation density would have reduced the scale of summer erosion. The weather conditions that have occurred during the last two decades in the southern Mezzogiorno are probably much like those that characterised drought cycles in the past, and they may provide a model for what the erosion cycles that have characterised the middle to late Holocene were like (Figure 3.30).

The increasing abundance of these species through time parallels the increased impact of people on the landscape. Ratios of the dominant pollen types included in each of the three groups for the core from Lago Grande di Monticchio reveal both the three-part division of the Holocene and an underlying five-part division that mirrors the five-part Blytt-Sernander system of northern Europe (Manten 1967; Iversen 1973) (Table 3.2). These periods are most obvious in the upper diagram of the pollen ratios from Lago Grande di Monticchio (Figure 3.32).

Human impact upon vegetation cover is best recorded in the pollen record. At Lago Grande di Monticchio there is a significant statistical break in the pollen assemblage just before 5,000 cal. BP (Allen et al. 2002: fig. 4; Watts et al. 1996). A principle component analysis of the pollen from the cores taken from Lago Grande di Monticchio covering the last 12,500 years results in the formation of three statistically significant groups. The time spans of these three groups coincide with the three major breaks in Holocene climate revealed in the MCM reconstruction of Gioia dell Colle climate, 1) Group A: 12,500–10,000; 2) Group B: 10,000–5500; 3) Group C: 5500–present (Figure 3.31).

The countryside during the Roman period was characterised by an increase in olive orchards and grape vineyards, and the growing abundance of cereal and weed pollen indicates that areas of grain were increasingly common on the landscape. Beech forests dominated the hills, and alder thickets were common along streams and on moist hillsides. The magnitude and duration of cycles of erosion increase after 2500 years ago, suggesting that soil exposure to summer rains after the wheat harvest aggravated erosion. The late Hellenistic period through the late Roman period was actually characterised by few episodes of erosion but mainly by a period of soil formation. In fact, the soil that is dated to about 2100 years ago in the Arroyo Italiano I and Vagnari North exposures corresponds to the top of Campbell’s Unit 1 (Campbell et al. 2011). Campbell’s Unit 2 above it is characterised by a 500-year unconformity, which coincides with the period of erosion illustrated by Boenzi et al. (2008) in their Figure 7, and to the erosion cycle we see in both the Arroyo Italiano I and Vagnari North exposures. Therefore, although both geomorphic and paleobotanical evidence indicate that the influence of agricultural people from the early Iron Age onward was increasingly significant in southern Basilicata, the Roman period was a relative period of stability.

In the PCA analysis (Figure 3.31), Group A (blue circle) is the late Pleistocene plant assemblage of the region and is dominated by pine (Pinus), birch (Betula), grasses (Gramineae/Poaceae). The dates associated with this group in the PCA analysis are all prior to 11,400 cal. BP. Group B (orange circle) is a subset of the Holocene plant pollen types (red circle). The plants of this period reflect a shift away from tree dominance to shrubby species, including hazel (Corylus) hornbeam (Carpinus), hop-hornbeam (Ostrya), maple (Acer), roses (Rosaceae) and ivy (Hedera). The weed Plantain (Plantago) becomes common indicating the arrival of agriculture. In addition, pistachio (Pistacia) must also have been brought in by people at this time. The dates associated with this group in the PCA analysis is primarily from about 7000 to 2500 years. Group C (red circle outside the orange circle) encompasses roughly the last 2500 years. Oak (Quercus) and alder (Alnus), beech (Fagus) and chestnut (Castanea) were the dominant trees. Chestnut was introduced from the Middle East. Olive (Olea), grapes (Vitis), and cereals (Cerealia-type) also appear.

The siting of Vagnari vicus An archaeological field survey of the area lying between the Bradano and Basento rivers suggested 36

Wigand: 3. The Landscape Context of Vagnari, Past and Present

a

b

c Figure 3.32a–c Comparison of the pollen rations of the three dominant periods from Lago Grande di Monticchio. The tripartite division of the Holocene vegetation of southern Italy is clear, but there is also an underlying five-part division, which mirrors the Blytt-Sernander system of northern Europe. The lowest diagram compares the Blytt-Sernander system with the erosion cycles of southern Italy and the general cultural phases of the region. The cultural stages are general, but a significant change in land use and climate seems to occur at the end of the Bronze Age. This is seen elsewhere in the Mediterranean as well (Weiss 1982).

37

The Making of a Roman Imperial Estate relocation of settlements from flood plains to marine or river terraces may have been related to a change in climate as well. Agricultural areas seem to be associated with farmhouses and are often located along the stream courses. They suggest that this association with rivers, and of their lower rank tributaries, was related to small scale manufacturing and to the improved transport of products. Clearly, the streams flanking Vagnari were never transport conduits, but the upper Basentello did have a trade route associated with it. Springs also play a role in site location, but the current streams on either side of the Vagnari village site appear ephemeral. However, the stream just to the south of the cemetery appears to have a spring associated with a slab of conglomerate that may have come down from the plateau (Figure 3.33). This location would have been more accessible 2000 years ago, because the stream between the village and the cemetery may not have been as deep as it is now or have reached as far upslope as it does today.

Figure 3.33 Possible spring locality south of the Vagnari cemetery site. This stream drainage is the one in which the Arroyo Italiano exposure is located. Even today it has running water during the summer. Image P. Wigand, after a Google Earth photo.

that previous mapping of archaeological materials may have contained biases that favoured the discovery of archaeological site locations on both coastal terraces and other terraces over other locations (Abbott 2011). However, a more recent study by Di Leo et al. (2018) may lend some additional insight into the location of the Vagnari village and other sites, although their study concentrated upon the siting of Greek settlements in the Metaponto area. They constructed a model describing the interactions between sites and parameters such as elevation, slope, aspect, landforms, land use, and distances from rivers. They produced sensibility maps which were to aid archaeologists in locating new sites in the area. The primary aim was to determine the importance of the main geomorphological features of the study area that affected settlement site selection, within the reconstructed climate of the period. What was clear was that the occupation of mid-altitude marine terraces (or mid-altitude river terraces — the author’s addition) was preferred, and as a result agriculture spread upon these locations, also in part due to the recognition by the settlers that these areas had better developed soils. The realisation that these landforms had better conditions for the development of agriculture may have played a role in regional settlement from 4100 to 2000 cal. BP. Di Leo et al. (2018) suggest that these agricultural peoples may have expanded inland in a search not only for the most fertile lands, but also for better protected areas, away from the coast. The gradual

Today the area around Vagnari has been heavily ploughed, usually in an up and down-slope manner, so that grain, primarily pasta wheat (Triticum durum), can be grown. Ploughing in this manner has resulted in severe down-slope movement of topsoil. In many areas the rubified, last interglacial soil is easily visible as reddish patches in the dark brown of the upper hill slopes. In just as many places, the underlying, grey, marine/coastal sediments of the Adriatic-Bradanic Foredeep have been exposed. In addition, there is a growing contribution of Aeolian dust to the soils of southern Italy from the deserts of North Africa (Blanco 2003). This has probably occurred numerous times during the middle Holocene as well. Around Gravina, native vegetation comprised of macchia Mediterranea/ thorn shrub occurs on north-facing slopes in scattered clumps, along steep-sided stream channels. In places, some hilltops are covered by isolated stands of pineoak woodland which has been restored or allowed to go wild as part of the national park known as the Parco Nazionale dell’Alta Murgia. However, as global warming has progressed, the danger posed by wildfires has increased and threatens to destroy what little native vegetation is left. It promises to result in inestimable damage to the landscape through uncontrolled erosion.

38

CHAPTER 4

Vagnari Vicus: Buildings and Chronology Maureen Carroll, Kelsey Madden, and Jonathan Moulton

The Periods of Occupation on the Site (Plates 1–2)

Phase 5

For brevity, all cardinal directions are abbreviated: N = north, S= south, W = west, E = east.

Early 3rd century AD. The porch is added to the N side of the long portico building. Several other rooms were tacked onto earlier structures in the S part of the buildings. The winery of Phase 3 goes out of use, and some of the dolia defossa in it are probably removed now.

Phase 1 2nd–mid-1st century BC. The occupation of this phase is attested by soil deposits on a natural deposit of bluegrey clay and marl that was levelled to create a flat platform or a floor-like surface. A deep cylindrical pit filled with a rich assortment of finds, potentially from a late Republican votive deposit, also belongs in this period.

Phase 6 Second and third quarter 3rd century AD. The portico building is destroyed or severely damaged by a fire c. AD 225–250. Adjacent rooms are stripped of furnishings at this point, and the roof collapses over some of the rooms around the middle of the 3rd century or shortly after. Some buildings are dismantled, for which traces of scaffolding survive, and walls are robbed of their material.

Phase 2 Early 1st century AD. The early imperial phase consists of backfilled pits, the first masonry walls, and a stone drain running E–W. The structures of this period consistently have a SW–NE orientation that was retained through all subsequent phases. Throughout the 1st century, and into the very early 2nd century, the vicus witnessed several new alterations to the built environment.

Phase 7 To this phase in our trenches belong the topsoil deposits and those contexts churned up through ploughing activity; they cannot be more closely dated, although they do contain material of the 4th and 5th century AD.

Phase 3

Phase O

2nd century AD. This period sees a diversification in the economic life of the site. Viticulture is established, and the creation of the winery (cella vinaria) in the early 2nd century AD resulted in the repurposing or destruction of some earlier structures. Some buildings of Phase 2 are patched up and re-used or added to, and Phase 2 drains are either made redundant or are diverted.

This includes deposits and features that cannot be dated. A note on the room plan and the 3-D reconstruction When excavating contiguous trenches every summer, it is easy to lose sight of how buildings and features fit together in the bigger picture, and it was not always obvious how our buildings might relate to those explored earlier by the Edinburgh and Foggia teams. For that reason, we put together all the results of the trenches and the geophysical surveys in the N half of the vicus and attempted to join up features and create an overview of the buildings. This, then, formed the basis of the room plan presented here (Figure 4.1), and the

Phase 4 Late 2nd–early 3rd century AD. The portico building is added to the earlier suite of rooms and an older drain extended underneath it. This building, extending for a length of c. 28m, is subdivided or screened off into sections with wattle-and-daub walls or bins. 39

The Making of a Roman Imperial Estate

Figure 4.1 Simplified plan of excavated remains in the north-west part of the vicus, with attested and estimated room arrangements. Plan J. Moulton.

room plan was transformed into a 3-D reconstruction of those structures (Plate 3 and Figures 23.1 and 23.4). This entailed a certain amount of deduction and speculation, and a survey of other 3-D reconstructions of Roman sites helped us to visualise heights and angles of walls, roofs, and other features. In the end, the reconstruction is meant to provide an informed impression, rather than a totally accurate image, of what the built environment in the vicus at Vagnari looked like. The image captures the vicus around AD 200, just before the winery was stripped of its dolia and before the fire destroyed part of the buildings and the site was mined for its building materials.

The built environment in the vicus Phase 1 No architectural remains of habitation in this phase were encountered within the excavated area, but various features, such as deposits, a cylindrical pit, and tool marks on a ground surface, confirm that the site certainly was occupied. The natural deposit (4019/6030) of blue-grey clay with marl and other carbonate inclusions (probably the sub-Apennine blue clay formation of the lowest Pleistocene period — see Chapter 3) in the E part of the excavation was noticeably flat and level, and the linear tool marks on the surface in one area of it indicate that it had been 40

Carroll, Madden and Moulton: 4. Vagnari Vicus — Buildings and Chronology

coloured phase plans in Plates 1–2. It had four stratified fills in it: 4032 at the top, with ceramics of the 1st century BC, some broken tile, and fairly large stones; a deposit devoid of pottery below that, but containing an iron knife blade (4033); a sterile deposit (4035) as the next fill down; and, at the bottom, a clayey deposit (4039) containing a wide variety of pottery vessels (large fragments or intact vessels), fusiform perfume flasks, lamps, loom weights, iron implements, and animal bone belonging to the period around the middle of the 2nd and early 1st century BC (see Chapters 5, 17, and 19; Figure 4.4). This assemblage in the bottom fill (4039) resembles those placed in burials of the 3rd to 1st century BC in S Italy, for example, in the S cemetery at Heraclea (Giardino 1990). A similar selection of objects, in particular fusiform perfume flasks, loom weights, and iron implements, is encountered also in contemporary votive deposits in sanctuaries in the region (Muscetta 2009). The material in the cylindrical pit, therefore, is unlikely to be the clear-out of a normal domestic dwelling or to represent household waste. It bears no resemblance, for example, to a pit at Botromagno-Silvium that was filled c. 80–70 BC with tile, kitchen and household refuse, carbonised wood from domestic fires, and broken pottery, leading the excavators to interpret this as a domestic latrine (Small et al. 1994). The material in the Vagnari pit, instead, appears to represent objects that belong to a special context of the 2nd to early 1st century BC.

Figure 4.2 Levelled natural chalk and clay deposit with visible tool marks from Phase 1 (foreground). Photo M. Carroll.

Figure 4.3 Cylindrical pit 4021. Photo M. Carroll.

worked to make a suitable surface for occupation (Figure 4.2). This deposit is located in what would later become Rooms F and K, extending E just beyond wall 5003 and wall 4009/5026/6022, and ending abruptly at an outcrop of yellowish, sandy and pebbly clay at the E limit of the excavation. Interestingly, the tool marks on the levelled blue-grey clay and marl run in a SW– NE direction. We suggest that the directionality of the marks in the deposit must reflect the orientation of property boundaries and the buildings in the settlement of the 2nd and 1st century BC. It is that same SW–NE orientation that is then retained for the vicus and all the buildings in it from the 1st century AD. Lying directly on top of the sub-Apennine blue clay surface was a deposit of soil (5059) that had diagnostic pottery in it dating to the second and first centuries BC. Further firm evidence of occupation in this period is furnished by a deep, cylindrical pit with vertical walls (4021) that cuts through the blue-grey clay and marl layer and the underlying sand and clay (Figure 4.3). This pit can be seen on the plan in Figure 1.6 and on the

Figure 4.4 Grey-gloss oil lamp from the lowest deposit (4039) in cylindrical pit 4021. Photo M. Carroll.

41

The Making of a Roman Imperial Estate by other such assemblages in Puglia (Cassano 2011; De Venuto et al. 2016). Although it remains unknown to what structure or complex the pit at Vagnari might have belonged, it might be evidence for a consecration or purification ritual that took place when the late Republican settlement was established in the 2nd century BC. The votive deposit appears to have been sealed by the sterile deposit 4035, perhaps marking the end of this function of the pit. The top deposit (4032), at any rate, is a bit later in date (1st century BC), and the broken tile and stone in it signal the different character of this material.

The presence of the animal bones in the lowest pit deposit (4039) may provide a clue as to the nature of this special context. Several articulated elements of sheep/ goat and pigs were identified, suggesting deposition of the bones while they still had connective tissue on them (see Chapter 19). This indicates that they were put straight into the pit, rather than remixed into the fill from an earlier deposit, otherwise these elements would have become separated or lost. This argues against the material having been redeposited from burials. Although the bones represent mainly head and foot parts, they can be considered consistent with the remnants of a meal, perhaps a ritual meal, even if they are not the best cuts. Also, the remains of red deer were contained in the deposit 4039. We suggest, therefore, that this lowest material in the pit, with its mix of wild and domestic fauna and other artefacts, is clearly more than simply food debris or refuse and may represent material consistent with a votive deposit, as paralleled

Phase 2 (Figure 4.5) The earliest structural remains belong in this period. All buildings are orientated SW–NE, and all of the features within the confines of their walls follow that orientation.

Figure 4.5 Excavated remains belonging to Phase 2. Plan J. Moulton.

42

Carroll, Madden and Moulton: 4. Vagnari Vicus — Buildings and Chronology

of the construction trench. The intact wall 4017 had been robbed out in the N, leaving only robber trench 4057/6050 as evidence behind; here, all the stones had been removed, the original salvage crew leaving only the bed of yellow sand on the floor of the construction cut (Figure 4.6). The N end of the wall 3010 was intact for 2.10m, where it ends at the portico built in Phase 4. At its S end, wall 4017 abuts E–W wall 4006, also consisting of hewn limestone blocks up to 56cm in length, resting on foundations of rough conglomerate stones in a bed of yellow sand. Just above its foundations, the two preserved courses of limestone blocks were dressed on the N inner side to which a plaster and vertical tile facing adhered (see Chapter 13) (Figure 4.7). This wall is intact for 4.10m, but it is preserved at the W end only as a robber trench (6016). The W wall of the two rooms is formed by walls 4062 and 3037. Both are the remains of robber trenches, and nothing at all remains in 3037, but a few remnants of the wall that was removed in robber trench 4062, namely the foundation bedding of yellow sand we see in all other contemporary masonry, survived, indicating that this W wall belongs to Phase 2 and delimits Rooms B and C on this side (Figure 4.8). We have no idea how far Room B extended to the N, as construction of the portico building (Room D) in Phase 4 obliterated any earlier remains.

Figure 4.6 Robber trench (4057/6050) of the Phase 1 wall 4017, view from the north. Photo M. Carroll.

Rooms B and C were divided by an E–W wall 4059/3047, only the foundations of which survive, because it was dismantled to make way for another wall in Phase 3 (4013). Directly beneath wall 4059/3047 was a deposit (5032) that is dated by the ceramics to the second half of the 1st century BC and the 1st century AD. Continuing to the S of Rooms B and C are other rooms, although only one of them was completely preserved or excavated. Both Room L and Room J are bounded on the N by wall 4006 and robber trench 6016. Room J is located immediately W of Room L, its W wall (6012) running at a 90° angle to wall 6016. Of wall 6012 only a shallow trench 3.72m long and 52cm wide remains, as it was completely robbed out at a later, unknown date (Figure 4.9). Room L, with a cobblestone floor set in grey clay (4015), is delimited on the E and S wide by walls that are preserved only as a robber trench (4034) (Figure 4.10). The W wall of Room L has been reconstructed in the position in which the cobblestone floor (4015) in the room ended, although no tangible traces of a wall or robber trench survive here. The floor is made of the same kind of large ovoid and smooth pebbles that can still be found on the surface of the fields after heavy rains erode the soil, resulting from flood gravels deposited in the late Pliocene and early Pleistocene on marine marls. The diagnostic pottery in the soil 4023 that collected on the cobblestone floor indicates that the room was still being used in the 2nd century AD (Phase 4). Based on the surviving floor,

Figure 4.7 Wall 4006, faced on the north inner side (top) with a vertical tile facing. Photo M. Carroll.

Rooms B, C, F, G, J, K and L The limits of Rooms B and C can be determined on the W, E and S side. The E and S walls of Rooms B and C cut through the levelled blue-grey clay and marl platform 4019 of Phase 1, and the deposits in these rooms are at least 30cm deeper than the clay and marl deposit. The E wall (4017/3010) runs in a N–S direction for a length of c. 7m in total. Its masonry consists of cut limestone blocks up to 40cm long, set in a hard, yellowish, sandy mortar, all laid on a bed of yellow sand at the bottom 43

The Making of a Roman Imperial Estate

Figure 4.8 Lower right: wall 4006; background: wall 4059; lower right, remnant of wall 4062. View from the south. Photo M. Carroll.

Figure 4.9 Drone photograph of trenches in 2018. Lower right: robber trenches of Phase 2 buildings; upper trench: wall 6022 (Phase 5). Photo V. Ferrari and G. Ceraudo.

44

Carroll, Madden and Moulton: 4. Vagnari Vicus — Buildings and Chronology

Room L had an E–W width of 3m and a N–S width of 2m, for a total internal space of 6m2. The extent of Room K in this period is unknown, as the remaining walls are of a later date (Phase 5). But the natural, levelled blue-grey clay and marl (4019/6030) is a prominent feature of this room, and cutting slightly into the deposit is a single rectangular stone base (4016) which must have supported something upright, perhaps a pillar or post. Room F lay to the E of Rooms B and C, bounded in the E by wall 5003; Room G shares this wall. Four courses of this well constructed wall are preserved, with cut limestone blocks resting on two lower courses of large and medium-sized conglomerate chunks. The stones are bonded with fine, hard pinkish-buff mortar. The wall runs N–S for 3.5m, but because it was rebuilt and altered in Phase 4, we cannot tell how far N it originally extended. At the S end it has a return of 45° ending in a smooth, vertical face that must represent the E part of a doorway into and out of Room F (Figure 4.11). The cornerstone of the wall here is a cut limestone block measuring 30 × 30 × 30cm. The construction cut (5041) for wall 5003 is visible only on the E side of the wall at its S end where it cuts through the geology; the dark soil fill (5042) of this 2.5m-long cut contained pottery of the 1st and 2nd century AD. The later pottery in this feature may have ended up here as a result of repairs and alterations to the walls at this time.

Figure 4.10 Cobblestone floor (4015) of Room L, bounded by wall 4006. View from the west. Photo M. Carroll.

Drains Drainage of wastewater was clearly a concern in this period, as a sturdily built E–W stone drain (5045/4054) under the floors of Rooms F and C indicates. The walls of the drain consisted of cut limestone blocks resting on a foundation of cobblestones; the drain floor was made of flat limestone slabs (Figure 4.12). The stone capping slabs had been removed in antiquity, but the mortar fixing the slabs to the top of the side walls was still in place. It is uncertain how far W this drain originally extended, as in a second phase of its existence the W end was removed, as a robber trench (3048) indicates, and, at the same time, the drain was diverted to a new branch (5013/2024) heading NW. This new section had walls and a floor of roughly cut limestone and pieces of conglomerate of varying sizes; the internal height was 33cm, the internal width 29cm, and all internal surfaces were heavily coated with mortar (Figure 4.13). The capping stones were intact only at the W end. The slope of the whole drain (5045/4045/5013) in this second phase is about 3.3%, a difference in height of ca. 50cm from E to W over a distance of ca. 15m, utilising the natural down-slope of the terrain. The lower sediments in the original drain (5045) contained material dating to the 1st century AD, the fill (5020) of the extension to

Figure 4.11 Wall 5003, view from the south. Photo M. Carroll.

45

The Making of a Roman Imperial Estate

Figure 4.14 View, from the west, of wall 5003 in which a gap for drain 5045 has been blocked up with masonry and mortar (middle of photo) when the drain was given up. The drain may have emptied waste from a possible latrine in Room G, behind this wall. Photo M. Carroll.

Figure 4.12 View, from the east, of the stone drain 5045/4054, with parts of the side walls of the drain removed to show the cobblestone foundations. Photo M. Carroll.

Figure 4.15 Grain storage pit 5008 with remains of its backfill containing pottery, loom weights, animal bone, and metal. Photo M. Carroll.

Figure 4.16 Grain storage pit 6061 with the remnants of the ceramic dolium still in situ. Photo M. Carroll.

Figure 4.13 View, from the west, of the stone-built drain 5013/2024. Photo M. Carroll.

46

Carroll, Madden and Moulton: 4. Vagnari Vicus — Buildings and Chronology

the drain (5013), however, had no diagnostic material in it. In Phase 3, with the construction of the cella vinaria in Room A, the drain was largely destroyed, but a part of it in the E continued in use. A new drain (5051) was added to the stump of the original drain 5045 in Room F, directing the flow of water now in a N direction. This drain remained in use and was even extended in Phase 4, until it was abandoned in the late 2nd or very early 3rd century AD (see below).

hydrology of the site and the issue of water supply is offered in Chapter 23. Storage pits A total of 19 circular or ovoid storage pits can be assigned to this period, with a concentration in Rooms B, C, and F. We are confident that more pits exist, particularly in Room B, but, unfortunately, in 2015 when we were exploring this area we did not dig deep enough to reveal them, due to time constraints. Two pits (5034 and 5037) are located underneath the foundations of wall 4059 between Rooms B and C, and pit 4067 lies beneath wall 4062, the W boundary wall of Rooms B and C, giving us a relative sequence of activities here and indicating clearly that there were successive phases of development of the site in the 1st century AD. All pits were backfilled, with mixed refuse, some of it diagnostic enough to indicate that the backfilling took place primarily in the 1st century and perhaps just at the beginning of the 2nd century AD (Figure 4.15). A piece of white Pentelic marble revetment from the region of Athens in the backfill of one of these pits suggests that in Phase 2 at least some of the rooms in the vicus once had interior decoration with this material (see Chapter 14).

Drain 5045 starts in the E at wall 5003 which separates Room F in the W from Room G in the E, and the wall is crucial in understanding where the wastewater was originating from. The drain seems to be contemporaneous with wall 5003 and the wall, in fact, incorporates the drain and is built on it. A close examination of the masonry in the wall directly above drain 5045 shows that the original opening in that wall was blocked up roughly with chunks of stone and grey mortar, almost certainly when the drain finally was abandoned in Phase 5 (Figure 4.14). This opening, therefore, must have been the entry point for the wastewater and its origin must be sought in Room G. It is possible that wastewater flowed through a downpipe built into the upper level of wall 5003, as is attested for private toilets on the first floor of buildings in Pompeii where the (usually) terracotta downpipe emptied into a drain (Jansen 1997). However, the Vagnari drain is so substantially built and capacious, and in use for so long, that its connection with a simple toilet seems unlikely. Although nothing has survived in Room G, possibly due to the high level of natural geological features with few ancient deposits on top, most of which modern ploughing has disturbed, we suggest that there might have been a public latrine in this area that necessitated the kind of substantial capacity and durability evident in our structure. If that were the case, the somewhat elevated floor and seats of the latrine would have been located immediately behind wall 5003 in Room G. Either stone or wooden seats are possible (Jansen 2015), although neither has left a trace, as they could have been removed when the vicus buildings went into decline and abandonment. Human waste certainly entered the drainage system here, as soil samples taken from a silt deposit (5046) in the drain tested positive for Giardia duodenalis, a parasite transmitted by the faecaloral route (Ledger et al. 202) (see Chapter 22). We do not know how large Room G was in this phase, as it is not until Phase 5 that we have more evidence for walls marking the edges of the room. A latrine need not have been very big; the latrine at the villa at S. Pietro in Tolve in Basilicata, for example, was only 2 × 4.10m in size, and its waste material flowed into a stone drain making use of the down-slope of the terrain from the late 1st to the mid-3rd century AD (Di Giuseppe 2008: 361–62, fig. 7). It is unknown where the water needed to flush out a latrine in the vicus came from, but a discussion of the

Two circular pits in Rooms K and N clearly were created to hold a ceramic storage vessel. One of them, circular pit 6061, c. 50cm in diameter, still contained the lower part of a ceramic vessel that was backfilled (6062) with pottery, fragments of iron implements, and charcoal (Figure 4.16). The vessel had a base diameter of c. 24cm and its surviving height was 30cm; if the in situ remains represent about one third of this vessel, the intact container could have been up to 1m in height. The other pit, dug as a circular basin (4020), no longer contained a ceramic container, but the basin was lined with smooth mortar in which the impression of a dolium with a flat base 26cm in diameter was preserved at the bottom (fills 4025, 4026) (Figure 4.17). The internal diameter of the basin measured 68cm, indicating that the dolium in it had a large storage capacity. It is highly likely that not just these two pits, but all of them, once contained a vessel for the storage of foodstuffs, even though the storage container rarely was preserved. The location of many of them (5068, 4053, 4063, 4064, 4065, 4066, 5068, 6045, 6052, 5053, 5055), tucked out of the way at the foot of the walls of the rooms, may offer support for this idea (Figure 4.18). In the rooms of the so-called barracks at the imperial property of Villa Magna near Anagni, for example, pits for dolia and smaller storage vessels typically were placed in the corners of the rooms and against the walls (Andrews and Privitera 2016: 128–32; Andrews 2016: 234, figs. 6.5, 6.7). At Villa Magna they are interpreted 47

The Making of a Roman Imperial Estate

Figure 4.17 Mortar-lined grain storage pit 4020 with the impression of a flat-bottomed dolium preserved at the bottom. Photo M. Carroll.

Figure 4.18 Excavating a back-filled grain storage pit 4053, view from the east, with wall 4006 (left) and drain 4054 (right). Photo M. Carroll.

Figure 4.19 Excavated remains belonging to Phase 3. Plan J. Moulton.

48

Carroll, Madden and Moulton: 4. Vagnari Vicus — Buildings and Chronology

as containers for grain storage (Rice 2016a: 143–44). The archaeobotanical remains in the fill of two of these pits at Vagnari (pit 6052 and pit 4053) suggest that they also served this purpose here. Both botanical samples from these fills can be identified as material from late-stage crop processing which is a strong indicator of storage (see Chapter 20). Thus, the rooms in which the storage vessels are located must have had a residential purpose or been associated with collection and distribution of foodstuffs for the inhabitants of the vicus or for those in this sector of it. Grain can spoil when stored, however, and would not be kept for very long, so it is likely that the contents of the containers were used up fairly rapidly and replaced for the next period of consumption (van Oyen 2020: 40–43). In the next phase, Phase 3, storage facilities of this kind were no longer needed or were in operation in any of the rooms of the vicus we excavated. The addition of the cella vinaria may have altered the nature and purpose of this sector of the site. Phase 3 (Figure 4.19) The winery, Room A An important addition to the vicus in this phase was a cella vinaria (Room A), or winery, indicating indirectly that vineyards were part of the imperial exploitation of the landscape and that the production of wine had become a staple of the estate economy (Carroll 2016). Viticulture required a considerable capital investment, especially if it involved preparing new ground for vines, as it must have done at Vagnari, necessitating a waiting period of at least two years in which the vines need to be tended but do not bear fruit and cannot yet bring in revenues (Kehoe 1988: 100–03). The facilities needed for the pressing of grapes, the collection and fermentation of the juice, and the storage of the wine in large ceramic vats (see Chapter 6) represented further expenses for landowners. Not all of these facilities, however, were revealed in our excavations. Where the wine was stored in ceramic dolia defossa is clear, but no traces of the mechanical press (torcularium), a timber device usually on a masonry base, the treading tank (calcatorium), or the basins (laci) for collecting the must were found at Vagnari, however, although these should have been in a room in the immediate vicinity of or at least close to the cella vinaria. The only explanation we can offer is that the tanks must have been dismantled and removed once the winery went out of use, or our trenches just missed the relevant facilities. As for the torcularium, mechanical presses were not always present in Roman wineries, especially in the more modest ones, and so the lack of one in the small winery at Vagnari might be because it never existed in the first place (Rossiter 1981: 348–51). Even at the luxurious imperial villa of the Antonines at Villa Magna no trace of a torcularium was found archaeologically anywhere near the treading

Figure 4.20 The cella vinaria at Vagnari, showing intact or partially intact dolia in mortar basins (6, 7, 9), robbed-out dolium impressions (2, 5, 8, 10), and circular darker traces of dolium holes not excavated (1, 3, 4). Plan J. Moulton.

tank, leading Fentress and Maiuro (2011: 347–48) to suggest that the grapes trodden in the tank might have been carried off in baskets for pressing elsewhere. Some of the walls of Room A survived at foundation level, whereas others had been robbed for their material. It is clear that the winery was added onto Rooms B and C, as walls 3037 and 4062 already existed in Phase 2. The W wall (2018) of the cella vinaria was largely robbed out later, at least in the S section, and this was the case with the S boundary wall (6049) as well. The N wall may have been positioned roughly where the S wall (2011, 3008) of the portico building was erected in Phase 4, but we cannot be absolutely certain. The floor of the cella vinaria was a layer of greyish-white mortar c. 10cm thick laid on a bed of broken brick, tile, and sand (2012, 3007, 3009, 4008, 6003). The pottery from the deposits (5017) directly beneath that floor includes Regional Red Slip, with one piece imitating Italian sigillata form Conspectus 37, dating from about AD 20 to 100, giving 49

The Making of a Roman Imperial Estate us a terminus post quem for the winery. Stratigraphic relationships between features of Phase 2 and Phase 3 also confirm this chronological assessment. The cella vinaria at Vagnari was modest in size, measuring 5.50 by 8.20m internally (c. 45m2), if we assume that its N boundary wall corresponds in some way with the later placement of the portico building. The wine was stored in an unknown number of dolia defossa inserted in rows into the mortar floor of the winery, of which 10 have left traces archaeologically (Figure 4.20). No dolia were detected in the area of Rooms M and D, so the cella vinaria probably did not extend beyond what we revealed in Room A. There would have been enough space for up to 18 of them in Room A if they had been arranged closely in three rows, but we cannot confirm that there had been that many (Figure 4.21). In places, the natural bedrock consisting of calcareous conglomerate, was not very deep and this may have prevented the insertion of some dolia or prohibited the depth at which the basins for them could have been dug. Each dolium was placed in a circular depression that was lined with smoothed mortar; around the rim of the mortar basin ran a slightly raised collar of mortar (Figure 4.22). A partially preserved channel (3025) cut into the mortar floor may have directed overspills or water from the cleaning of the containers and the floor out of the room, although its W end could no longer be recognised (Figure 4.23). The complete lack of any roofing material retrieved from the winery (Room A) suggests that it was open to the sky, as the wine storage room was at most other Roman sites. For wooden posts supporting an awning or canopy over the dolia, as at Villa Magna, there is no evidence (Booms et al. 2016: 97– 98, fig. 5.34). The dolia are described and the fabric and residue analysis of the sherds discussed in Chapter 6, but it is worth noting here that the Vagnari dolia defossa were made in workshops on the Tyrrhenian coast of Italy and transported to the estate, rather than having been sourced locally in Apulia (Montana et al. 2021).

Figure 4.21 Mortar-lined dolium basins, with a large fragment of a dolium preserved in basin 3020 (left) and just the basin for 3012 (right). View from the north. Photo M. Carroll.

Figure 4.22 Mortar-lined basin 3012 for a dolium, view from the north. Photo M. Carroll.

Changes to Rooms B and C Rooms B and C are further defined in this period by the erection of one new wall and the re-use of others from Phase 2. The limits are still represented by wall 3037/6042 (on the W), wall 4017/3010 (on the E), and wall 4006/6016 (on the S); all these walls belong to Phase 2. This space was once divided into two rooms in Phase 2 by wall 5037/3047, but that now went out of use and was replaced by another wall (4013) built slightly S of it (only present as a robber trench). This now determines the size of Room C as 15m2, 3m (N–S) × 5m (E–W). The size of Room B is less certain, because of the unknown exact placement of the N wall.

Roman dolia stayed in place in the ground for many years in the wineries in which they were placed, and they were considered a permanent part of a cella vinaria. Peña (2007: 46–47, 216) recently suggested that such containers may well have been in use for 20 or 30 years, or even longer, as is possibly the case in the Caseggiato dei Doli at Ostia, where dolia appear to have been used for 40–70 years. Ethnographic parallels in the Mediterranean also point to the use of wine storage jars for over 100 years (Carrato 2017: 180). With this in mind, and if wine was being produced and stored at Vagnari for about a century until the winery was given up in Phase 5, the dolia in the cella vinaria might have been replaced individually or in groups when necessary or used for the entirety of the lifetime of the winery, but it is impossible to be more specific.

What Rooms B and C were used for is unclear. They were badly disturbed in the stripping of the building in Phase 6, and not even a recognisable floor was detected. Here we found panes of window glass (Chapter 7) and (sometimes large) pieces of marble revetment (Chapter 14) left over from the demolition (see below). The 50

Carroll, Madden and Moulton: 4. Vagnari Vicus — Buildings and Chronology

broken pieces of marble revetment represent what was left when larger slabs of marble were pried off the walls and floors, and they all have remnants of the mortar adhering to them that once fixed the cladding in place. Rooms B and C were the only ones in the building to provide contextual evidence for this material, and this can be interpreted in two ways. Either this material once was part of the furnishings of Rooms B and C, or the pieces of marble veneer and window glass simply ended up in these two rooms after the dismantling gang had finished in Phase 6, as might be the case at the late Roman bath house on the Philippiani estate at Gerace on Sicily (Wilson 2020: 492).

Figure 4.24 The western part of drain 4054/5013 cut by the placement of dolium 3020 in Room A. View from the east. Photo M. Carroll.

At any rate, storage of foodstuffs had played an important role in Rooms B and C in Phase 2, but pits for dolia and storage vessels ceased to be dug in Phase 3, so perhaps the rooms no longer had a domestic or residential character. It may be that with the establishment of the winery, these rooms functioned in a different way that we cannot determine.

Figure 4.25 View from the southeast of stone-built drain 5045 (left), blocked up so the water and waste could be diverted into stone drain 5051 (right). Photo M. Carroll.

Changes to the drainage system The creation of the cella vinaria put an end to the functioning of the stone-built drain that ran E–W and serviced Room G, the possible latrine, in Phase 2. The W parts of drain 4054/5013 were cut by the placement of dolium 3012 and dolium 3020 in Room A (Figure 4.24). A short stretch of the original drain 5045 where it exited wall 5003 was retained, then blocked up after 80cm to the W and diverted in a N direction as stone-built drain 5051/3035; later, in Phase 4, it was extended further N as drain 1009 (Figure 4.25). Thus, in Phase 3, the downslope of the Vagnari plateau from E to W ceased to be factored in for drainage in these buildings, and, instead, the S to N down-slope of the terrain was utilised to dispose of water and waste.

Figure 4.23 Channel 3025 cut through the mortar floor of the cella vinaria, view from the east. Photo M. Carroll.

51

The Making of a Roman Imperial Estate

Figure 4.26 Excavated remains belonging to Phase 4. Plan J. Moulton.

indicate, this building was c. 28m in length (E–W) and 4.25m wide (N–S); it had an internal width of 3m. The remains of a threshold and the E edge of a door frame, a large upright limestone slab, were excavated c. 13m E of the W edge of the building, so that access could be gained from the N, but the width of the door cannot be determined, as the W edge is now missing (Figure 4.29). This section of wall 1005 was of a slightly different build in that the rubble foundations were mortared and topped by a course of re-used tegulae providing a flat surface onto which the surviving upper course of wellmortared limestone pieces were fixed. Perhaps this gave the doorframe additional stability. There probably was at least a second door along the N front of the building.

Phase 4 (Figure 4.26) The portico building (Room D) and adjoining changes In this phase, a long rectangular building on the N edge of the vicus was added onto Rooms A, B, F, and I. The external walls (1008, 1013, 2004, 2011, 2013, 3008, 3039), preserved to a maximum of four courses, are between 50 and 70cm thick and consisted of pieces of limestone, conglomerate, and broken tile bonded in clay (Figure 4.27). That this wing was attached to the older rooms to the S is evident when viewing the join between the W wall (2018) of the cella vinaria (Room A) and the W limiting wall (2013) of Room D (Figure 4.28). The construction cut of wall 2013 visibly cuts through the marl and soil on which wall 2018 (the W wall of the winery) rests and is filled with darker soil. As our excavations, the trench dug in previous excavations (Favia et al. 2011: 123–25, fig. 4.49), and the geophysics

The building was simply appointed, with a very hard, beaten earth floor (2025/3033) and stone walls skimmed internally with a coarse grey plaster (Figure 4.30). It is possible that the building was internally articulated in 52

Carroll, Madden and Moulton: 4. Vagnari Vicus — Buildings and Chronology

Figure 4.27 View from the south at east-west wall 3008 (top) abutting onto the older north-south running wall 3010 of Phase 2 (right). Photo M. Carroll.

Figure 4.29 View from the southeast of the northern wall 1005 of the portico building (Room D) with a doorway and threshold. Photo M. Carroll.

Figure 4.28 View from the east of wall 2018, the western wall of the cella vinaria in Phase 2 (right), cut by wall 2013, the western wall of the portico building, in Phase 4 (left). Photo M. Carroll.

some way with wattle and daub walls, as this material lay in a burnt deposit 15cm thick (2010/3011) on the earth floor (Figure 4.31). The deposit was most obvious in the furthest W part of Room D. The burnt daub consisted of clay mixed with some sand, small stones or pebbles, tile and pottery fragments, and lots of straw, potentially from the stubble or stalks of harvested and discarded wheat that was one of the cereal crops on the estate (Figure 4.32). It was heat reddened and had many black patches in it where organic material had burned; lots of charcoal also was present. Some of the indentations in

the daub may represent wood or wattle, and the surface of the daub that would have been visible is smoothed and roughly skimmed with grey plaster. Wattle and daub (craticii), as a cheap and quick method to create partitions, was flagged up by Vitruvius (On Architecture 2.8.20), but he was aware that it could easily ‘catch fire, like torches’. We did consider that the burnt material could have come from walls of pisé or rammed earth, or even cob, but we rejected this idea because such walls usually were very thick and the material used in them normally is considerably finer (Di Giuseppe 2008: 53

The Making of a Roman Imperial Estate 359; Russell and Fentress 2016; Giada et al. 2019; Wilson 2020: 482–83). A much closer parallel for our material comes from the rural craft site at Podere Marzuolo in S Tuscany, where burnt wattle and daub was found in a layer of collapse and in the form of two in situ screens or dividers in a context of the 1st century AD (van Oyen et al. 2019: 77, fig. 6, originally interpreted as roofing material). This material, like ours, is friable and coarse grained, with impressions of the wattle frame. We have reconstructed the portico building (Room D) as a two-storey structure, because it is a type known broadly at Roman sites, especially with the later addition of a portico or porch (Room E) on the N front (Andrews and Privitera 2016a). The upper rooms of the building may have been used for storage of various goods or as very basic accommodations. We propose that the ground floor was at least in part also used for storage, potentially grain storage. The ground floor at the W end of the building appears to have been used in this manner, to judge by the cereal remains contained in the burnt daub on the floor; these remains indicate late-stage crop processing and point to storage (see Chapter 20). They were found in the daub only along the walls of Room D, not in the middle of the room. It may be that the wattle and daub that collapsed on the floor comes from partitions or even bins made of this material that were located along the walls of the room. When the partitions burned in the fire, the grain burned too. Since the 1st century BC, large rectangular and undifferentiated buildings, internally

Figure 4.30 Wall 3008 of Room D with its internal plaster skim in situ. Photo M. Carroll.

Figure 4.31 Burnt daub 2010/3011 lying on the beaten earth floor in Room D in the portico building. Photo M. Carroll.

Figure 4.32 Chunks of burnt daub 2010/3011 with black patches and burned out organic material. Photo M. Carroll.

54

Carroll, Madden and Moulton: 4. Vagnari Vicus — Buildings and Chronology

divided based on ad hoc needs, were regularly used to store grain in loose heaps, sacks or bins (van Oyen 2015: 111). One of these, the granary at Blera (località Selvasecca) is of similar size (28 × 6m) to the portico building at Vagnari (van Oyen 2015: 103, Table 1.). But such buildings could be polyfunctional, like the edificio meridionale at Drei Canè at Mezzocorona, where wooden walls dividing spaces into compartments and a second storey (albeit supported on central piers) made the building ideal for storing foodstuffs such as cereals, but also other commodities (Pellegrino 2017: 461, fig. 8). In fact, in the E section of Room D, where we opened our first trenches in 2012, there was no hint of a collapsed burnt daub deposit, so it is possible that only one part of the building, the W part, was used to store cereals. Other changes to the buildings neighbouring the portico building to the S include the N rebuild or extension of wall 5003 built in Phase 2. Walls 1014 and 5028, were added and keyed into the N end of wall 5003 for a length of c. 3.75m and were made of the same material as other walls of Phase 4. In this case, however, we have a good cross-section through wall 5028, showing that the five surviving courses consisted of a foundation course of roughly cut limestone pieces, followed by a levelling

Figure 4.33 The various courses of wall 5028. Photo M. Carroll.

Figure 4.34 Sondage with walls 4012 and 4018. Photo M. Carroll.

55

The Making of a Roman Imperial Estate course of pebbles on which flat tiles rest, and then three upper courses of irregular limestone pieces (Figure 4.33). This wall possibly replaces an older wall, maybe even 5003, on the same alignment. It meets and abuts the S wall (1013) of the portico, and it divides Room F from Room I. The S wall of Room I is represented by E–W wall 5025, a poorly constructed wall of unmortared conglomerate, limestone and broken tile that is superficially keyed into wall 5003. We did not excavate the E wall of this room, but it almost certainly is a continuation of a NS wall excavated by Small, so that Room I probably had an internal size of c. 3.75 × 4m. In 2016, we opened up a modest sondage in the S part of the site, and in this trench, we encountered two walls of different phases (Figure 4.34). We have referred to this poorly understood configuration of walls as Room O on our plan. Wall 4012, probably belonging to Phase 4, on the basis of the masonry technique (conglomerate, stones, clay bonding), had a SW–NE orientation and continued for a length of 1.70m; it was 50cm wide. The other wall 4018 was added in Phase 5. Figure 4.35 Stone-built drain 3035, the northern end of drain 5045 in Phase 3. Photo M. Carroll.

Figure 4.36 View from the north of stone-built drain 1009, the Phase 4 of the Phase 3 drain 3035/5051 (left). The northern extension of this drain, under the Phase 5 portico porch, can be seen here. Stone-built drain 1012 (right) under the floor of the Phase 4 portico is shown to the right. Photo M. Carroll.

56

Carroll, Madden and Moulton: 4. Vagnari Vicus — Buildings and Chronology

Phase 5 (Figure 4.38) The portico building extended to the N (Room E) Not very long after the portico building was constructed, a portico or porch (Room E) was added for the entire length of the structure on the N side. The walls (1005, 2005) of this portico are 50–58cm wide and constructed of un-mortared conglomerate, limestone, and broken tile (Figure 4.39). This porch (Room E) had an internal width of 1m, so it was just wide enough for people to pass along the front of the building and enter it. The floor of the porch was an earthen one. We have reconstructed a low wall on the N side of the porch which either had timber supports or brick columns supporting a sloping roof which would have offered protection from sun and rain. The brick segmental tiles for columns 26cm in diameter (see Chapter 13) could well have been used in this context, as such columns, although slender, could easily have supported the porch roof.

Figure 4.37 Northern Phase 5 extension of stone-built drain 1009 where it pierces and flows through the northern outside wall 1008 of the portico porch. View from the south. Photo M. Carroll.

Changes to the drainage system

Changes to the drainage system

When the portico building was built, the Phase 3 stone drain (3035/5051) in Room F was extended to the N to run under the floor of Room D as drain 1009 and to empty out down the slope beyond it (Figures 4.35 and 4.36; see Figure 4.25 for drain 5051). Drain 1009 had upright walls of cut limestone and an internal width of 15cm at the S and 29cm at the N. The stone slabs used to cover the drain were still in place almost everywhere, and when the capping stones were removed, a second course of capping stones was revealed under them, suggesting that the drain had been repaired or doubly strengthened at some point. In this phase, this drainage system from beginning to end had a length of c. 10m.

The drain 5045 originating at wall 5003 in Phase 2 now ceased to function completely, and the opening for the drain in that wall was blocked up. On top of that junction between drain and wall in Room F, a shallow pit was dug (5047) that either was a hearth or a disposal area for burnt debris from a hearth (see below). Drain 1012, built in Phase 4, was extended now between the N wall (1008) of the older portico building (Room D) and the N wall (1005) of the added portico (Room E). Where it went under the floor of Room E, this extension made use of upright slabs of conglomerate for the walls and a mixture of flat stones and tegulae for the floor. Where it went through wall 1005, a peaked ‘tunnel’ of angled conglomerate slabs was built into the wall to drain out further N (Figures 4.36–37). Outside the portico, immediately N of the stone drain 1012, a small 1 × 1m sondage revealed the remains of a linear cut (1023) about 30cm wide in the soil; the dark fill (1024) of the cut was clear. Unfortunately, torrential rainfall and flooding in the 2012 season prohibited us from further investigating this feature, but it could be that it is a remnant of a culvert cut into the ground and lined with tile, possibly imbrices, that allowed wastewater to flow out of the building.

Parallel to this drain, another one (1012) made of stone and capped with stone slabs was constructed 1.75m further E under the floor of Room D and Room E (Figure 4.36). We could follow this for 5m to the N limits of our trenches. The drain had an internal width of 26cm, and it varied in depth between 14–25cm from the top of the walls to the stone floor. The walls of the drain were made of upright cut slabs of limestone and the floor consisted of flat stone slabs. But this was not consistently the case over the whole length of the drain, because large, upright chunks of conglomerate were used to create the drain walls and tegulae were used for the floor for the last c. 1.5m at the N end, suggesting that this last section was added on as an extension (Figure 4.37). We propose that this happened in Phase 5 (see drain 1012N below). We do not know where this drain originated in the S, but if we reconstruct it continuing straight for several metres in that direction, it would have gone under the floor of Room I and perhaps beyond. It was not found in Room G beyond that, so it may have veered to the E somewhere in the vicinity of Room H, but, because we did not excavate that area, we cannot be certain.

Rebuilds and additions The buildings underwent rebuilding and extension in the SE. Room G is now defined on its E side by a wall (5030), but its full length is unknown. If we are correct in reconstructing a public latrine in Room G that went out of use when the drain emptying it (5045/5051/3035/1009) was blocked up and made useless in this same phase, the room’s reconfiguration and new E boundary wall could tie in nicely here. An extension to wall 5003 at 57

The Making of a Roman Imperial Estate

Figure 4.38 Excavated remains belonging to Phase 5. Plan J. Moulton.

its S end is indicated by walls 4009/6022 for a length of 6.5m, ending at other contemporaneous walls (6047 and 6048) set at a right angle for a preserved length of 2.9m. Both wall 6047 and 6048 were keyed into wall 6022, and their masonry and bonding are the same (Figure 4.40). The material used to build wall 4009 included a piece of a large dolium lid from the winery (Chapter 6, Cat. 6.6, Figure 6.4), indicating that by now the winery had gone out of use. All these walls are poorly built, using large chunks of quarried conglomerate, smaller pieces of limestone, and broken tile held together with a clay bonding, and they frame Room G to give it an internal width of c. 1.75m and an internal length of just under 9m. It is unknown how much further W wall 6048 continued, but we have reconstructed this as the N boundary wall of Room N. The room we have labelled as Room H, which would have shared wall 5030 with Room G, was not investigated by us, rather it is indicated by the excavations by Alastair Small.

In the area of Room O, a NS orientated (4018) wall with a dog-leg is added to the Phase 4 wall 4012. The wall is very roughly constructed, using very large, irregular chunks of quarried conglomerate, some cut stone, and broken tile in a clay bonding. It runs for a preserved 1.13m to the N, then turns at a 90° angle to continue to the W for 70cm; both sections are about 46cm in width. Further W in the vicus, S of the former cella vinaria (Room A), another NS wall (6011) was erected that we have reconstructed as the eastern boundary wall of Room M. It is preserved for 4m, but it clearly was robbed of its material later, as only a scattering of large conglomerate chunks is left on the floor of the construction trench (Figure 4.9). This wall must have abutted the S wall (6049) of Room A, although the considerable extent of disturbance and robbing in this area in Phase 6 meant that absolute proof of this wall joint could not be furnished. 58

Carroll, Madden and Moulton: 4. Vagnari Vicus — Buildings and Chronology

Figure 4.39 Excavating the porch (Room E) of the portico building, with the external wall 2005 on the right. View from the east. Photo M. Carroll.

on the floor with signs of burning on them (Figures 4.41–42). All are up against or in close proximity to the E wall (5003) of Room F. Hearths 5064 and 5067 lie on top of the juncture between the now defunct drain 5045/5051 and wall 5003. The drain’s capping stones already had been removed by this time and the drain no longer was in use. It is likely that these hearths are an indication of domestic activity, such as cooking, and it is noticeable that they are all concentrated in Room F. It may be that in this period Room F, in fact, was slightly enlarged to the S. It is difficult to confirm absolutely, but the stone foundation (4099) lying on the levelled blue-grey clay and marl outcrop may have been a support for a post or pillar, or it may be a remnant of a wall that continued E to meet and abut wall 4009 and W to walls 4006 and 4034. Room F originally had a wider S entrance, marked by the dog-leg in wall 5003, but with the installation of hearths and a potential increase in domestic activity, the room may have needed to be closed off more effectively.

Figure 4.40 Detail of wall 6047. Photo M. Carroll.

Hearths It is in Phase 5 that we have ample evidence of the use of domestic fire within the buildings, in particular in Room F. Here, the relevant features are hearths consisting of either ovoid pits containing copious amounts of charcoal mixed with soil or of re-used tegulae laid flat

Only one other hearth (1006) of this period lay just outside and to the N of the porch (Room E) of the portico building. This consisted of broken tiles grouped together around the edge of flat, broken tegulae, and it 59

The Making of a Roman Imperial Estate removal of dolia (Carrato 2017: 178–79, fig. 88). Of the 80 dolia in the cella vinaria of the farm of the 1st and 2nd century AD at La Mougère de Vauguière in S Gaul, for example, 44 had been removed completely, leaving only the circular depressions in which they originally stood (Carrato 2017: 415–17, figs. 316–17). At Villa Magna, some of the dolia from the earlier Antonine winery were moved to a 6th-century winery on the site of the villa where they were reused (Fentress 2016: 229, figs. 6.3–6.4). In mortar basin 3020 (No. 7 on Figure 4.20) at Vagnari, the base of a dolium was still in place, whilst in the basin 3038 (No. 9 on Figure 4.20), the lower third of a dolium was in situ (Figure 4.43). Large quantities of sherds from smashed dolia were retrieved in the backfill of these two mortar basins: over 4kg in basin 3012 and over 4.5kg in basin 3038. This suggests that the dolia in these cases were already broken and abandoned or that they broke when they were being retrieved, with the debris dumped back into the basin. Some of the dolia, however, may have been retrieved intact; there was no trace of a dolium, for example, in mortar basin 3012 (No. 6 on Figure 4.20), so perhaps this is one that was

Figure 4.41 Tile hearth 5070. Photo M. Carroll.

Figure 4.42 Tile hearth 5044. Photo M. Carroll.

was filled with burnt organic material (1006) consisting primarily of winter wild oat (Avenasterilis L.) and a small amount of durum wheat (Triticum durum Desf.) and various weeds (see Chapter 20). The composition of the sample suggests that this assemblage is likely to represent weeds removed from cereal crop fields (with some durum accidently removed with the weeds) and intentionally discarded onto the hearth as waste or fuel.

Figure 4.43 Lower third of the dolium 3038, with a student as scale. Photo M. Carroll.

The end of the winery What remains in the cella vinaria at Vagnari, apart from the walls of the room, is the evidence for the removal of the dolia defossa in antiquity. When the winery went out of use, some of the dolia were either removed completely, to be re-used elsewhere, as in the Villa of Augustus at Somma Vesuviana, or smashed into pieces to be used secondarily as building material, as they were at the villa of Settefinestre (Aoyagi et al. 2018; Celuzza 1985). As a valuable commodity, some of them might have been sold. A passage in Apuleius (Metamorphosis 9.5-7) indicates this practice, and it is also suggested by the frequent circular voids in winery floors left by the

Figure 4.44 Human skeletons in the dolium 3038. Photo M. Carroll.

60

Carroll, Madden and Moulton: 4. Vagnari Vicus — Buildings and Chronology

taken for re-sale. In other cases, the removal of the dolia may have resulted also in the destruction of the mortar basins in which they sat (Nos. 2, 5, 8, and 10 on Figure 4.20). As mentioned above, a large piece of a dolium lid (Chapter 6, Cat. 6.6) was secondarily used as building material in wall 4009, separating Rooms K and G, at this time. It was in dolium defossum 3038 that we found two human skeletons, although these can be attributed to Phase 6 or later (see below, and Chapter 21) (Figure 4.44).

gang (see Chapters 7, 10, 11, and 14) between AD 230–250, although this must have been done under the protection of a largely intact roof because all of the churned up deposits (3040, 4022, 4024, 4027, 4028, 5004, 5033) containing the remains of room fittings lay stratigraphically below the spread of collapsed roof tile (1004, 3027, 4007). These deposits in Room C contained lenses of charcoal and ash (4052 and 5014), indicating that some burning had taken place (Figure 4.47). As the excavations of the bath house at Gerace on Sicily have demonstrated, the stripping and backfilling of rooms could be part of the same operation, with salvage material being moved around in the process (Wilson 2020: 490).

Diagnostic pottery in the backfill of the mortar basins has a range of dates. The bottom fill (3032) of dolium 3020 is later than the 2nd century, although 2nd-century material also is present, but above that is another fill (3028) with pottery ranging from about AD 230 into the 4th century AD. The top fill (3041) in dolium 3038 dates to the period between the mid-2nd and mid-3rd century. This suggests that the main removal of wine vats or the destruction phase of the cella vinaria took place in the late 2nd century or the 3rd century, with some later disturbances resulting in the deposition of more recent material (see Phase 6, below). Pieces of smashed dolia were recovered also in other contexts: 1780g in deposits on the winery floor and over 3kg collected in Room B. Other large base fragments of dolia (Chapter 6, Cat. 6.4 and 6.5) and a big piece of a tall, cylindrical dolium (Chapter 6, Cat. 6.3) were retrieved from the deposits representing the demolition of the buildings, but they were found in the area of Rooms B and C, rather than in the winery. The winery destruction, therefore, corresponds to Phase 5, or at the very latest to Phase 6, when the buildings suffered from a fire, dismantling, and collapse.

The roof then came down (or was brought down) over most of Rooms B, C, and L around the middle of the 3rd century or shortly after (1004, 3027, 4007) (Figure 4.48). The E–W wall (4013) separating Room B from Room C was still standing when the roof came down, leaving a void where the tiles fell on either side of the wall (Figure 4.49); the stones of the wall were subsequently removed. The tiles also stacked up against the still standing wall 4006 in Room C as they fell and accumulated at the N foot of the wall. We did not retrieve roof tile collapse in any other rooms in the vicus. It would appear, therefore, that this collapse had been taken away or robbed out for later buildings, perhaps for those S of the ravine, and it had been taken away rather thoroughly. This was not the case for the roofs over Rooms B, C, and L, which, for some reason, only partially were ‘mined’ for tiles. What was left behind clearly was no longer needed, as was the case at the end of the later settlement S of the ravine; when that settlement was abandoned in the late 5th century AD, there cannot have been any demand for this material anywhere in the vicinity, and the tiles were not retrieved (Favia et al. 2011a: 177, 184, 199, fig. 5.91). According to Tony Rook (2013), a tiled roof weighed about 90kg per m2. In Rooms B and C, we retrieved c. 1500kg (c. 1000kg of tegulae and 500kg of imbrices), but if we take the internal dimensions of these two rooms, 35m2, the roof would have weighed 3150kg, so a little less than half of the tiles has been left in this case; but if we take external size of the rooms, 53.4m2, the roof would have weighed 4806kg, so only about one third of the tiles had not been retrieved.

Phase 6 (Figure 4.45) Fire, damage, and material retrieval There was a fire in the portico building and the internal wattle and daub walls or compartments burned down and collapsed in Room D (AD 225–250). The roof of the portico building must also have fallen in at this point, but, apart from some imbrices on the porch floor (Room E) the tiles seem to have been salvaged for use in another (unknown) location or for another purpose. Presumably the fire in the portico building had an impact on the neighbouring buildings and rooms, although it is impossible to say whether it made any of them completely uninhabitable.

There appears to have been considerable industrial activity involving iron working going on in the NW corner of what was left of Room B, as we recovered almost 2.5kg of iron and iron slag in deposits mixed with charcoal (3014, 3022). The iron artefacts include smithing hearth bottoms, furnace slags, and furnace lining, as well as iron objects such as nails and a blade (see Chapter 11). The assemblage is a clear indication

After the fire in the portico building, the period of dismantling and quarrying for building materials began. Rooms B and C were stripped of their furnishings — marble revetment, window glass (Figure 4.46), and metal —, or salvaged material was cached here and the rejected items dumped by the dismantling 61

The Making of a Roman Imperial Estate

Figure 4.45 Excavated remains belonging to Phase 6. Plan J. Moulton.

6031 where it just grazes wall 6022 on its W side (Figure 4.51). This was excavated into the natural levelled bluegrey clay and marl 4019/6030 as a hole with a diameter of 36cm, but at a depth of 17cm from the top, the outline of the wooden post itself is clear; the post was 14cm in diameter, and was rammed in to a depth of 48cm in total. With the help of timber scaffolding supported on wooden posts driven into the ground, material from buildings could have been dismantled from the top of the walls to a level at which the stones could be broken out while the demolition team stood on the ground.

of iron working, both the formation of new pieces and the reshaping and maintenance of iron objects. This activity may have been a recurring one for a period of time, months or even longer. The iron-working debris here perhaps gives us insight into the presence of squatters or salvagers at the site because these deposits lay on top of or mixed with the debris of the collapsed roof, but still within the ruins of the building. They are located in Room B mainly in the corner between wall 3037 and wall 3008, where upright walls, even if damaged, could still have offered shelter to the smiths. The pottery from this deposit gives us a general range from c. AD 200–300.

It is not always possible to determine when some of the walls were quarried for their stone, as not all fills of robber trenches contain diagnostic material allowing us to pinpoint this chronologically. For example, linear robber trench 4064 was completely devoid of any material and it does not seem to connect to another feature, and so remains unphased. Pottery in the fill

Some buildings were dismantled now for their stone, for which the scaffolding holes along various walls, including Rooms F, K, L, and M is evidence (holes 4051, 5056, 5057, 6031, 6037, 6042, and 6043) (Figure 4.50). A particularly well-preserved post hole survives in hole 62

Carroll, Madden and Moulton: 4. Vagnari Vicus — Buildings and Chronology

Figure 4.48 Collapsed roof tiles 1004, 3027, 4007 in Rooms B, C, and L. Running through the deposits is wall 4006. View from the east. Photo M. Carroll.

Figure 4.46 Large fragments of a glass windowpane in situ. Photo M. Carroll.

Figure 4.49 Collapsed roof, with tiles on either side of wall 4013 which separated Room B from Room C, later robbed out leaving this void. View from the west. Photo M. Carroll. Figure 4.47 Deposits 4052 and 5014 in Room C, with traces of burning. View from the north. Photo M. Carroll.

(4029) of robber trench 4034, marking the E end of Room L, ranges in date from c. AD 100–300. The same is the case for the fill (6055) of the robber trench 6049 marking the S end of Room A. Soil deposits (3005, 3006) overlying the S wall (3008) of the portico building contained ceramics dating to the second half of the 3rd and the early 4th century, and a coin of Constantine (see Chapter 8) was retrieved in the deposit (4004) overlying the collapsed roof. A deposit of dark soil (5015) overlying stone drain 5051 in Room F contained pottery of the late 2nd century AD, but also a bronze coin dating to AD 330–335, and it is, therefore, difficult to know what phase of occupation such a deposit belongs.

Figure 4.50 Wall 6022 foundations and post holes 6031, 6042 and 6043 from scaffolding used to dismantle the wall. View from the north. Photo M. Carroll.

63

The Making of a Roman Imperial Estate dolium and came upon it by chance. Since the soil on top of the skeletons has material in it dating to the mid2nd to mid-3rd century AD, this must represent soil that was churned up while digging down to dispose of the human remains and then redeposited to cover them up. Someone, therefore, was living and dying at Vagnari in the 4th century, although it is a mystery why these individuals were not buried in the cemetery. Before the end of the 4th century, however, the inhabited area had shifted to the S side of the ravine, and several new buildings were constructed there and in use until they were destroyed in the late 5th century (Small 2011a: 29–33; Favia et al. 2011a: 143–229). Figure 4.51 Post hole 6031 from scaffolding in Phase 6. Photo M. Carroll.

Phase 7 Late Roman/Post-Roman and modern disturbance

These finds suggest that there was still some activity of an undefined nature on the site in the 4th century AD, but there is no indication how many or which buildings in this part of the vicus were inhabited as late as that. It may be that other parts of the vicus were still occupied at that time, and Small (2011a: 29) and Dalton (2014: 92– 93) interpret the cistern or reservoir S of the area of our excavations as still in use in the 4th century, until it was filled in deliberately in towards the end of that century. They also date the remains of a flimsy structure on the back-filled cistern, possibly an animal stall, to the early 5th century.

The top one or two soil layers on site were disturbed and repeatedly churned up by modern ploughing to a depth of c. 30cm. The soil was mixed with material that remained from the latest Roman occupation of the site and its demolition, and from some 4th-century activity on site; this includes a coin of Constantius II, dating to AD 354 to 358, in the soil (1002) immediately beneath the plough soil (1001). The modern disturbance of the soil overlying the ruins of the vicus not only churned up late material to the surface, however, as we also found a silver victoriatus in excellent condition that was minted in Rome in 211 BC (see Chapter 8).

A very few graves in the cemetery also date to the 4th century (Small et al. 2007: 136–37), and a new C14 date on the leg bone of one of the two skeletons in dolium 3038 now gives us a date range with the highest probability between c. 332–410 (see Chapter 21) (Figure 4.44). This is puzzling because in order to deposit the bodies (or skeletons?) into a partially preserved dolium suggests that the dolium remnant was still visible or that they dug down through the accumulated soil on top of the

In earlier excavations, an isolated inhumation burial, C14 dated to AD 420–610, was found N of the ravine in the ruins of the vicus, and the remains of apsidal huts were detected on the S side of the ravine, suggesting that some people were still occupying the decayed remains of the site before it was abandoned altogether (Small 2011a: 33–34; Favia and Prowse 2011).

64

CHAPTER 5

The Vessel Pottery and Lamps David R. Griffiths

with contributions from Coralie Clover Silvium, Gravina in Puglia. At Vagnari, material resulting from excavations at the cemetery was analysed by Kenrick (Small et al. 2007), and his publication of the late Republican and imperial-period pottery from the vicus has been invaluable in his 2011 study (Kenrick 2011: 373–76), especially the detailed descriptions of fabrics and assemblages from well-dated contexts from different areas of the site (see various sections in Small 2011). In addition, the extended type series produced by Kenrick based on material recovered from the cemetery has been essential to dating much of the material presented here (Small et al. 2007; unpublished data was kindly provided by the cemetery site director, Tracy Prowse, in advance of publication). Further work by Kenrick (2014) helped clarify the often complex and varied terminology used to describe certain pottery wares, especially for those vessels with a red slip (e.g. colour-coated, red slip wares, Regional Red Slip wares etc.). The extensive research on ceramic oil lamps from the area by De Stefano (2007; 2014) has proved invaluable to the analysis of the material considered here, as has the work by Disantarosa (2011; 2014) on amphorae from the sites at Vagnari and in the Basentello valley.

Introduction A total of 8592 sherds of pottery weighing over 87kg was recovered from archaeological excavations at Vagnari vicus between 2012 and 2018 (Table 5.1). Over half of the assemblage, 4699 sherds weighing 51.6kg, was recovered from archaeologically secure stratified deposits from six phases of activity (Phases 1 to 6). The remainder were recovered from heavily disturbed deposits and topsoil (Phase 7), and features that could not be placed into the stratigraphic sequence of the site (here referred to as Phase O). The pottery from Phases 7 and O contained material dating from the 1st century BC to perhaps the 5th century AD; the data on these phases have been combined in Tables 5.1 and 5.2, and only brief summaries of wares and their chronology are provided; however, certain pottery groups and vessels of intrinsic interest are discussed in more detail. The relative proportions of each ware class are presented in Table 5.2, followed by brief overviews of the quantities and types of pottery present. Pottery assemblage groups from certain phases and deposits, which are of particular interest or are key to dating deposits, are discussed in more detail within their phase summary, below.

Methods

Background

The author worked on the assemblage over four seasons of excavation, two weeks each in 2012, 2013, 2018, and three weeks in 2019; Coralie Clover recorded and illustrated several pottery groups during the summer of 2017 which have been integrated into the assemblage recorded by the author. In 2018, recording of pottery was undertaken by the author and two students, Katie Hullock and Sarah Hayes, from the University of Sheffield, funded through the Roman Society, to begin training as Roman pottery specialists. During the final study season in 2019, the author was assisted again by Sarah Hayes; further assistance with illustrations was provided by Kelsey Madden and Jonathan Moulton. To those mentioned, I offer my profound thanks.

Previous pottery analysis in the Vagnari region has been undertaken by a team led by Alastair Small (1992) working on material from excavations at the Iron Age and Roman Republican settlement on Botromagno-

The recording of the pottery was undertaken following a two-stage process, which was followed by further analysis after fieldwork. The pottery was recorded into a MS Access database, with the raw data extracted and

The broad phases and chronologies of the site were: Phase 1: 2nd– mid-1st century BC Phase 2: early 1st century AD Phase 3: 2nd century AD Phase 4: late 2nd–early 3rd century AD Phase 5: early 3rd century AD Phase 6: second and third quarter 3rd century AD Phase 7: topsoil and plough soil Phase O: out-of-phase deposits

65

The Making of a Roman Imperial Estate

Phase

Class

1

2

3

4

5

6

7 and 0

Total count/weight

Amphorae

3/623

2/696

1/83

2/344

54/2534

83/4739

145/9019

Fine Wares

69/993

65/489

8/60

12/612

77/259

633/4144

782/3920

1646/10477

Plain Wares

25/321

139/1493

12/187

17/74

89/602

1505/15776

1670/16033

3457/34486

Coarse Wares

73/433

175/3303

14/206

11/115

129/1053

1528/16894

1324/10964

3254/32968

2/92

2/5

8/12

37/158

30/89

79/356

Ceramic Oil Lamps Unguentaria Total

4/58

3/15

4/25

11/88

176/2520

383/5986

35/536

40/801

308/2275

3757/39506

3893/35770

8592/87394

Table 5.1 Bulk data by phase by count/weight (grams).

Phase

Class

1

2

3

4

5

6

7 and 0

Total

Amphorae

24.7%

11.6%

15.5%

0.0%

15.1%

6.4%

13.2%

10.3%

Fine Wares

39.4%

8.2%

11.2%

76.4%

11.4%

10.5%

11.0%

12.0%

Plain Wares

12.7%

24.9%

34.9%

9.2%

26.5%

39.9%

44.8%

39.5%

Coarse Wares

17.2%

55.2%

38.4%

14.4%

46.3%

42.8%

30.7%

37.7%

Ceramic Oil Lamps

3.7%

0.1%

0.0%

0.0%

0.5%

0.4%

0.2%

0.4%

Unguentaria

2.3%

0.0%

0.0%

0.0%

0.2%

0.0%

0.1%

0.1%

100.0%

100.0%

100.0%

100.0%

100.0%

100.0%

100.0%

100.0%

Total

Table 5.2 Relative proportions by ware class (by weight) and phase.

presented in the appendices as MS Excel spreadsheets. Stage One: initial bulk sorting of the pottery assemblage separated material into broad ware classes: amphorae (AM), coarse wares (CW), plain wares (PW), fine wares (FW), thin-walled wares (TW), ceramic oil lamps (LA), and unguentaria (UG). In order to progress through large assemblages during relatively short periods of fieldwork, bulk quantification was by sherd count and weight (in grams). Stage Two: diagnostic sherds (predominantly vessel rims, but also handles, bases, decorated body sherds) were recorded in greater detail to include fabric class, form, surface treatment, and decoration, where applicable. Each diagnostic sherd or multiple sherds from a single vessel were assigned a unique Featured Vessel (FV) number; the details of featured vessels form the primary record in Appendix 4. Quantification for featured vessels used estimated vessel equivalents (EVEs), based on the percentage of the vessel rim preserved; these data form the basis for the functional analysis of the assemblage below.

Ceramic oil lamps and unguentaria were recorded following the methods outlined above, and are integrated into this chapter, with items of note highlighted where relevant. Dolia were recorded and analysed by Carroll and colleagues, with results presented in Chapter 6 (see also Montana et al. 2021). To partly place the pottery assemblage within the context of the social, economic, and chronological history of the inhabitants of Vagnari vicus, a functional analysis of the pottery assemblage by phase is presented. This type of analysis aims to identify and highlight any changes over time for production and supply of pottery, the range, quantity, and relative proportions of vessels and vessel types in use in a particular phase, and whether any changes reflect the nature of occupation at the vicus, what activities were taking place (domestic, industrial, agricultural), and/or if the material culture reflects the social status of those who consumed it. This is a broad-brush approach, but the hope is that the data presented may be compared with other excavations in the region, and/or with other imperial estates in Italy and beyond.

As with many Mediterranean Roman-period sites, often with large pottery assemblages, many or most sherds in the later phases of sites with long periods of activity contain increasing quantities of residual material (Reynolds 2019). Where recognised, residual-in-phase sherds are noted in the appendices, and for functional analysis, excluded from assemblages relevant to a later phase.

The results are presented in the following sections: phased overview, functional analysis, and a discussion of chronology, production, and supply and consumption of pottery at Vagnari vicus. The appendices present a site fabric series with full descriptions (Appendix 3) and a catalogue of featured vessels (Appendix 4) by 66

Griffiths: 5. The Vessel Pottery and Lamps

phase and feature; illustrations of selected featured vessels (FV) by phase are presented in Plates 4–20. Fragments of oil lamps are presented in Figures 5.6 and 5.7, and they are catalogued in Appendix 5.

large fragments of Grey Glaze plates/lid plates (FV1027, 90/70–1 BC, Plate 4.8; FV1028, 90/70–1 BC, Plate 4.9; and FV1029, Plate 4.10; bowls (FV1031, 110/80–30 BC, Plate 4.12); unguentaria, possibly representing a single vessel (FV1023/1024, both c. 150–100 BC, Plate 4.5–4.6); and a near complete Grey Glaze lamp FV642, 150–100/75 BC (Plate 4.17), along with a fragment of a second lamp FV639. The remains included two coarse ware jars (FV1035, Plate 4.14; FV1037, Plate 4.16) and a lid (FV1036, Plate 4.15). The vessels recovered from pit 4021 date firmly to the late 2nd and 1st century BC.

Phased overview The results provide indications of pottery supply and consumption at the site, for perhaps some four or five hundred years of activity. Phase 1: 2nd- mid-1st century BC

In addition to the pit group, a further five sherds from a Grey Glaze bowl (FV309, 2nd to 1st century BC) were recovered from the Phase 1 soil deposit 5059.

A total of 176 sherds weighing 2520g (average sherd weight 14.3g) was recovered from Phase 1 deposits (Table 5.1), of which 20 were considered diagnostic and assigned featured vessel numbers (Appendix 4). The material in the group dates predominantly between the 2nd and 1st centuries BC, with a likely date for the deposition in cylindrical pit 4021 of the late 2nd to early 1st century BC. The group was relatively small; however, it is important in that it provides evidence for the earliest phase of activity at the site. While amphorae formed 24.7% (by weight) of the Phase 1 assemblage, they were represented by only three body sherds (of indeterminate source). Coarse and plain ware utilitarian pottery formed 17.2% and 12.7% respectively, which was the lowest relative proportions by phase of the whole Vagnari assemblage. The fine wares formed 39.4%, with ceramic oil lamps and unguentaria forming 3.7% and 2.3%, respectively.

Phase 2: early 1st century AD A total of 383 sherds weighing 5986g (average sherd weight 15.6g) was recovered from Phase 2 deposits (Table 5.1), of which 47 were considered diagnostic and assigned featured vessel numbers (Appendix 4). The Phase 2 pottery assemblage was the second largest, both by weight and count, from all stratified deposits. Amphorae formed 11.6% (Table 5.2), with coarse and plain wares forming the greatest component of the assemblage (by weight), at 55.2% and 24.9%, respectively; the fine wares formed 8.2%; and two lamp fragments were recovered forming 0.1% of the assemblage. Pottery dating to the 1st century BC was present, however, most of the assemblage dates from the 1st century AD and into the early part of the 2nd.

The fine wares, ceramic oil lamps, and unguentaria from Phase 1 formed the highest relative proportions by weight of any stratified phase (Table 5.2, Plate 4). This phenomenon was likely due, in part, to the small size of this group; however, the nature of the pottery assemblage and its deposition in pit 4021 was unusual, and not likely from the general disposal of rubbish. The presence of near complete vessels and large sherds suggests limited post-depositional disturbance (e.g. redeposited in antiquity or disturbed by subsequent ploughing of the site over millennia). The presence and combination of near complete ceramic oil lamps, the remains of two unguentaria, large fragments and near complete fine table ware vessels, two clay loom weights (see Chapter 17), and articulated animal bones (see Chapter 19) may suggest that this collection of artefacts and material was originally part of a votive deposit in the pit. No other such groups of artefacts were recovered from this programme of excavations at the vicus.

Featured Vessels

Featured Vessels

The Grey Glaze sherds present in Phase 2 were residual, that is left over from earlier activity at the site and not in use in the late 1st to 2nd century AD. While the fine wares included three sherds of Italian Sigillata and six sherds of African Red Slip ware, none of these was considered diagnostic. Several thin-walled ware sherds, dating broadly between the 1st century BC to 1st century AD, were recovered; these were drinking vessels, including the remains of two cups (FV303, Plate 5.15, and FV310, Plate 5.18), and a bowl (FV308). There were eight Regional Red Slip ware featured vessels (plates, bowls, cups, and a jar), some of which were in forms such as plates (Figure 5.1) copying those originally produced in Italian Sigillata in the 1st centuries BC and AD, while cup FV1040 (Plate 5.3) may date to the 2nd century AD. In addition, an Eastern Sigillata red slip bowl FV300 (Plate 5.13, Figure 5.2) (Hayes 1985, Type 60) was recovered, produced in the eastern Mediterranean during the 2nd century AD.

Most featured vessels were recovered from cylindrical pit 4021 (fills 4032, 4033 and 4039), consisting of predominantly fine table wares. The fine wares included

Coarse ware vessels were predominantly jars (86% of the CW group), with casseroles and clibani (portable ceramic bread baking ovens) forming 10.1% and 3.9%, 67

The Making of a Roman Imperial Estate

Figure 5.1 Regional Red Slip plate from backfilled grain storage pit 5037, Phase 2. Photo M. Carroll.

11.17, Figure 5.4), of which fragments from the vessel found in two deposits: FV1157, context 3045, Phase O, and FV1161, context 3027, Phase 6 (residual in phase). The fabrics of these vessels are very coarse with numerous volcanic inclusions, indicating production in a region with volcanic geological character, such as close to Rome or Campania (Montana et al. 2021). The large size of vessel FV780 (Plate 6.1), and the presence of a large ceramic mortarium, approximately 60cm in diameter (FV1157/1161, Plate 11.17), indicate food preparation on a scale greater than a domestic household, perhaps for communal meals taken by estate workers. In addition, the supply of these specialist vessels, and the very large storage dolia, from possibly the same source many hundreds of miles away from Vagnari, may provide evidence of a centralised supply system for key components of estate infrastructure.

Figure 5.2 Eastern Sigillata red slip bowl FV300. Photo M. Carroll.

respectively, of the coarse ware assemblage (Cubberley et al. 1988). During this phase of activity, the first ceramic imports to this part of Vagnari vicus were of African origin; all were found in the fill (3023) of drain 4054, and were cook-ware vessels (base fragments, FVs 562, 563, and 564) dating from approximately AD 70 through to the 3rd century, and possibly later.

Other featured vessels included coarse wares, predominantly jars, with several collared jars, a form that was common in the 1st and 2nd centuries AD at Vagnari (Figure 5.3). Thirteen plain ware featured vessels were present in Phase 2 deposits, including pitchers, flagons, lids, and a cup, seven of which were found in circular storage pit 4053. The foot/spike of an amphora (FV1041, Plate 5.12) was found in the fill 4060 of this pit; this vessel was produced in a regional coarse fabric (fabric code AM4), akin to Kenrick’s coarseware Sandy 1 (Kenrick 2011: 376), and it provides

An exceptionally large bowl or basin base fragment (FV780, Plate 6.1) was recovered from storage pit 6045. The vessel fabric is similar in composition to other specialist ceramic items recovered, especially the dolia (see Chapter 6, and Montana et al. 2021), and the large mortarium for mixing and grinding ingredients (Plate 68

Griffiths: 5. The Vessel Pottery and Lamps

and 34.9%, respectively; fine wares formed 11.2%. No lamps or unguentaria fragments were considered diagnostic. Featured Vessels There were few fine table wares in Phase 3; however, undiagnostic body sherds of African Red Slip ware (ARS3) and Regional Red Slip ware were present in the bulk pottery (Appendix 4). A Regional Red Slip ware cup (FV294) was recovered from a levelling deposit (5017) under the mortar floor (2012) of the cella vinaria. Two coarse ware jars were present (FVs 559 and 587), the latter, FV587 (Plate 6.7), being an unusual form with a bifid outer rim. Analysis of the fabric suggests an Albanian source (fabric code CW22, Appendix 3). An African cook-ware casserole (FV147, Hayes Form 197) and an African Black-top lid (FV148, Plate 6.6) were recovered from the mortar floor of the winery (context 2012), and both types date from the late 2nd and to end of the 3rd century AD.

Figure 5.3 Fragment of a large, collared jar FV1043 from Phase 2. Photo M. Carroll.

Phase 4: late 2nd–early 3rd century AD A total of 40 sherds weighing 801g (average sherd weight 20.2g) was recovered from Phase 4 deposits (Table 5.1), of which four were considered diagnostic and assigned featured vessel numbers (Appendix 4). As noted above, the assemblage from this phase was relatively small, and any results from statistical analyses and comparisons are limited. No amphora sherds were recovered. Coarse and plain ware utilitarian pottery formed 14.4% and 9.2% (Table 5.2) respectively, with fine wares forming 76.4%. No lamps or unguentaria were recovered. Featured Vessels The base of Regional Red Slip ware platter (FV202) was recovered from the north-south wall 5008. The vessel is a regional imitation of Italian Sigillata platters common in the late 1st century BC and through to 1st century AD; the vessel was residual in this phase. A relatively poorly finished Regional Red Slip ware small cup (FV107, Plate 7.1), with a thickly applied brown painted outer surface, was recovered from the upper fill of stone drain 1009, and may date from the 1st to 3rd centuries AD. A coarse ware lid (FV205, Plate 7.2) and the rim of a clibanus (FV204, Plate 7.3), a portable bread baking oven, were recovered from north-south wall 5008.

Figure 5.4 Plain ware bowl FV1151. Photo M. Carroll.

evidence for regional production of transport vessels. The remains of two ceramic oil lamps (FVs 210 and 626) were recovered from Phase 2 deposits, both dating broadly from the late 1st century BC to 1st century AD. Phase 3: 2nd century AD A total of 35 sherds weighing 536g (average sherd weight 15.3g) was recovered from Phase 3 deposits (Table 5.1), of which seven were considered diagnostic and assigned featured vessel numbers (Appendix 4). The Phase 3 and 4 assemblages were significantly smaller that from other phases of activity, therefore, any results from statistical analyses and comparisons are limited. Amphorae formed 15.5% (by weight) (Table 5.2), with coarse and plain ware pottery forming 38.4%

Phase 5: early 3rd century AD A total of 308 sherds weighing 2275g (average sherd weight 7.4g) were recovered from Phase 5 deposits (Table 5.1), of which 43 were considered diagnostic and assigned featured vessel numbers. Amphorae formed 15.1% (Table 5.2), with coarse and plain wares forming the greatest component of the assemblage, at 69

The Making of a Roman Imperial Estate 46.3% and 26.5%, respectively; fine wares formed 11.4%. Ceramic oil lamps and unguentaria formed 0.5% and 0.2%, respectively.

(contemporary and residual) and assigned featured vessel numbers. The Phase 6 pottery assemblage was by far the largest group, by both count and weight, from all stratified deposits. Amphorae formed 6.4% of the group (Table 5.2), with coarse and plain wares forming the greatest component of the assemblage, at 42.8% and 39.9%, respectively; fine wares formed 10.5%. Ceramic oil lamps formed 0.4%. No unguentaria fragments were recovered.

Featured Vessels The assemblage from this phase was relatively large and contained much material considered residual. The fills of three dolia defossa (contexts 3012, 3020, and 3038) were excavated, and contained sherds dating to the late 2nd and 3rd centuries AD. Dolium 3038 contained the remains of two humans, one of which was radiocarbon dated to c. AD 332–410 (see Chapter 4, Chapter 21).

Much of the Phase 6 assemblage was recovered from deposits relating to major structural changes in the portico building and the demolition or collapse of a tile roof (see Chapter 4). These events resulted in a wellstratified sequence of deposits, with the roof collapse/ demolition overlying several deposits which contained pottery mixed in with roofing materials and deposits overlying the remains of the roof. The nature of the sequence of these deposits provides good dating of the collapse/demolition, and valuable information on the range of pottery vessels in use immediately prior to, and following, these major structure changes. Following an overview of the wares present in this phase (as in previous sections), the collapse/destruction sequence assemblages are examined in detail by stratigraphic sequence (Plates 8–12). This sequence provides excellent evidence for the supply and consumption of ceramics at Vagnari in the 3rd century AD.

The fine ware assemblage contained residual sherds from Grey Glaze and Thin-Walled ware vessels, along with a Regional Red Slip ware plate (FV790, Plate 7.10), a fusiform unguentarium (FV794), and fragments of lamps, all broadly dating from the late 1st century BC to 1st century AD (e.g. FV791 and 793). Potentially contemporary Regional Red Slip ware sherds included those of a bowl (FV324), the rim of a vessel of uncertain form (FV329), and a lid (FV767); all of these vessel types date broadly from the 1st to 3rd century AD. In addition, an African Red Slip ware large bowl rim (Hayes Form 45B, FV396) was recovered, dating from approximately the second quarter of the 3rd to the middle of the 4th century AD. Coarse ware vessels included a small jar/beaker (FV774) and a bowl (FV388), the remains of three casseroles (FVs 332, 754 and 758), and a pan with handle (FV759). Jars formed the greatest number of coarse ware vessels present, along with two lids, one (FV1064) potentially dating to the 3rd or 4th century AD (an Ikäheimo D.II form). In addition, a fragment of a multi-holed strainer (FV461) was recovered from the fill of a dolium (context 3041).

The Phase 6 assemblage was the largest phased pottery assemblage from the excavations and contained much residual material, including Black Gloss and Grey Glaze wares from 2nd to 1st century BC activity, Italian Sigillata, Thin-Walled wares from the 1st century BC to the 1st century AD, the rim of a moulded-relief bowl (late 3rd to 1st century BC), an African Red Slip ware bowl (Hayes Form 8A, late 1st to mid-2nd century AD), and ‘picture lamp’ fragments dating from the late 1st century BC to the end of the 1st century AD.

Plain ware vessels included the remains of two flagons (FVs 397 and 750), three jars (FVs 460, 746 and 766; Plate 7.8), and two lids (FVs 747 and 749). These plain ware vessel types were long-lived and potentially date from the 1st to 3rd/4th centuries AD.

Amphorae Several amphora fragments from various sources were recovered from Phase 6 deposits. These included items possibly produced in the Vagnari region, many similar in composition to regional slipped, plain, and coarse wares. This group of material was relatively small and included a handle (FV437) and a body sherd (FV1004), however FV1005 (Plate 10.1) provides the best example of an amphora produced locally/regionally, with a partial rim, neck and handle preserved. Two amphora fragments (unfortunately, only relatively undiagnostic body sherds were found) of African origin were recovered in North African Lime-rich fabric (Tomber and Dore 1998: 101, NAF AM 1), including a partial handle (FV472), and a body sherd (FV466), dating broadly from the mid-2nd to 3rd centuries AD.

A small quantity of ceramic oil lamps fragments was recovered, but only one was identifiable to a form: a rim and shoulder fragment of an ‘a perline’ type (FV645), produced between the last quarter of the 1st century AD until the end of the 2nd century, but possibly still in use early in the 3rd century at Vagnari (see also De Stefano 2014: 145; De Stefano 2011: 215–18). Phase 6: second and third quarter 3rd century AD A total of 3757 sherds weighing 39506g (average sherd weight 10.5g) was recovered from Phase 6 deposits (Table 5. 1) of which 282 were considered diagnostic 70

Griffiths: 5. The Vessel Pottery and Lamps

An extensive discussion on the distribution and use of African amphorae at Vagnari and in the surrounding area between the end of the 2nd and beginning of the 7th century AD is provided by Disantarosa (2011: 392–97 and 2014: 160–65). A fragment of an amphora foot (FV595, Plate 11.16) of unknown source was also recovered. Amphorette In addition to the transport amphorae remains mentioned above, fragments of two amphorette (FV 436, Plate 10.7, and FV557) were recovered. These vessels are small and ‘amphora shaped’ and were possibly used to transport high-value products, such as perfumes, in relatively small quantities; alternatively, they may be considered novelty vessels or had ‘ritual’ functions.

Figure 5.5 Maker’s stamp on the mortarium FV1157/1161. Photo M. Carroll.

Fine wares

A large mortarium (FV1157/1161, Plate 11.17) rim fragment (weighing 987g) was recovered from the tile collapse (context 3027); the fragment joins other parts of the same vessel found in a soil deposit 3045 (Phase O) around the stump (3047) of east-west wall 4059. The mortarium was probably in use during the 1st or 2nd century AD and showed evidence of wear in its bowl. The vessel was stamped with a maker’s mark (Figure 5.5), which might read: [---]V f AV[---] above [---]t ALiS (reading courtesy of Silvia Pallecchi). The thick walls and hard fabric would have resulted in a very robust vessel for grinding ingredients, and it may have had a long use-life. This was the only ceramic mortarium recovered during this campaign of excavations at Vagnari.

Contemporary fine ware featured vessels were many, including a large quantity (15) of African Red Slip ware bowls of Hayes Forms 9B (AD 150–200), 14A (AD 175–225), 17B (AD 200–300), and 45A (AD 230–350), and dishes of Hayes Forms 16.16 (AD 175–225) and 31.4 (AD 200–250). In addition, small rim sherds each of a possible Hayes Form 50B dish, dating to the 4th and 5th centuries, a Hayes Form 59 dish (AD 300–400+), and a Hayes Form 61B dish, dating to the 5th century AD, also were recovered. Given the potentially late date of these sherds, and their small size, they were likely intrusive from a later phase of activity. Regional Red Slip wares formed the largest quantity of fine ware featured vessels (35) from this phase, some of which may have been residual; however, some of the vessel forms may be long-lived. The forms present were mostly bowls, dishes, and cups, a jar, and a handle fragment; a pyxis lid (FV1077, Plate 8.9) is also worthy of note. Many of the Regional Red Slip ware forms were also common in unslipped/undecorated plain ware fabrics.

Plain Wares A range of plain table wares was recovered, including many for serving liquids (e.g. pitchers and flagons), and for consuming food and drinks (e.g. cups and bowls/ dishes). Plain wares of utilitarian functions consisted of jars and a large quantity of lids. It was likely that many of the plain wares were local/regional in origin.

Coarse wares

Ceramic oil lamps

A total of 117 featured vessels in coarse wares was recovered, predominantly utilitarian jars and lids, and cook-wares, such as casseroles and large pans/dishes; also present was a range of bowls and dishes. Some of the cook-wares were of local/regional origin, but there were several from further afield, especially Campania, but also a range of imports from North Africa. The forms present were predominantly casseroles, along with large internally slipped dishes and pans, and cookware lids. Also recovered were the remains of three clibani (FVs 251, 263, and 779).

While many ceramic oil lamp fragments from Phase 6 deposits were clearly residual (dating from the late 1st BC to 1st c AD), many were of the ‘a perline’ type (with raised dots on the top) common in the region date from the late 1st century AD through the 2nd and possibly into the 3rd century AD (Figures 5.6 and 5.7; Appendix 5). These lamps are generally thought to have been produced until the end of the 2nd century, but there is certainly a strong possibility for production and use into the 3rd century AD. However, it is worth noting here that there was complete absence of known 3rd and 71

The Making of a Roman Imperial Estate 4th century AD ceramic oil lamp types recovered from this programme of excavations at Vagnari.

coarse wares included two dishes (FVs 732 and 738), a jar (FV737), and a lid (FV735). Fine wares included a Regional Red Slip ware dish (FV724), a plain ware bowl (FV628), and a small jar/beaker (FV741). Two ceramic oil lamp fragments were recovered (FVs 629 and 635) of ‘a perline’ type, dating from the late 1st to 2nd/3rd century AD.

Key tile collapse/demolition deposits Deposits below tile collapse, c. AD 225+ 3014 Deposit of charcoal and slag under tile collapse (Plate 10)

4027 Soil under tile collapse (Plate 10)

The pottery recovered from context 3014 was a relatively large group with 21 featured vessels. These included amphorae FVs 1004 and 1005 (Plate 10.1), and additional body sherds (see above, and Appendix 4). A well-dated group of African Red Slip fine and cookware vessels was recovered, including a single small rim sherd (11g) (FV1016) of uncertain form, along with two casseroles, FV1012 (Hayes Form 23B, c. AD 225/250–325) and FV1015 (Hayes Form 197, c. AD 175–300), and a lid (FV1013). In addition, Regional Red Slip and plain ware bowls/dishes, a cup, and a beaker, were present. Plain and coarse utilitarian wares included a plain ware basin (FV1000, Plate 10.2), and numerous lids, along with a coarse ware collared jar (FV1006). No ceramic oil lamp fragments were recovered.

A total of six featured vessels was recovered from this deposit. The deposit included a Regional Red Slip ware bowl (FV1073) and cup (FV1074), both dating from the second half of the 2nd century and into the 3rd century AD. Other items included three coarse ware lids (FV 1079, Plate 10.8; FV 1080, Plate 10.9; and FV1081, Plate 10.11), and a plain ware lid (FV1129, Plate 10.10). 4028 Soil under tile collapse (Plate 10) A total of 15 featured vessels was recovered from this deposit. The deposit included African cook-wares in the form of a casserole (FV1066, Hayes Form 197, AD 175–300) and lid (FV1063). Numerous coarse and plain ware lids were also present. Table wares included a Regional Red Slip ware cup (FV1075), a coarse ware dish (FV1098), two bowls (FV1096, Plate 10.15, and FV1097, Plate 10.16), a plain ware cup (FV1121, Plate 10.14) and two flagons (FV1124, Plate 10.12; FV1125, Plate 10.13).

3040 Soil under tile collapse (Plate 10) This was a relatively large group with 24 featured vessels. The deposit included an amphora handle (FV437) and amphoretta foot fragment (FV436, Plate 10.7). As with context 3014, the African vessels provide the most robust dating evidence: an African Red Slip ware bowl (FV450, Hayes Form 45A, c. AD 230–350) was recovered, along with casseroles FV449 (Hayes Form 197, c. AD 175–300), and FVs 451, 452, and 454 (Hayes Form 23B, c. AD 225/250–325), and a cook-ware lid (FV448). Italian Sigillata and thin-walled wares were present, but all were residual in this phase. Other fine wares included a Regional Red Slip ware cup (FV456, Plate 10.5), dating to the 2nd or possibly early 3rd century AD, and a plain ware bowl/dish (FV435, Plate 10.4) and a flagon (FV441, Plate 10.3). Coarse and plain utilitarian wares included a wide-mouthed jar (FV447) and two lids (FVs 444 and 446), along with plain ware jars (FV433 and FV442) and lids (FV432, Plate 10.6, and FV435, Plate 10.4).

4030 Soil under tile collapse (Plate 8) This was a large deposit from which 42 featured vessels were recovered. Fine wares included an African Red Slip ware bowl FV1054 (Plate 8.7) of Hayes Form 17B, AD 200–300, and two Regional Red Slip ware bowls (FV 1076, Plate 8.8, and FV1078, Plate 8.10), and a pyxislid (FV1077, Plate 10.9). Also present were plain ware flagons (FV1126, Plate 8.1, and FV1128, Plate 8.2), a cup (FV1127, Plate 8.5), and a bowl (FV1151, Figure 5.4 and Plate 8.6). In addition, there were numerous coarse ware bowls (FV 1102, Plate 8.14, and FV 1104, Plate 8.16), and dishes (FV 1103, Plate 8.15, FV 1105, Plate 8.17, FV1106, Plate 8.18, and FV 1107, Plate 8.19), possibly for use at table. African cook-wares were present and provide robust dates for this deposit; these include casseroles (FVs 1068 and 1069, Plate 8.13) of Hayes Form 23B, AD 225/250–300, a pan (FV1065, Plate 8.12, Ikäheimo G, AD 200–300), and a lid (FV1062, Plate 9.5, Ikäheimo B.I.). Utilitarian wares included a widemouthed coarse ware jar (FV1112, Plate 8.24), other jars (FV1110, Plate 8.22; FV1111, Plate 8.23; FV1117, Plate 8.20; and FV1120, Plate 8.21), and numerous coarse and plain ware lids. Two small ceramic oil lamp fragments of ‘a perline’ type were recovered (FVs 632 and 633), dating broadly from the late 1st and through the 2nd century AD, but possibly still in use during this

4024 Soil under tile collapse A total of 14 featured vessels was recovered from this deposit. The deposit included African cook-ware casseroles FV739 and FV 45 (Hayes Form 23B, c. AD 225/250–325), along with a dish with an internal slipped (non-stick) coating (FV736). Other cook-wares included a casserole in a form common at Vagnari (FV731), and a local/regional imitation (FV739) of the African cookware Hayes Form 23B, in coarse ware fabric CW13. Other 72

Griffiths: 5. The Vessel Pottery and Lamps

Figure 5.6 Lamp fragments from Phases 2, 5, 6, and 7. Photo D.R. Griffiths and H. Russ.

73

The Making of a Roman Imperial Estate

Figure 5.7 Lamp fragments from Phase O. Photo D.R. Griffiths and H. Russ.

74

Griffiths: 5. The Vessel Pottery and Lamps

phase. A residual, early (c. 25 BC–AD 100) mould-made ceramic oil lamp (FV641) also was present.

common at the site. In addition, two local/regional coarse ware vessels (FVs 91 and 93) of uncertain form were also recovered, as well as a red slipped ceramic oil lamp (FV104), dating from the mid-1st to 2nd century AD.

Residual fine wares included a Black Gloss cup (FV1070, Plate 8.3), an Italian Sigillata cup (FV1071, Plate 8.4), and an Eastern Sigillata ware dish (FV1061, Plate 8.11, of Hayes 1985, Form 60, c. AD 100–150; see also FV300 (Figure 5.2, Plate 5.13), context 5042, the fill of the construction cut for wall 5003, Phase 2).

Residual featured vessels included a Black Gloss bowl/ dish base (FV95), two Grey Glaze bowls (FV96, Plate 11.4, and FV103, Plate 11.3), and an African Red Slip ware bowl (FV99, Hayes Form 9, AD 100–200).

5004 Soil under tile collapse (Plate 10)

3027 Tile collapse (Plate 11)

This was a relatively small group from which 11 featured vessels were recovered. The deposit included a Regional Red Slip ware cup (FV193, Plate 10.25) and plain ware cup (FV607, Plate 10.24). Utilitarian wares included a coarse ware collared jar (FV200), two lids, and up to five plain ware lids. Two sherds of Grey Glaze ware (FV194) were present in this group, but they were residual. Tile collapse

Only two pottery fragments were recovered from this deposit: an amphora foot (FV595, Plate 11.16) of unknown source and a large ceramic mortarium (FV1161, Plate 11.17) with a maker’s stamp (Figure 5.5). The latter likely was residual in this phase, as these types date broadly to the 1st century AD. Additional sherds from this mortarium join with others (FV1157) found in context 3045, Phase O.

1004 Tile collapse (Plate 11)

Deposits above tile collapse

This deposit contained the bulk of the tile remains from the collapse/demolition event. A total of 27 featured vessels was recovered from this deposit. While there were only a few well-dated 3rd-century AD vessels, deposition of the assemblage must have taken place after c. AD 225+, as the tile collapse overlays deposits material dating later than this.

3022 Soil on tile collapse A total of 22 featured vessels were recovered from this deposit. Fine table wares included two sherds of African Red Slip ware, FVs 558 and 581 (Hayes Form 31.4, c. AD 200–250), possibly from the same vessel; a Regional Red Slip ware bowl or dish (FV582), and a vessel of uncertain form (FV583). Plain ware vessels included a cup (FV605), a flagon (FV588), a lid (FV585, Plate 12.5), and fragments of a handle (FV586). The plain ware cup (FV605) is of particular interest, as it was over-fired during its production, perhaps at Vagnari, as the fabric resembled the greenish clay found in the ravine at Gravina (Montana et al. 2021); it was likely a ‘second’, and not particularly visually impressive, but functioned perfectly well as a ‘cup’ (Figure 5.8). Utilitarian coarse wares included a collared jar (FV589), two lid-seated jars (FVs 590 and 592), two other jars (FVs 591 and 593), and a lid (FV556, Plate 12.6). A small fragment (43g) of a possible amphoretta foot (FV557) also was recovered.

A quantity of Regional Red Slip table wares (possibly 1st/2nd to 3rd century AD) were recovered, including two cups (FV97, Plate 11.6, and FV100, Plate 11.5), two bowls (FV 83, Plate 11.8, and FV 98), a dish (FV101, Plate 11.7), and a flagon (FV82, Plate 11.1). A relatively large jar (FV84, Plate 11.9), with a wide-mouth and a band of incised linear decoration around the upper body and all-over brown-red painted decoration, was recovered. Plain ware vessels included two jars (FV 85, Plate 11.11, and FV86, Plate 11.10), a pitcher (FV87, Plate 11.2), three lids (FVs 79–81, Plate 11.12–13), and two handles (FVs 88 and 90), the latter relatively large, perhaps from a pitcher or a small table amphora. Two coarse ware jars in Butrint Ware (produced in modern Albania across the Adriatic Sea) were recovered from this deposit; the first, FV92 (Plate 11.15), is a relatively plain form with a triangular rim (see a similar example from Butrint (Reynolds 2020: 263, table B.6 and 259, fig. B.3.1, no. 530.64), while the second, FV94 (Plate 11.14), has a deep lid seat and linear bands of incised decoration to the upper body, similar to other examples from Butrint (Reynolds 2020: 263, table B.6; 259, fig. 3.3; 237, fig. A.3.j). The fabric (CW22, see full description in Appendix 3) of these two vessels corresponds closely with Kenrick’s Vagnari fabric Sandy 8 (Kenrick 2014: 127), however, the forms of FV92 and FV94 are not

Three items were retrieved from this deposit which were residual, including a Black Gloss lid (FV584, Plate 12.4), a Thin-Walled ware body sherd (FV580), and a ceramic oil lamp fragment (FV647). 4004 Soil on tile collapse and wall 4006 While the bulk pottery assemblage recovered from this deposit was relatively large (344 sherds weighing 2.4kg), only two items were considered as featured vessels; these were two ‘a perline’ type lamps (FVs 630 and 634), dating from the late 1st to the end of the 2nd century AD. However, the presence of a 4th-century 75

The Making of a Roman Imperial Estate vessel function based on fabric and form characteristics (e.g. Types), for example, some types were obviously primarily for cooking, such as casseroles and large pans/dishes with internal red slip coatings — often with evidence of heavy burning to their outer body; others, such as the wide range of jar types may have had a variety of functions, including cooking, but might also have been used for storage and perhaps transportation (see Table 5.3 for the range of types within each functional category and their relative proportions within each phase). The functional categories in this analysis are relatively broad; the complete dataset is presented in Appendix 4 (featured vessels). This data has the potential to be used in different ways to address specific questions by future scholars. The range of vessel types and their relative proportions (by estimated vessel equivalents -EVEs) were combined in Table 5.3. The main categories are: dining, those vessels for use ‘at the table’ for presenting/serving (e.g. flagons and pitchers, large platters), and those for consuming food (e.g. plates and bowls) and drink (e.g. cups and beakers); cooking, e.g. those vessels used in direct contact with heat to cook food (e.g. clibani, large shallow pans/dishes with non-stick internal red slip coating, and casseroles); utilitarian, a catch-all category for many items which may have had multiple functions in antiquity and may not fit solely in one of the categories here, such as jars, and vessels for preparing food, e.g. mortaria and large bowls. Devices for lighting (ceramic oil lamps) and unguentaria (vessels for perfumes and cosmetics) are assigned individual categories. Transport amphorae and coarse and plain ware lids were excluded. Those vessels clearly residual in each phase were excluded from functional analysis, as their presence in deposits from ‘later’ phases does not represent contemporary activity; they were produced, used, and subsequently broken at a much earlier in date. These are noted as ‘residual’ in Appendix 4.

Figure 5.8 Misfired plain ware cup FV605. Photo M. Carroll.

coin (Constantine or Constantine sons, see Chapter 8) suggests these lamps were residual. Phases 7 (heavily disturbed deposits) and O (out-of-phase material) Due to issues of deposition and post-depositional disturbance, material from topsoil deposits allocated to Phase 7 and from stratigraphic uncertain deposits of Phase O were not subject to detailed analyses. However, a full catalogue of featured vessels is presented in Appendix 4, with selected illustrations for Phase 7 and Phase O presented in Plates 15–16 and Plates 17–19, respectively. A total of 2235 sherds weighing 20459g (average sherd weight 9.1g) was recovered from Phase 7 deposits (Table 5.1), of which 94 were considered diagnostic and allocated featured vessel numbers. The pottery recovered from Phase 7 could be assigned to the date range from the 1st century BC through to the 4th and/ or possibly 5th century AD.

Phases 3 and 4 were not comparable, due to the small pottery assemblages recovered; also, material from Phase 7 was not considered due to the heavily disturbed nature of their deposits; and Phase O vessels were not considered due to uncertainly stratigraphic relationships and, therefore, phase of activity. However, their relative proportions are included in Table 5.3 for completeness.

A total of 1658 sherds weighing 15311g (average sherd weight 9.2g) was recovered from Phase O deposits (Table 5.1), of which 172 were considered diagnostic and assigned featured vessel numbers.

Phase 1: 2nd- mid-1st century BC: n = 483.5 EVEs

Functional Analysis by Estimated Vessel Equivalent EVES (Tables 5.3 and 5.4)

As outlined above (Phase 1 discussion), the pottery assemblage from Phase 1 was unusual, almost all of it from two fills in cylindrical pit 4021. The assemblage probably was not formed through ‘normal’ deposition, i.e. as rubbish disposal of already broken vessels; most items likely stem from a primary deposit such as a votive offering (see Chapter 4 and Chapter 19). The high

In this section, the securely phased pottery assemblages have been analysed according to vessel function (based on vessel form and fabric characteristics) specifically, to identify any potential changes over time of the relative proportions of the various functional classes. While there are many potential issues with assigning 76

Griffiths: 5. The Vessel Pottery and Lamps

Phase

Function/form Dining (table wares) Beaker

1

2

3

4

5

6

7

0

Total

26.9%

59.1%

0.0%

40.0%

32.3%

53.3%

21.8%

35.5%

43.2%

0.0%

0.0%

0.0%

0.0%

0.0%

1.1%

0.0%

2.4%

1.1%

26.9%

12.9%

0.0%

0.0%

28.7%

43.9%

40.0%

36.4%

37.3%

0.0%

0.0%

0.0%

0.0%

0.0%

5.7%

12.6%

1.1%

4.0%

Cup

53.8%

37.4%

0.0%

100.0%

0.0%

23.9%

0.0%

12.7%

23.5%

Dish

0.0%

0.0%

0.0%

0.0%

0.0%

9.9%

0.0%

6.2%

6.8%

Flagon

7.7%

24.1%

0.0%

0.0%

57.4%

12.3%

0.0%

18.2%

15.2%

Jar

0.0%

2.1%

0.0%

0.0%

13.9%

0.3%

7.9%

2.1%

1.5%

Pitcher

0.0%

20.2%

0.0%

0.0%

0.0%

2.3%

39.5%

12.8%

7.7%

Bowl Bowl/dish

11.5%

3.3%

0.0%

0.0%

0.0%

0.0%

0.0%

1.5%

1.3%

Pyxis-lid

Plate

0.0%

0.0%

0.0%

0.0%

0.0%

0.7%

0.0%

0.0%

0.4%

Uncertain

0.0%

0.0%

0.0%

0.0%

0.0%

0.0%

0.0%

6.7%

1.3%

0.0%

5.3%

72.3%

60.0%

32.0%

13.9%

1.6%

10.8%

11.2%

Cooking Casserole

0.0%

72.1%

25.5%

0.0%

100.0%

58.3%

0.0%

53.6%

59.4%

Clibinus

0.0%

27.9%

0.0%

100.0%

0.0%

5.1%

0.0%

32.8%

12.6%

Dish

0.0%

0.0%

0.0%

0.0%

0.0%

12.3%

0.0%

12.0%

9.6%

Jar

0.0%

0.0%

74.5%

0.0%

0.0%

21.3%

0.0%

0.0%

15.1%

Pan

0.0%

0.0%

0.0%

0.0%

0.0%

3.0%

100.0%

1.6%

3.3%

Utilitarian

24.0%

32.6%

27.7%

0.0%

35.7%

29.5%

76.6%

36.1%

35.0%

Bowl

0.0%

0.0%

0.0%

0.0%

0.0%

0.0%

0.0%

4.6%

1.1%

100.0%

100.0%

100.0%

0.0%

100.0%

93.6%

55.8%

78.3%

83.9%

Mortarium

0.0%

0.0%

0.0%

0.0%

0.0%

1.5%

0.0%

10.3%

3.0%

Uncertain

0.0%

0.0%

0.0%

0.0%

0.0%

4.9%

44.2%

6.8%

11.9%

Lighting

28.4%

3.1%

0.0%

0.0%

0.0%

3.3%

0.0%

17.5%

8.6%

Lamp

100.0%

100.0%

0.0%

0.0%

0.0%

100.0%

0.0%

100.0%

100.0%

Jar

Unguentaria

20.7%

0.0%

0.0%

0.0%

0.0%

0.0%

0.0%

0.0%

2.0%

100.0%

0.0%

0.0%

0.0%

0.0%

0.0%

0.0%

0.0%

100.0%

Total

100.0%

100.0%

100.0%

100.0%

100.0%

100.0%

100.0%

100.0%

100.0%

N=

483.5

407

32.5

12.5

189

2285

436

1160.5

5006

Unguentaria

Table 5.3 Functional analysis by category and form using relative proportions (based on estimate vessel equivalents -EVEs- of featured vessels).

Phase

Function

1

2

3

4

5

6

7

0

Total

26.9%

59.1%

0.0%

40.0%

32.3%

53.3%

21.8%

35.5%

43.2%

0.0%

5.3%

72.3%

60.0%

32.0%

13.9%

1.6%

10.8%

11.2%

Utilitarian

24.0%

32.6%

27.7%

0.0%

35.7%

29.5%

76.6%

36.1%

35.0%

Lighting

28.4%

3.1%

0.0%

0.0%

0.0%

3.3%

0.0%

17.5%

8.6%

Unguentaria

20.7%

0.0%

0.0%

0.0%

0.0%

0.0%

0.0%

0.0%

2.0%

Total

100.0%

100.0%

100.0%

100.0%

100.0%

100.0%

100.0%

100.0%

100.0%

N=

483.5

407

32.5

12.5

189

2285

436

1160.5

5006

Dining (table wares) Cooking

Table 5.4 Functional analysis using relative proportions (based on estimate vessel equivalents -EVEs- of featured vessels).

77

The Making of a Roman Imperial Estate relative proportions (Tables 5.3 and 5.4) of lighting equipment (28.4%) and unguentaria (20.7%) in this assemblage result from the complete or near-complete vessels present, and clearly highlights that this artefact group was different from any other from in all phases. Fine table wares formed 26.9% of the assemblage, however, many of these were, like the ceramic oil lamps and unguentaria, large fragments or near-complete vessels. Utilitarian vessels formed 24%; no vessels recovered were used specifically for cooking. Fine table wares, lighting devices, and unguentaria combined make up 76% of the entire group from Phase 1.

wares to 32.3%, much lower than 59.1% in Phase 2. The relative proportion of utilitarian wares reduced to 35.7%, down from 32.6% in Phase 2. No ceramic oil lamps or unguentaria rims were present. Phase 6: second and third quarter 3rd century AD: n = 2285 EVEs The relative proportion of cook-wares were greatly reduced, down from 32.0% in Phase 5 to 13.9%. The relative proportions of fine table wares increased substantially from 32.3% in Phase 5 to 53.3%, comparable with the Phase 2 assemblage. Utilitarian wares were slightly reduced from 35.7% to 29.5%. Part of the single mortarium recovered was found a deposit from this phase, however, as noted earlier, this may be residual from an earlier phase. Lighting equipment formed 3.3%, but it is possible that some of these lamps were residual.

Phase 2: early 1st century AD: n = 407 EVEs The nature of the composition and deposition of the Phase 2 pottery assemblage and the assemblages from subsequent phases (with the exceptions of Phases 4 and 5, 7, and O, whose comparative limitations have been outlined above) allows for detailed functional comparisons.

Functional Analysis: Discussion

Vessels for cooking formed 5.3% of the assemblage, a relatively low proportion when compared with later phases (Tables 5.3 and 5.4). Fine table wares were the highest relative proportion, at 59.1%, the greatest from any phase (but only marginally higher than in Phase 6, which has 53.3%). While the utilitarian wares were greater than from Phase 1, with 32.6% compared to 24.0%, the group compares well with all other phases (except Phases 3 and 4). A large coarse ware vessel base (FV780, Plate 6.1) from a bowl or basin was recovered from Phase 2, but because no part of its rim was preserved, it was not included in this analysis. However, it was notable due to its large size, and certainly had a more specialised function; its primary use may have been outside of a domestic setting, and, as with the mortarium (FV1157/1161, Plate 11.17), it may have been more suited to preparing food on a large scale, perhaps for estate workers. Vessels for lighting formed 3.1%, comparable with Phase 6 (however, many ceramic oil lamps from Phase 6 may be residual). No unguentaria rims were present.

Within the utilitarian vessels were few items clearly intended for food preparation, such as the large vessel FV780 bowl/basin in Phase 2, and the mortarium (FV1157/1161) in Phases 6 and O. Vessels for heating or cooking food were completely absent in the Phase 1 assemblage, and only formed 5.3% in Phase 2. However, by Phase 5, the relative proportions of cook-wares increased substantially to 32.0%, but reduced again in Phase 6, to 13.9%. In contrast, the relative proportion of fine table wares reduced greatly from 59.1% in Phase 2 to 32.3% in Phase 5, it but increased again later in Phase 6 to 53.3%.

Phases 3 and 4: 2nd to late 2nd-early 3rd century AD: n = 32.5 and 12.5 EVEs respectively

The relative proportions of vessels for certain functions, and the changes which took place over perhaps a 500year period (Tables 5.3 and 5.4), may imply that there were also changes in the ‘nature’ of occupation in this part of Vagnari vicus. When considered alongside other evidence, especially the structural remains and features excavated, the pottery may reflect changes in the use of structures and space, from, perhaps, votive (e.g. Phase 1) and early domestic activity (possibly a wellappointed house) in the earliest phases, with a change to an area of communal activity, with some evidence for large scale food preparation, perhaps for those working on the estate.

Ceramic vessels for lighting were almost completely absent in phases later than Phase 2 (3.1% in Phase 6); this phenomenon may also reflect a change in this part of the site, from some domestic activity which required lighting after-dark to a less domestic task. After Phase 1, the relative proportion of vessels of more utilitarian functions remain relatively consistent throughout the later activity at the site.

Due to the small size of the assemblages from Phase 3 (32.5 EVEs) and Phase 4 (12.5 EVEs), comparison of functional analysis would not be statistically robust. However, for completeness, data is presented in Tables 5.3 and 5.4. Phase 5: early 3rd century AD: n = 189.0 EVEs Cook-wares formed 32.0% of the assemblage, a significant increase from 5.3% in Phase 2. There was a reduction in the relative proportions of fine table 78

Griffiths: 5. The Vessel Pottery and Lamps

However, one must caution inferences of human activity based on the pottery data. While there was evidence for changes over time in the functional composition of the pottery assemblage, can one infer social and economic activity (and possibly status) solely from the use of pottery, and/or identify areas of function (domestic, industrial etc.) and changes in function over time? Any conclusions drawn from this type of analysis are tentative, but there were changes in the composition of types of pottery vessels use at the vicus over many hundreds of years. The results of this data also raise a number of questions: what was happening in Phase 5 to see such a dramatic shift in the proportions of cookwares and fine table wares? Was this pattern ‘real’ or were there issues of residuality within the pottery assemblage which influence the data? Are these patterns reflected in the certain classes over time, for example, specialist vessels and the supply of fine table wares? Also, the size of assemblages may have influenced this data, and the later date of the Phase 6 assemblage certainly includes a greater proportion of residual sherds (not necessarily identified during analysis and excluded from the data).

Nationally distributed pottery Nationally distributed ceramics were present in relatively large quantities and from a wide range of sources. These wares included fine table wares (e.g. Italian sigillata, possibly from northern Italy), and black glazed ware (fabric BG2) and the remains of a mouldedrelief bowl (fabric MR2, ‘Megarian-type’), both of which may have been produced at Metaponto. Relatively large quantities of pottery were produced in Campania, including black glazed fine wares (fabric BG1, Campana C) and thin-walled wares (fabric TW6), plain wares (fabrics PW12 and 13), coarse wares (fabrics CWs 16 to 21), and possibly the mortarium (fabric MO1), ceramic oil lamps (fabric LA9), and unguentaria (fabric UG4). In addition, research on the provenance of the dolia at Vagnari, highlights that these large, specialist vessels were produced from clays found over a wide volcanic region which includes most of Latium, from the Vulsini Mountains and Bolsena Lake areas and southwards towards Rome, and further south to the mouth of the Garigliano River on the border with northern Campania (Chapter 6; Montana et al. 2021: 6). In addition to these nationally traded products, numerous amphora fragments were recovered from as yet unidentified sources within Italy (and overseas, see below).

The range of ceramics recovered from this programme of excavations correspond closely with those recovered from previous investigations at the site (e.g. Disantarosa 2011; Kenrick 2011 and 2014; De Stefano 2011 and 2014). While there was material dating from the 2nd/1st century BC to possibly the 5th century AD, the bulk of the material was recovered from securely stratified deposits related to activity in Period 6, the 3rd century AD and later (see Table 5.1).

Overseas imports In addition to regional and nationally distributed wares, the inhabitants of the site also consumed pottery and amphora-borne commodities from overseas. While Disantarosa (2011, 2014) previously identified relatively large quantities of overseas amphorae from a wide range of sources at Vagnari, few were recovered during the most recent work at the site. Overseas amphora fragments included body sherds from North Africa and a rim and neck fragment of a Beltrán II B amphora of Spanish origin (FV120).

Sources of Pottery and Conclusions Due to time and practical constraints brought about by the coronavirus epidemic of 2020–21, it was not possible in the preparation of this chapter to directly and statistically compare this assemblage with those recovered from previous investigations in different parts of the settlement at Vagnari (C. Small 2011; Kenrick 2011; Kenrick 2014), the nearby cemetery, or the villa at San Felice, and other sites in the region.

In addition to the amphorae, there was a range of fine and coarse wares from North Africa, including coarse ware pans/dishes, casseroles and lids, and fine table wares, including bowls and dishes. Relatively few fragments of pottery from overseas were recovered from the earlier phases of activity at the site but they included four base sherds of African cook-ware (FVs 562, 563, and 564) from Phase 2, a small undiagnostic body sherd of fine ware (fabric ARS2), and a fragment of a casserole (FV147, Hayes Form 197, c. AD 175–300) from Phase 3. Amphora body sherds of African origin were recovered from Phase 4 deposits. Phase 5 deposits contained single sherds each of a large fine ware bowl (Hayes Form 45B, c. AD 230–350) and a cook-ware lid (possibly Ikäheimo D.II., c. AD 150–250). The greatest quantity of featured vessels (39) of African origin was recovered from Phase 6 deposits (see catalogue and

It is nevertheless clear that the sources of the pottery consumed at Vagnari vicus range widely, with the bulk of the material recovered consisting of regionally produced pottery in various fabrics, including coarse, plain, and fine wares. As previously noted by Kenrick (2014: 139), vessel types present at Vagnari have parallels at Ordona, located c. 100km to the northwest (Annese 2000; De Stefano 2008); this is also true of the assemblage presented here. Examples of fine wares from the region include grey and black glazed wares and a single moulded-relief bowl.

79

The Making of a Roman Imperial Estate Appendix 4 for details). While some (FVs 99, 1017, 1052, and 1053) were produced and in use prior to the 3rd century AD, for example the Hayes Form 8A, 9, and 9B bowls, the majority date to the 3rd to 4th centuries AD. A total of five featured vessels was recovered from Phase 7 and eight featured vessels belonged to Phase O, which included vessels of late 2nd- to early 3rd-century date (Hayes Form 14A, FV154, from context 5015 and FV 812 and 834, from context 6039); other vessels had a date range from the late 2nd to 4th centuries AD.

produced there. However, the presence of transport vessels from outside of the region, such as elsewhere in Italy, Southern Spain, and North Africa (and other sources identified by Disantarosa (2011; 2014), suggests that certain goods which were not necessarily produced at the estate were desired and consumed by those living at Vagnari and imported as required. Ceramic oil lamps The ceramic oil lamp assemblage consisted of 79 fragments (weighing 356 grams), forming 0.4% (by weight) of the overall pottery assemblage (Figures 5.6 and 5.7). The vessels date from as early as the mid2nd to early 1st century BC up to the end of the 2nd or possibly early 3rd centuries AD (Appendix 4). A near complete Grey Glaze lamp (FV642, Plate 4.17) was recovered from Phase 1 (and a second fragment in the same fabric, LA8 (Appendix 3), and similar examples have been found at Gravina (Prag 1992: 216, no. 1686, Type IV, Period VIIIa, late 2nd to 1st BC) and Ordona (Delplace 1974: type IB2a, mid-2nd to early 1st century BC). At least thirteen examples of ‘a perline´ type were recovered; this a globular type common at Vagnari (and elsewhere), both in the vicus and the cemetery (De Stefano 2014), and throughout the region and the Italian peninsula from the late 1st century AD until the end of the 2nd/early 3rd centuries AD. It was not possible to identify many of the lamp fragments accurately to type due to the absence of diagnostic features, particularly nozzles; however, the majority of the partial body fragments corresponded to Loeschke (1919) shoulder forms for globular lamps of the late 1st century BC through to the 2nd century AD (Appendix 4).

Two featured vessels, from unknown sources in the Eastern Mediterranean (Eastern Sigillata), were recovered from Phase 2 (FV300, Plate 5.13, Figure 5.2) and Phase 6 (FV1061, Plate 8.11) deposits; both were Hayes form 60. A total of six featured vessels made across the Adriatic Sea were recovered (allocated fabric code CWOX18 ‘Butrint Ware’). The earliest example was FV587 (Plate 6.7) from a Phase 3 deposit, followed by FVs 92 (Plate 11.15), 94 (Plate 11.14), and 237 from Phase 6 deposits; FVs 408 and 464 were found in out-of-phase deposits. The fabric is distinct (see CW22, Appendix 1), with abundant coarse and angular calcite inclusions, often visible on the surface of the vessel. The vessels were storage jars and/or cook-wares and included a single lid (FV408) and handle fragment (FV464). Specialist vessels While the categorisation of specialist vessels is subjective, the use of the term here refers specifically to transport amphorae, ceramic oil lamps, clibani, and the single mortarium (FV1157/1161, Plate 11.17), and other overly large vessels, for example FV780 (Plate 6.1) the large size of which may indicate use outside of the domestic environment (see also dolia in Chapter 6).

What was significant in this assemblage was the absence of any ceramic oil lamps securely dating to the 3rd century AD and later. These were found in earlier excavations in 5th-century contexts at Vagnari (Favia et al. 2011a: 209, 214, fig. 5.114, P1190; 209, P1182; 186, P896; 188, P891 and P892; Hayes 1972, type II, mid-5th to 6th century AD) and elsewhere in the region (De Stefano 2014). This absence of later Roman-period lamps is unusual considering the presence of large quantities of 3rd to 4th century AD pottery present in Phase 6 and later deposits. One reason may be that ‘function’ in this part of the vicus was not directly related to domestic activity, where lighting equipment was predominantly used once the sun had set.

Amphorae The amphorae assemblage was relatively small (145 sherds weighing 9019 grams, forming 10.3% (by weight) of the overall pottery assemblage). However, 615 sherds of amphorae (from multiple sources), were recovered from previous investigations, both fieldwalking and excavation (between 1998 and 2004) at Vagnari (Disantarosa 2011: 387). The assemblage presented by Disantarosa (2011: 387–406) represents perhaps the range from the 4th century BC to the 7th century AD. The relatively small size of the amphorae assemblage recovered during this programme of excavations may be the result of the organisation of rubbish disposal at the site, or it may be that different activities were taking place in this part of the vicus at different times. Given the nature of the settlement at Vagnari, a productive rural estate, one may assume that many transport amphorae were used for the export of commodities

Mortaria Only one mortarium (FV1157/1161, Plate 11.17) was recovered, fragments of which were found in context 3027 (Phase 6) and context 3045 (uncertain phase). Ceramic mortaria are rare at the site, as much of the mixing and grinding was done in stone rather than 80

Griffiths: 5. The Vessel Pottery and Lamps

ceramic mortars (see Chapter 14). The single mortarium was relatively large, approximately 60cm in diameter, a size that would allow large quantities of ingredients to be prepared, possibly for greater numbers of people than in a domestic setting. The other significant feature of this vessel is the clay from which it was formed: the fabric is very similar to that of the imported dolia (Montana et al. 2021; and Chapter 6) made of a very coarse fabric with large volcanic inclusions and produced outside Apulia (see comments above).

coarse ware bulk assemblage. A clibanus was a small, portable oven generally believed to have been used for baking bread (see Cubberly et al. 1988 for a full discussion of this vessel type). Cubberley et al. (1988: 98) suggest that these small portable ovens were often found at rural and urban sites, and perhaps used by many levels of society in Roman Italy. The vessels considered here were formed in a range of fabrics (CW3, 6, 7, 8, 11, 13 and 16), many of which were likely produced in the region (see fabrics CW3, 6, 7, 8 and 11, all belonging to fabric Group CWOX1); however, fabrics CW13 (with clear ‘glassy’ volcanic(?) inclusions), similar to Kenrick’s Sandy 4 (Kenrick 2011: 376), and CW16 (with black and green-tinged volcanic inclusions) were produced outside of the region, likely in Campania and/ or possibly another ‘volcanic’ region on the west side of the Italian peninsula.

Large vessel of uncertain function In addition to the mortarium mention above, a second large coarse ware vessel was recovered that may have had a specialist function (FV780, Phase 2; Appendix 4). This was a large bowl or dish, with a small bead footring with a diameter of 18cm and a body diameter of at least 45cm, although it was clear the vessel was originally much larger (Plate 6.1). The fabric (CW19, Appendix 3) was very coarse, with abundant black and green-tinged volcanic inclusions (very similar to the mortarium), indicating it was produced in a volcanic region of Italy. These fabrics are common in Campania (see fabrics CW16–21, Appendix 3). The inside surface was very smooth and may have been used for mixing or grinding ingredients to a fine paste.

Overall, then, the pottery assemblage considered in this chapter highlights the complex nature of supply and consumption of this important class of material to the imperial estate over perhaps 500 years. As discussed above, the bulk of the material was produced in the region, however, the wide-range of imports, from Italy and overseas, emphasises the complexities and long-distance connections for the transport of goods (both amphora-borne commodities and pottery vessels themselves) throughout the Roman empire. The pottery assemblage presented in this chapter adds to the already substantial dataset of material from this imperial estate.

Clibani Thirteen clibani featured vessels were recovered (Appendix 4), with many body sherds present in the

81

CHAPTER 6

The Dolia Defossa and Viticulture at Vagnari Maureen Carroll, Giuseppe Montana, Luciana Randazzo, Donatella Barca, and Benjamin Stern

Figure 6.1 Roman funerary relief (from Rome?) depicting a man and woman (left) on the premises of their vineyard and winery full of dolia. Photo M. Carroll, courtesy of National Museums Liverpool.

with a flat base of 26cm in diameter, still stood in its circular unlined pit in the area of the later Room N (Figure 4.16). These two earlier dolia, however, were probably grain storage vessels, unlike the dolia defossa in the cella vinaria of the 2nd century AD which contained wine. It is on these latter that we concentrate here. Two other fragments of dolia, a rim and a flat lid with holes, both made of reddish clay, were recovered in deposits that cannot be dated (Phase O) and we cannot be sure where they originally were used (Plate 20.1, 20.3).

The dolia defossa The dolia defossa in the cella vinaria, —the winery—, were buried up to their necks in the ground, just as Roman agrarians recommended, thereby keeping the temperature of the wine constant and cool, a necessary measure in hot climate zones (Pliny the Elder, Natural History 14.27). At Vagnari, the circular depressions dug to accommodate the dolia were lined with smoothed mortar which might have helped to protect them from any extremes in temperature. This system of lining an excavated circular depression with mortar in order to place a dolium in it has an older pedigree at Vagnari. One such truncated mortar basin (4020) with a flat base 24cm in diameter (reflecting the size of the base of the now missing dolium) was found in the Room K area, and it dates to the 1st century AD (Phase 2) (Figure 4.17). Another ceramic dolium of the same phase (6061), also

Dolia defossa had a flat ceramic lid (operculum) that rested on the vessel’s flat rim and covered the opening, but only one fragmentary lid has survived at Vagnari (Cat. 6.6, Figure 6.4, Plate 20.2). A 2nd-century AD funerary relief in Liverpool showing a man and woman clasping hands in front of their winery indicates, however, that sometimes also slatted wooden lids could 82

Carroll et al.: 6. The Dolia Defossa and Viticulture at Vagnari

cover the mouths of dolia (Angelicoussis 2009) (Figures 6.1 and 6.2). There often was a larger domed ceramic lid (tectarium) placed on top of the operculum to further protect the contents from rain or sun, although these rarely are still found on archaeological sites and none was recognised at Vagnari. As far as we can tell, the Vagnari dolia defossa (Cat. 6.1– 6.2) resting in their mortar basins were globular or ovoid and with a flat bottom between c. 25–30cm in diameter. They correspond to Carrato’s Type 5 (‘globulaire’) and Type 11 (‘conique’) (Carrato 2017: 126, fig. 64; 309–11, figs. 195–97). Although no dolium was left in mortar basin 3012, the impression at the bottom of the basin suggests that the vessel in it had a flat base with a diameter of c. 30cm. Other fragments of dolium bases (Cat. 6.4–6.5, Plate 20.4–5), not found in the winery, range between 29 and 38cm in diameter. A different shape is the cylindrical dolium (Cat. 6.3) that was not found in a mortar basin (Figure 6.3). This survived in one large body sherd with an angular carinated edge at one end (whether the shoulder or lower body is hard to say), making the shape as a Carrato Type 4 (‘tronconique’) clearly identifiable (Carrato 2017: 126,

Figure 6.2 Detail of the relief in Figure 6.1, showing lidded dolia in the winery. Photo M. Carroll.

Figure 6.3 Remains of a tall, cylindrical dolium (Cat. 6.3). Photo M. Carroll.

83

The Making of a Roman Imperial Estate

Figure 6.4 Fragment of a ceramic lid from a dolium (Cat. 6.6). Photo M. Carroll.

fig. 64; 309, fig. 195). It had a body diameter of c. 1.20m. These are all shapes typical of dolia on Italian land sites.

litres at the Villa Regina at Boscoreale, and smaller and larger dolia in Regio I, insula 22 at Pompeii containing at least 133 litres and as much as 734 litres, to two types holding 480 litres or 790–1730 litres at San Giusto in Puglia (De Caro 1994: 41; Cheung and Tibbott 2020: 177; Pietropaolo 1998: 64). Some of that variability existed at Vagnari as well, but, given the fragmentary nature of the vessels, and the varying sizes of mortar basins, any attempt to pin down specific capacities must remain somewhat hypothetical.

It is difficult to calculate the capacity of the globular dolia at Vagnari, as none of them is intact and none of the mortar basins in which they were placed has survived to a height of more than 65cm. However, there are some useful measurement data. The bottom third of the globular dolium in basin 3038 (Cat. 6.1) had a maximum preserved internal body width of 96cm and it may originally have had an internal diameter of at least 1m or 1.10m higher up in the middle of the vessel. Mortar basin 3020 was large, with an internal diameter of 1.40m, suggesting that the globular dolium (Cat. 6.2) resting in it might have had an internal diameter of around 1.20m, given that the walls of the vessel can be as thick as 7cm. The fragmentary dolium lid (Cat. 6.6) with a diameter of 52cm suggests that at least one of the dolia had an internal rim diameter of c. 42cm, as the lids never quite reached to the outer edge of the vessel rim (Figure 6.4, Plate 20.2). The Vagnari dolia probably fall within the range of those from the Pisanella villa at Boscoreale that were 1.36m tall and had a maximum internal diameter of 1.26m or of those at the villa of N. Popidius Narcissus Maior in Scafati, as these had an internal rim diameter of 42cm and an internal body diameter of 1.10m (Pasqui 1897: 486–89; De Spagnolis 2002: cat. no. 179–80). There is great variability in the capacity of dolia defossa in wineries throughout Italy, ranging, for example, from those containing 392.50

Much larger than any of these dolia defossa that were used for the fermentation and storage of wine in the cella vinaria at Vagnari or on other land sites were the enormous dolia manufactured to transport Italian wine in bulk on specially constructed tanker ships around the western Mediterranean from about the mid-1st century BC to the middle of the 2nd century AD (Gianfrotta 1989; Gianfrotta 1998; Rendini 1991; Heslin 2011; Rice 2016: 176–80; Cibecchini 2020: 187–88). They typically are of globular, cylindrical or pear shape (Carrato 2017: 126, fig. 64; 304–08, figs. 190–94 — Types 9, 11, 12, 19). Dolia found on wrecks, such as the Diano Marina, La Garoupe, Gran Ribaud D, Ladispoli, and La Giraglia, along the NW coast of Italy, the south coast of France, and the islands in between, had capacities between 1,500 and 3,000 litres, and they are the true giants of the Roman dolia family (Pallarés 1995/1996; Gianfrotta and Hesnard 1987; Hesnard et al. 1988; Dell’Amico and Pallarés 2005;

84

Carroll et al.: 6. The Dolia Defossa and Viticulture at Vagnari

Figure 6.5 Lower third of dolium 3038 (Cat. 6.1) in situ with numerous cracks. Dark layers of the pine pitch used to coat the interior surface are still visible here. Photo M. Carroll.

Figure 6.6 Finger indentations on the top of the vertical wall of the dolium (Cat. 6.1), an indication of having been coil built. Photo M. Carroll.

85

The Making of a Roman Imperial Estate Carre 1993; Marlier and Sibella 2002; Sciallano and Marlier 2008; Broekaert 2013; Cibecchini 2020).

lead-alloy (Carrato 2017: 184–86; Cheung and Tibbott 2020: 179–84). Cato (On Agriculture 39.1), writing in the mid-second century BC, specifically recommended lead as an ideal material with which to mend wine vats. We found no lead repairs in place in any of the dolium sections or fragments at Vagnari, possibly because the lead was removed when the dolia went out of use, to be re-melted for another purpose. It was clear, however, that the dolium in mortar basin 3038 (Cat. 6.1) had not been repaired with clamps, suggesting that it might have been abandoned once it began to leak.

In the absence of any workshop data recorded in the Roman period, more recent production of dolia can help to illuminate how work-intensive and timeconsuming the making of these storage vessels is. There is pertinent ethnographic data on the production of enormous dolia in Spain (tinajas) and in Greece (pithoi) in the 19th and 20th centuries (Romero Vidal and Cabasa Calpe 1999; Gouin and Vogt 2002; Carrato 2017: 135–52). These vessels could take between 10 days and two months to dry in advance of firing in specially built kilns at 900–1000°C for eight to 24 hours, depending on the size of the vessels (Carrato 2017: 147–49). In the Artenova workshop near Florence, where dolia are being reproduced for modern buyers, each dolium with a 500-litre capacity is coil-built slowly, one coil at a time from the bottom up, over a period of 15–20 days, and another month is then needed to dry them in preparation for firing at 1000°C in large kilns for three days (Artenova 2021). Recent experimentation in the making of a dolium in France involved the drying of the complete vessel for five months before it was fired in the kiln (Caillaud 2020: 142–43). This puts into perspective how long it must have taken to produce dolia in antiquity.

At Vagnari, the fabric of the dolia defossa was distinctively different from that present in all the other ceramics from the site, as it was a distinctive orangeyred with black, glassy particles and reddish-brown grit fragments (Figure 6.7). The surviving lid fragment (Cat. 6.6), on the other hand, is made of a fairly fine buff clay, suggesting it was made somewhere else. It is not unusual that dolia and lids might be made of different clays and in different workshops, as they are in Gallia Narbonensis, where some analysed lids were made in Rome, but the vessels were local (Carratto 2017: 173; Capelli and Cibecchini 2020: 236–37). The recognition that it is important to conduct fabric analyses on dolia retrieved in archaeological excavations, whether on land or in the sea, is a fairly recent phenomenon (Capelli and Cibecchini 2020). The vast majority of the gigantic transport dolia found in shipwrecks in the western Mediterranean that have been subjected to a fabric analysis are now known to originate in the Tyrrhenian coastal zone of western Italy, between southern Etruria and northern Campania (Capelli and Cibecchini 2020: 228). An invaluable opportunity presented itself at Vagnari, therefore, to conduct an archaeometric analysis of the clay to ascertain whether the dolia defossa were made at Vagnari or anywhere else in the region or if they were sourced from figlinae located further away. This research was carried out at the Universities of Palermo and Calabria, and a synopsis of the results is presented below.

There are clues on the surviving Roman dolia themselves as to the methods with which they were built. A recent examination of dolia defossa in Regio I, insula 22, in Pompeii has confirmed that these vessels were coil-built, the transition points where the coils met still being easily visible (Cheung and Tibbott 2020: 178–79, figs. 3–4). Such joints often cracked when the dolium was drying or during firing or at a later stage during use. Details on the intact lower third of a dolium defossum still in situ in mortar basin 3038 (Cat. 6.1) at Vagnari suggest that this one also was coil built (Figure 6.5). The top vertical edge of the remaining part of the dolium is slightly undulating, preserving the indentations made by the fingers of the potter to create a surface onto which the next coil above would better adhere (Figure 6.6). A similar undulating break between two coils has been noted on Roman dolia in Gallia Narbonensis (Caratto 2017: 139, fig. 70). The vertical cracks radiating from the base of this Vagnari dolium are possibly stress fractures that started in firing and might not have been noticeable during the use-life of the vessel, but worsened with time and/or exposure to the fermenting juice in the dolium. It is clear that the joint between the wall and the flat base of this dolium was another weak point, as it is all around it that a crack runs. Cracks in storage vessels often necessitated repairs consisting of tenons, clamps, plugs or staples of lead or

Figure 6.7 Fragment of a dolium with black volcanic temper and reddish-brown grit. Photo M. Carroll.

86

Carroll et al.: 6. The Dolia Defossa and Viticulture at Vagnari

Figure 6.8 Map of central and southern Italy showing the Roman Magmatic Province and the Ernici-Roccamonfina Magmatic Province, as well as various relevant sites. Volcanoes are marked as red triangles. Plan by Helen Goodchild, with data from Jarvis et al. 2008 and the Geoportale Nazionale.

fragments, trachytic rock fragments, clinopyroxene, sanidine, biotite, amphibole, garnet, olivine, and plagioclase, strongly suggests that the clay came from the Tyrrhenian coast of Latium and Campania which is affected by geologically recent volcanic activity (Pliocene-Pleistocene) composed of deposits of potassic and ultrapotassic rocks. To obtain further confirmation, and to restrict as much as possible the potential areas of provenance, chemical microanalysis was performed on the clinopyroxene contained in the dolia pastes (major elements using SEM-EDS and trace elements using ICPMS with laser ablation). The obtained results strongly support the hypothesis of provenance from the wide volcanic region which covers most of the regional territory of Latium, from the Vulsini Mountains and Bolsena Lake areas (Tuscan border) southwards towards Rome, in one zone, and as far south as the area around the Garigliano river basin (Latium-Campania border), in the other zone (Figure 6.8). This territory constitutes the Roman Magmatic Province and the ErniciRoccamonfina Magmatic Province, and it is markedly characterised by recent Quaternary volcanism with prevalent silica undersaturated and ultrapotassic eruptive products belonging to the HKS ultrapotassic volcanic series (Peccerillo 2017: 81–124, 125–43).

Fabric analysis — Where were the dolia defossa made? Ten samples of dolia were subjected to mineralogicalpetrographic analysis (optical microscopy under transmitted polarised light on thin sections) and chemical analysis (ICP-MS coupled with ICP-OES for a total of 52 determined chemical elements). The detailed results of this research can be found in Montana et al. 2021, so we restrict ourselves here to a summary. The Vagnari dolia have homogeneous textural and compositional characteristics which are very different from those verified in ceramic building material samples of local production (see Chapter 13). This result suggests that they were imported from other regions. In fact, they present a medium-high frequency of aplastic inclusions (20–30% area) with an evident bimodal sorting that reveals the prevalence of the coarse and very coarse sand classes (0.5–2 mm) on the one hand, and of the very fine ones (0.06–0.125 mm) on the other. From the compositional point of view, the presence of abundant inclusions of a volcanic nature is peculiar (monomineralic grains and lithic fragments). The composition of the volcanic inclusions in the dolia found at Vagnari, such as leucite-bearing rock 87

The Making of a Roman Imperial Estate The dolia defossa, therefore, were made of clay that came from the hinterlands of the Tyrrhenian coast of Latium and Campania, with Rome (Roman Magmatic Province) or Minturnae (Ernici-Roccamonfina Province) as key sites. The better match for the temper and clay of the Vagnari dolia is the region around Rome, in the Tiber valley, but the hinterland of Minturnae as the source of the materials cannot be ruled out completely. We can definitely reject the attribution of the volcanic products to Monte Vulture 80km northwest of Vagnari, however, as these are easily distinguishable from those evident in the Vagnari dolia. They can also be distinguished clearly from the Campanian volcanic products from the Phlegraean Fields, Ischia, and Somma Vesuviana in southwest Italy.

dolia defossa for wine should be re-lined with pitch 40 days prior to the grape harvest, and the 10th-century Geoponica (6.4), which drew on earlier Roman books on agricultural pursuits, advised the renewal of the pitch lining every year in July. Elsewhere, agricultural tasks collated in a calendar inscribed on a marble block of the 1st century AD in Rome included the re-lining with pitch in the month of September (CIL 6.2305). Tartaric acid is present in large amounts in grapes, and consequently in wine, but is not nearly so common in other plants. The survival of tartaric acid in ceramic vessels, therefore, is taken as an indicator of the presence of wine. In addition, syringic acid is a derivative of the red colour pigments in red wine, and this is taken as an indication of red wine rather than white. To detect both tartaric and syringic acid, twelve samples of dolia fragments were selected at the University of Bradford for High Performance Liquid Chromatography-Mass Spectrometry (HPLC-MSMS) analysis. Four fragments coated with the black substance were subjected to solvent extraction and Gas Chromatography-Mass Spectrometry (GCMS) analysis to test whether this was pitch and to determine whether other lipids are present.

Residue Analysis — What did the dolia contain? It was evident that the Vagnari dolia had been re-lined multiple times internally with coatings of a blackish substance which we were certain was pitch, and this feature was a reliable indication that the original contents of the dolia had been wine (Figure 6.5). Wine vats needed to be cleaned regularly, and even fumigated, to avoid contamination of the new wine with which they were filled annually. Pitch was the normal internal surface treatment for vats containing wine, unlike oil or grain dolia which could be coated with wax or gum (Peña 2007: 212–13). The Roman agrarian writer Columella (On Agriculture 12.18.5–7) recommended that

Sample

Vessel

3020

dolium pitch

3028

dolium

3032

dolium

3038 *

dolium

3038p

dolium pitch

4028 *

dolium

4029 * 4030 *

dolium

Sample preparation — lipid analysis Residues were separately removed from the interior and exterior surface of each sherd by scraping with a metal Degraded Pinaceae Pinaceae pitch No lipids (dehydroabietic acid, (retene) 7-oxodehydroabietic acid)

Surface sampled

Trace fatty acids (C16:0 and C18:0)

interior

Y

Y

interior

Y

Y

Y

exterior exterior interior

Y Y

Trace B-sitosterol monoacylglycerols

Y

Y

Y

Y

Y

exterior

Y

exterior

interior

interior

interior

exterior

cylindrical interior dolium exterior

Y

Y

Y

Table 6.1 Summary of analytical results, with lipids and selected sources identified. * = samples also analysed for polyphenols.

88

Y

Y Y

exterior

Y

Y

Y

interior

exterior

Y

Y

Y

Y

Carroll et al.: 6. The Dolia Defossa and Viticulture at Vagnari

spatula. The samples analysed are listed in Table 6.1. The resultant powders were extracted with three aliquots of ~2mL DCM:MeOH (dichloromethane:methanol 2:1, v/v), with ultrasonication for 5 minutes. followed by centrifugation (5 min 2000 rpm). Excess BSTFA (N,Obis(trimethylsilyl)trifluoroacetamide) with 1% TMCS (trimethylchlorosilane) was added to derivatise the sample which was warmed at 70°C for one hour. Excess derivatising agent was removed under a stream of nitrogen. The samples were diluted in DCM for analysis by GC-MS. A method blank was prepared and analysed alongside the samples.

acid (m/z 169/125), p-coumaric acid (m/z 162.9/119.3), myricetin (m/z 317/151), syringic acid (m/z 197/167 and m/z 197/182) and tartaric acid (m/z 149/87). Results of the residue analysis Eight samples (each with separate interior and exterior samples) were analysed by GCMS targeting the lipids (Table 6.1). No lipids were recovered from sample 4029 (both surfaces) or the exterior surfaces of 4028 and 4030. Trace fatty acids (unsaturated with 16 and 18 carbon chains) were the most common lipid recovered, however their abundances were extremely low in all the samples. The presence of monoacylglycerols in some of these samples could indicate the degradation of oils or fats, but these components are also at low concentrations. As fatty acids are ubiquitous, and commonly found as contaminants, it is not possible to comfortably attribute these to being part of the original dolia contents. The presence of β-sitosterol is indicative of plant material (this is a sterol only found in plants), but this cannot be used for any further source identification.

Sample preparation — polyphenol analysis Visible residues were separately removed from the interior and exterior surface of selected sherds by scraping with a metal spatula. The samples analysed were: 4029, 3038, 4030 and 4028 (both interior and exterior surfaces). The resultant samples were extracted with ~2mL of methanol, with ultrasonication for 15 minutes. Undissolved material was removed from the samples by filtration using a 0.22μm filter. The samples were then analysed by HPLC-MSMS.

Four samples (each with the interior and exterior separately sampled) were analysed by HPLCMS targeting polyphenols, including syringic and tartaric acids which are biomarkers for wine (Table 6.1) (Guasch-Jané et al. 2004; Stern et al. 2008). No polyphenols were found, and, therefore, there is no molecular identification of wine in these samples.

Instrumental (GC-MS) — lipid analysis Analysis was carried out by combined gas chromatography-mass spectrometry (GC-MS) using an Agilent 7890A Series GC connected to an 5975C Inert XL mass selective detector. The splitless injector and interface were maintained at 300°C and 340°C respectively. Helium was the carrier gas at constant flow. The temperature of the oven was programmed from 50°C (2 min) to 350°C (10 min) at 10°C/min. The GC was fitted with a 15m × 0.25mm, 0.25µm film thickness HP-5MS 5% Phenyl Methyl Siloxane phase fused silica column (Agilent J&W). The column was directly inserted into the ion source where electron impact (EI) spectra were obtained at 70 eV with full scan from m/z 50 to 800.

Five samples (3020, 3032, 3038, 3038p and 4030) contained a group of diterpenoids (7-oxodehydroabietic acid and dehydroabietic acid) which are indicative of a degraded Pinaceae resin. This was recovered from the interior surfaces only and confirms the use of this material as a sealing agent to either waterproof the vessels or prevent migration of the contents through the vessel walls. Two of these samples (3028 and 3038p) also contained the molecule retene, a biomarker for Pinaceae pitch (Evershed et al. 1985). A pitch is produced after extensive heating of the natural bled resin, so this is likely to be a deliberate production and use of pitch as a sealant on the interior surfaces of these dolia. Romanus et al. (2009) suggest that a pitch layer is effective at preventing the absorption of olive oil into the ceramic vessel, but it does allow the migration of polyphenols from wine. However, this was not apparent at Vagnari.

Instrumental (HPLC-MSMS) — polyphenol analysis Five polyphenols were target for analysis: caffeic acid, gallic acid, p-coumaric acid, syringic acid and tartaric acid. An XBridge Phenyl 3.5μm, 150mm × 4.6mm HPLC column was the stationary phase held at constant at 40°C. An injection volume of 10μL was used. The mobile phase was at a constant flow rate of 1.0 mL/min. Solvent A comprised of 0.1% formic acid in deionised water while solvent B consisted of HPLC grade methanol. The gradient elution program for the 90 minutes run was 90% A, 10% B. A photodiode array detector was set at 210nm with a final wavelength at 600nm. Resolution of 1.2nm, sampling rate of 1 spectra/sec. In addition, seven mass channels in MRM mode were used to identify the following: caffeic acid (m/z 179/135), gallic

In summary, absolute scientific proof that the Vagnari dolia defossa held wine could not be produced through an examination of the sherds. The liquid contents of the dolia had not left any trace in the vessel walls, but that is because the pine pitch applied and re-applied to the interior surfaces did its job in sealing the surfaces and 89

The Making of a Roman Imperial Estate stopping absorption of the wine into the clay extremely well. The remains of this substance used traditionally in the Roman treatment of wine storage vats, however, are a solid indication that the Vagnari dolia defossa were, indeed, used for the vintages of the imperial estate.

Since the early 1st century BC, dolia defossa were part of the repertoire of Minturnaean figlinae, as indicated by a vessel of that period from Minturnae bearing a stamp of the slave Dem…. Coionius (Nonnis 2015: 195; Gregori and Nonnis 2013: 92, fig. 7). And the Codonii and Pirani figlinae at Minturnae made dolia defossa that were acquired by villa owners north of Rome (Lazzeretti 1998). We have considerable information about the production sites and kilns used for making amphorae in the territory of Sinuessa and Minturnae, but no remains of dolium workshops or kilns have yet been located archaeologically (Guidobaldi and Pesando 1989: 44–45; Arthur 1991a: 155–57; Chiosi and Gasperetti 1994; Bellini and Trigona 2013: 268; Gregori and Nonnis 2013: 91). Some of the families operating out of Minturnae were particularly active in the bulk wine trade, two of the main ones being the Piranus and Cahius families. They not only made normal storage dolia defossa and the gigantic transport dolia, but also potentially built the ships to transport them (Johnson 1933: 126–28; Bellini and Trigona 2013; Gregori and Nonnis 2013; Heslin 2011: 165–66; Carre 2020).

Conclusions Normally, in regions where Roman wine and olive oil were produced in significant quantities, the kilns and workshops for dolia were located at or near the harvesting sites and estates. We can see this at Giancola near Brindisi in Puglia, along the Guadalquivir river in southern Spain, and at the villa of Saint-Bézard at Aspiran in the Hérault region of southern Gaul (Manacorda and Pallecchi 2012; Desbonnets et al. 2017; Carrato 2017: 143–47, figs. 73–77; 380–90, fig. 272). The fabric analysis of a number of dolia defossa from the Roman villa at Settefinestre and other sites in Rome also points to local or regional production (Capelli and Cibecchini 2020: 235). A fabric analysis of two dolium samples from Brindisi on the Adriatic coast suggests an Adriatic origin, possibly from the opposite coast (Capelli and Cibecchini 2020: 235). However, in contrast to this normal situation, the imperial administration responsible for the management and development of the estate at Vagnari, did not procure dolia defossa for the new winery in the 2nd century from any local or regional providers. Instead, figlinae around Rome or around Minturnae are the likely source of the Vagnari dolia, although the Tiber valley and Rome are a better match for the fabric of the vessels (Montana et al. 2021).

Unfortunately, no stamps have survived on the dolium fragments at Vagnari, so we cannot connect any known figlinae in the Roman Magmatic Province or the ErniciRoccamonfina Province with these vessels. Only the clay used to make the dolia defossa at Vagnari points to these areas. A fabric analysis of dolia defossa found at the villa at S. Giovanni on the island of Elba also only narrows down their place of manufacture to these two regions (Manca et al. 2016).

In Latium, especially around the capital city of Rome, heavy ceramics, including dolia, were manufactured in vast quantities in production centres owned by the Roman elite. Kilns at Scoppieto di Baschi, for example, made dolia and mortaria using clay deposits from the Vulsini volcanic zone, according to the archaeometric analysis of them (Nicoletta 2007; Comodi et al. 2007). Many of these brickyards around Rome gradually were transferred to imperial ownership by bequest or confiscation, and by the 2nd century AD the majority were owned and run by this group (Gasperoni 2003; Lazzeretti and Pallecchi 2005; Gliozzo and Filippi 2005; Lo Cascio 2005: 97; Maiuro 2012: 381–98). Imperial properties of the late 1st and 2nd centuries AD are known northeast of Gaeta and Formia at the foot of the Ernici volcano too, but the economic nature of them is unclear (Maiuro 2012: 272–75). The estates in the hinterland of the Roman towns of Minturnae, Suessa Aurunca, and Sinuessa, located in the Garigliano river basin and near the Roccamonfina volcano, were active in the production of wines, and it is certain that some estates around Minturnae were part of the imperial patrimony (Arthur 1991: 63–69, 81–83; Maiuro 2012: 218–19, 275–78; Cascella 2016; Woodhull 2018/2019).

We might expect the emperor to supply his estate at Vagnari with products locally made, as private vintners in Apulia did, and to conform with the normal market scenario in terms of balancing investment and expenditure against profit. This did not happen. Using suppliers for essential (and specialist) equipment from the other side of the peninsula makes little sense in this context and it seems excessive and unnecessary, unless there was a particular reason to do so. That reason might be that the imperial administration drew on figlinae and suppliers that belonged to the patrimonium Caesaris, as their outputs would be property of the emperor. The act of outfitting the winery this way might be explained as a decision of the imperial administration to move equipment from one property in the emperor’s possession in western Italy to another in eastern Italy. Transfers of property assets from one imperial estate to another might have represented a kind of economy on the part of the administration who perhaps sought to avoid spending cash outside of the patrimonium. It is possible that an imperial estate like Vagnari, that primarily produced grain, exchanged this with properties elsewhere in western Italy that produced other needed commodities, such as dolia defossa. 90

Carroll et al.: 6. The Dolia Defossa and Viticulture at Vagnari

Given the difficulty of transporting such bulky goods by land over the Apennines, it seems likely that the dolia destined for Vagnari were shipped by sea around the tip of the Italian peninsula (Figure 6.8), as water routes would have been preferable in general (Tchernia 2016, 90–93). They could have sailed from Ostia/Portus or Minturnae, either to a port on the Adriatic coast, or to the coast of the Ionian Sea, then making use of the rivers at least to a certain point in the interior, or even to the harbour at Tarentum, the starting point of the Via Appia; the last leg of the journey to Vagnari would have been overland.

Cat. 6.3 (Figure 6.3). Cylindrical dolium, with carinated shoulder(?). Context 4030. Body Dm 1.20m. Clay dark orangey-red, with grit and volcanic temper. Date: Found in Phase 6 context, but belonging to Phase 3. Cat. 6.4 (Plate 20.4). Dolium fragment of a vessel base. Context 4028. Base Dm 29cm. Reddish clay with grit and black volcanic temper. Date: Found in Phase 6 context, but belonging to Phase 3. Cat. 6.5 (Plate 20.5). Dolium fragment of a vessel base. Context 4029. Base Dm 38cm. Clay orangey-red with black volcanic temper. Date: Found in Phase 6 context, but belonging to Phase 3.

Catalogue Cat. 6.1 (Figure 6.5). Dolium, bottom third of the vessel. Context 3038, in mortar basin. Base Dm c. 25–30cm. Clay orangey-red, with grit and black volcanic temper. Date: Phase 3.

Cat. 6.6 (Figure 6.4, Plate 20.2). Fragment of a dolium lid fragment. Context 4009, re-used as building material. T 3cm; Dm 52cm. Fabric fairly fine buff clay. Date: Phase 3, built into wall of Phase 5.

Cat. 6.2 (Figure 4.21). Dolium, large fragments of the vessel base. Context 3020, in mortar basin 3020. Base Dm c. 25–30cm at bottom. Clay orangey-red, with grit and black volcanic temper. Date: Phase 3.

91

CHAPTER 7

Vessel and Window Glass Camilla Bertini and Victoria Lucas Only two fragments of vessel glass were recovered from Phase 3 contexts; these were a rim (Cat. 7.VF7) and a base fragment (Cat. 7.VF13, Plate 21.6, Figure 7.1.1) both from deposit 5017 directly below the floor of the winery, dated to between AD 20–100. Neither fragment was diagnostic, however the relatively tight dating of the deposit, the fact the glass is colourless and blown, and its association with the foundation of the winery place these sherds firmly within the 1st century AD. No fragments were recovered from contexts identified as belonging to Phase 4.

Vessel Glass Introduction A total of 132 fragments of vessel glass was recovered from excavations at Vagnari between 2015–2018. A catalogue of these fragments has been compiled with each fragment (or associated group of fragments) and given a unique identifier created by pairing up two elements: VF (vessel glass fragment) and a consecutive number (1, 2, etc.). Fragments are arranged in the catalogue by element (rims, handles, bases, decorated bodies, undecorated bodies). Entries comprise description of the fragment including an identification where the fragments are diagnostic; a record of colour and dimensions (where available); the small finds number; context and phase; and, where relevant, a reference to any associated fragments recorded elsewhere in the catalogue. The complete catalogue including both the vessel and window glass fragments can be found at the end of this chapter. Drawings of illustrative or diagnostic fragments have been included in Plates 21–22.

A fairly significant number of vessel glass fragments was recovered from context 4005, with a total of 28 fragments deriving from this context, the largest number of fragments from a single context at the site. Context 4005 was a soil deposit on top of the mortar floor of the winery. As it was only encountered in a sondage, its exact date and relationships are uncertain, but it most likely belongs to Phase 5 (the early 3rd century AD). None of these fragments are diagnostic or belong to diagnostic elements (handles, rims, bases etc.), and there are no recorded decorative elements on any of the fragments. All the fragments from 4005 are small in size, with the largest recorded fragment being just 5.3 × 2.9cm, and the smallest recorded fragment being 1cm × 0.9cm. There is a number of fragments with no recorded dimensions, but they can be assumed to be